eCAADe2016 Volume1 Lowres

eCAADe 2016
Complexity & Simplicity

Volume 1
Volume 1 - eCAADe 34 | 1
Editors
Aulikki Herneoja
Toni Österlund
Piia Markkanen
Oulu School of Architecture
University of Oulu
1st Edition, August 2016
Complexity & Simplicity - Proceedings of the 34th International Conference on Education and
Research in Computer Aided Architectural Design in Europe, Oulu, Finland, 24.-26. August
2016, Volume 1. Edited by Aulikki Herneoja, Toni Österlund and Piia Markkanen. Brussels: Ed-
ucation and Research in Computer Aided Architectural Design in Europe; Oulu: Oulu School
of Architecture, University of Oulu.
ISBN
9789491207105
978-94-91207-10-5
Copyright © 2016
Publisher: eCAADe (Education and Research in Computer Aided Architectural Design in Eu-
rope) and Oulu School of Architecture, University of Oulu.
Cover Design: Piia Markkanen and Toni Österlund
All rights reserved. Nothing from this publication may be produced, stored in computerised
system or published in any form or in any manner, including electronic, mechanical, repro-
graphic or photographic, without prior written permission from the publisher.
2 | eCAADe 34 - Volume 1
eCAADe 2016

Volume 1
Proceedings of the 34th International Conference on Education and Research in
Computer Aided Architectural Design in Europe
24.-26. August 2016

Oulu, Finland
Oulu School of Architecture
University of Oulu
Edited by
Aulikki Herneoja
Toni Österlund
Piia Markkanen
Volume 1 - eCAADe 34 | 3
Sponsors of the eCAADe 2016 Conference
Alfred Kordelin Foundation
4 | eCAADe 34 - Volume 1
Theme

The first volume of the conference proceedings of the 34th eCAADe conference, 24.-26. Au-
gust 2016, at Oulu School of Architecture, University of Oulu, Oulu, Finland contains 81 papers
grouped under 14 sub-themes. Both volumes contain altogether 157 accepted papers.
The theme of the 34th eCAADe Conference is Complexity & Simplicity. We invited the eCAADe
community to address the multifaceted notions of complexity and simplicity, which are en-
countered in architectural design processes. Approaches discussing the theme from the per-
spective of computer aided design education; design processes and methods; design tool de-
velopments; and novel design applications, as well as real world experiments and case studies
were welcomed. What is the role of complexity or simplicity as part of the design process?
Does the use of complex design methods offer simplicity to the design process itself? Is it
possible to design complexity with simple methods? Does the use of computation in design
necessitate complexity or offer means to control it?
Recent development in digital technologies and digital design tools enable us to address com-
plex situations in architectural environments, ranging in scale from structures and buildings
to urban contexts. We often expect technology to better help us manage the complexity of
life, to simplify our daily lives and tasks. However, these developments also raise the question
of whether design technologies encourage complexity at the expense of simplicity in both
the design process and lived environments. Does computation cause complexity? Or does it
enable simplicity?
In addition to the accepted papers, the first volume contains Keynote Papers, including keynote
speakers contribution concerning the themes of their keynote lectures and the Workshop Con-
tributions, including the papers summarizing the contents of workshops given.
Theme - Volume 1 - eCAADe 34 | 5

All the papers of these proceedings will be accessible via CuminCAD - Cumulative Index of
Computer Aided Architectural Design, http://cumincad.scix.net
eCAADe 2016 Acting Conference Chairs

Aulikki Herneoja, Toni Österlund, Piia Markkanen
6 | eCAADe 34 - Theme - Volume 1

Acknowledgements
Welcome to eCAADe 2016, the 34th eCAADe conference, in Oulu, Finland. It has been exactly
fifteen years since the previous eCAADe conference was held in Finland, hosted by Helsinki
University of Technology (HUT) in 2001. HUT was also in charge of organizing the 3rd eCAADe
conference in 1984. We are now very happy to welcome you back to Finland, this time to the
most northern architect school in the world, the Oulu School of Architecture, University of
Oulu.
The original idea of bringing the eCAADe conference to Oulu dates back to September
2012 after the eCAADe conference in Prague. Officially, the eCAADe Council granted us the
permission to organize the 34th eCAADe conference in Oulu in March 2013. Over the last four
years several people have helped us to make this conference happen. We thank especially
the former Dean of Oulu School of Architecture, now the Vice Rector for Education, University
of Oulu, Professor Helka-Liisa Hentilä and the present Dean of Oulu School of Architecture,
Professor Rainer Mahlamäki for their positive and supportive attitude.
During the lengthy process of organizing the eCAADe 2016 we have had the privilege to
experience the supportive and kind atmosphere of eCAADe Council, whose members have
helped us with all aspects of the organizing. We warmly thank both Presidents Johan Verbeke
(until 2016) and Joachim Kieferle (since 2016) for encouragement and unswerving support
during the organizing of this conference. We most warmly thank Bob Martens, the liaison with
the conference host and the previous conference organizer, for always kindly and patiently
guiding us through the multistage arrangement process of the conference. Also, we wish to
thank all the other previous conference organizers, Henri Achten, Rudi Stouffs and Emine Mine
Thompson, for sharing their experience and knowledge without hesitation.
Quality control is an emerging issue concerning the publishing of a conference proceed-
ings. We are grateful for eCAADe Council for being able to use the OpenConf system through-
out the publishing process. Authors uploaded their extended abstracts (length of 1000 to
1500 words and one optional image) for the double blind peer review process. Each abstract
Acknowledgements - Volume 1 - eCAADe 34 | 7

was evaluated by three reviewers. With the help of the OpenConf system we could, for exam-
ple, easily ensure that none of the reviewers came from the same institution as the authors.
Altogether, we received 241 abstracts from authors from 44 different countries. After the peer
review process, 165 papers were accepted for full paper submission. In the end, altogether
157 papers were presented in eCAADe 2016 and published in the proceedings. We are very
grateful for all the 98 reviewers from 29 different countries (see the List of Reviewers) for their
constructive and thorough comments for each author. We also continued the practice started
in eCAADe 2015 conference in Vienna of having all the session chairs to give prospective com-
ments of the papers to evoke the discourse at early stage between the author and session
chair for the 27 sessions of the conference. All the session chairs also participated the peer
review process of the extended abstracts. We owe the session chairs great gratitude for their
commitment and long term contribution to the process until the final paper presentations.
Parallel to these prospective comments editorial team gave comments to the authors too. We
thank and congratulate all authors for their hard work and support on using the ProceeDings
tool and finalizing their full papers carefully in time. In this last phase of editing the full pa-
pers we may not thank enough Gabriel Wurzer, the Master of the ProceeDings, who patiently
and relentlessly worked together with us and enabled us to successfully produce high quality
proceedings in time.
We owe great thanks to the keynote speakers; Professor Branko Kolarevic (Chair in Inte-
grated Design and co-directs the Laboratory for Integrative Design (LID), University of Calgary;
Canada), Professor Mette Ramsgaard Thomsen (Chair in Architecture and Digital Technologies,
Head of CITA, Centre for Information Technology and Architecture, Royal Academy of Fine Arts,
School of Architecture, Design and Conservation (KADK), Denmark) and Professor Toni Kotnik
(Chair in Design of Structures at Department of Architecture in Aalto ARTS, Finland), and their con-
tribution of writing the keynote papers concerning their keynote lecture themes.
The Panel Session is a traditional Thursday afternoon event in eCAADe conferences. This
year the panelist were Finnish colleagues, the top level practicing architects. The visionary
discussion was moderated by Director Hanna Harris (Architecture Information Centre Finland).
The panelists were; Professor Rainer Mahlamäki (Lahdelma & Mahlamäki Architects), Architect
8 | eCAADe 34 - Acknowledgements - Volume 1

Antti Nousjoki (ALA Architects Ltd) and Architect Toni Österlund (Geometrics Ltd). We owe great
thanks to the moderator and the panelists for their rewarding presentations and discussions.
Workshops are an integral part of eCAADe conferences and eCAADe 2016 made no excep-
tion. We thank all the workshop organizers for the event and for their contribution of short
papers (non-peer reviewed) about the contents of the workshop. We are also grateful for Wolf-
gang Dokonal and the eCAADe council for organizing the traditional PhD workshop for young
researchers and supporting the participants with a subsidy for traveling to Oulu.
In addition to printed and digital conference proceedings, the team supervised by Wolf-
gang Dokonal provides the eCAADe community video streams of the paper presentations.
Harri Hämeenkorpi and Ari Orajärvi from University of Oulu were locally in charge of recording
and video streaming with the help of architecture students from Oulu School of Architecture.
Great thanks for the whole team.
We would like to express our gratitude for the administrative help in organizing this con-
ference. eCAADe council and especially Nele De Meyere, has provided us valuable input and
lessons learned from past conferences. We have also had Konffa Ltd as highly-appreciate part-
ner for managing the conference services, ranging from the registration process to the actual
on-site registration services. A big thank goes you to Pia Hanski, Hanna Hentilä and Jenni
Ylisirniö. We will also thank Erweko Ltd and Mikko Träskelin for smooth process and high qual-
ity printing of the proceedings.
Organizing an international conference of this scale requires financial help. We are in-
debted for our sponsors – The Federation of Finnish Learned Societies, Alfred Kordelin General
Progress and Education Fund, Autodesk Inc., Bentley Systems International Ltd., and the Oulu
School of Architecture, University of Oulu. Their sponsorship enabled us to fulfill the quality
standards of eCAADe conference. As a special form of sponsorship, the members of the local
conference staff donated their time to help prepare and organize this conference. Thank you
all for helping us out.
eCAADe 2016 Acting Conference Chairs

Aulikki Herneoja, Toni Österlund, Piia Markkanen
Acknowledgements - Volume 1 - eCAADe 34 | 9

List of Reviewers
Persons marked with (*) acted as session chair, in addition to their role as reviewer
Sherif Abdelmohsen - American University in Antonio Fioravanti* - Sapienza University of Rome,
Cairo, Egypt Italy
Henri Achten* - Czech Technical University in Pia Fricker* - ETH Zurich - Swiss Federal Institute of
Prague, Czech Republic Technology in Zurich, Switzerland
Asterios Agkathidis - University of Liverpool, Harald Gatermann - Bochum University of Applied
United Kingdom Sciences, Germany
Aleksander Asanowicz - Bialystok University of Evelyn Gavrilou - University of Thessaly,
Technology, Poland Department of Architecture, Greece
Daniel Baerlecken - Georgia Institute of David Gerber* - University of Southern California,
Technology, United States United States
Thomas Grasl - SWAP Architects, Austria
Guenter Barczik - Erfurt School of Architecture,
Germany Aulikki Herneoja* - Oulu School of Architecture,
University of Oulu, Finland
Jakob Beetz - Eindhoven University of Technology,
Netherlands Christiane M. Herr - Xi'an Jiaotong-Liverpool
Anand Bhatt - Architexturez Imprints, India University, China
Stefan Boeykens - KU Leuven - University of Jia Hu - Philips Research North America, United
States
Leuven, Belgium Patrick Janssen* - National University of
Johannes Braumann - Association for Robots in Singapore, Singapore
Architecture & University for Arts and Design
Linz, Austria Anja Jutraz - Institute CoLab, Slovenia
Peter Buš - Faculty of Architecture, Czech Technical Matevz Juvancic - University of Ljubljana, Faculty
University in Prague, Czech Republic of Architecture, Slovenia
Anetta Kepczynska-Walczak* - Institute of
Gulen Cagdas - Istanbul Technical University, Architecture and Urban Planning, Lodz
Faculty of Architecture, Turkey University of Technology, Poland
Gabriela Celani - School of Civil Engineering, Joachim B. Kieferle* - Hochschule RheinMain,
Architecture and Urban Design at the Germany
University of Campinas, Brazil Kristina Kiesel - TU Wien, Austria
Tomo Cerovsek* - University of Ljubljana, Slovenia Arto Kiviniemi - University of Liverpool, United
Angelos Chronis - Institute for Advanced Kingdom
Architecture of Catalonia, Spain Erik Kjems - Aalborg University, Denmark
Birgul Colakoglu* - İstanbul Technical University, Robert Klinc - University of Ljubljana, Slovenia
Turkey Michael Knight - University of Liverpool, United
Mahesh Daas - University of Kansas, United States Kingdom
Wolfgang Dokonal* - University of Technology Tuba Kocaturk - University of Liverpool, School of
Graz, Austria Architecture, United Kingdom
Dirk Donath - Bauhaus University, Germany Volker Koch - Karlsruhe Institute of Technology,
Jose Duarte - Faculty of Architecture, University of Germany
Lisbon, Portugal Branko Kolarevic - University of Calgary, Canada
Athanassios Economou - Georgia Insitute of Reinhard König - ETH Zurich - Swiss Federal
Technology, United States Institute of Technology in Zurich, Switzerland
List of Reviewers - Volume 1 - eCAADe 34 | 11

Jose Kos - Federal University of Santa Catarinaa, Henrika Pihlajaniemi* - Oulu School of
Brazil Architecture, University of Oulu, Finland
Krzysztof Koszewski - Faculty of Architecture, Ulrich Pont - TU Wien, Department of Building
Warsaw University of Technology, Poland Physics and Building Ecology, Austria
Toni Kotnik - Aalto University, Finland Ahmad Rafi - Mutimedia University, Malaysia
Matthias Kulcke* - Hamburg University of Rabee Reffat - Assiut University, Egypt
Technology, HafenCity University Hamburg, Blair Satterfield - University of British Columbia,
Germany School of Architecture, Canada
Andrew Li - Kyoto Institute of Technology, Japan Marc Aurel Schnabel - Victoria University of
Thorsten Lomker - Zayed University Dubai, United Wellington, New Zealand
Arab Emirates Sven Schneider - Bauhaus University, Weimar,
Wolfgang Lorenz* - TU Wien – Institute of Germany
Architectural Sciences, Austria Tim Schork* - RAW - The Monash Future Materials
Russell Loveridge - NCCR Digital Fabrication, ETH & Processes Laboratory, Monash University,
Zurich - Swiss Federal Institute of Technology Australia
in Zurich, Switzerland Davide Simeone - Sapienza University of Rome,
Earl Mark - University of Virginia, School of Italy
Architecture, United States A. Benjamin Spaeth - Welsh School of Architecture,
Bob Martens* - TU Wien, Austria Cardiff University, United Kingdom
Tom Maver - Glasgow School of Art, United Martijn Stellingwerff - Delft University of
Kingdom Technology, Netherlands
AnnaLisa Meyboom - School of Architecture and Rudi Stouffs* - National University of Singapore,
Landscape Architecture, Canada Singapore
Volker Mueller - Bentley Systems, United States Tiemen Strobbe - Gent University, Belgium
Michael Mullins - Aalborg University, Denmark Kjeld Svidt* - Aalborg University, Denmark
Marc Muylle - University of Antwerp, Belgium Martin Tamke* - The Royal Danish Academy of
Fine Arts, Schools of Architecture, Design and
Pirouz Nourian - Delft University of Technology, Conservation, Denmark
Netherlands Emine Mine Thompson* - Northumbria University,
Kateřina Nováková* - Faculty of Architecture, United Kingdom
Czech Technical University in Prague, Czech
Republic Armando Trento* - Sapienza University of Rome,
Italy
Toni Österlund* - Oulu School of Architecture, Bige Tuncer - Singapore University of Technology
University of Oulu, Finland
and Design, Singapore
Rivka Oxman - Technion – Israel Institute of
Technology , Israel Žiga Turk - University of Ljubljana, Slovenia
Mine Özkar* - Istanbul Technical University, Turkey Jos van Leeuwen - Hague University of Applied
Sciences, Netherlands
Burak Pak* - KU Leuven - University of Leuven Johan Verbeke* - KU Leuven - University of
Faculty of Architecture, Belgium Leuven, Belgium
Vera Parlac - University of Calgary, Canada Spela Verovsek - University of Ljubljana, Slovenia
Jenni Partanen - Tampere University of Maria Voyatzaki - Aristotle University of
Technology, Finland Thessaloniki, School of Architecture, Greece
Chengzhi Peng - University of Sheffield, United Ivo Vrouwe - KU Leuven - University of Leuven,
Kingdom Belgium
Frank Petzold - Technical University of Munich, Jerzy Wojtowicz - Warsaw University of
Germany Technology, Poland
12 | eCAADe 34 - List of Reviewers - Volume 1

Stefan Wrona - Faculty of Architecture, Warsaw
University of Technology, Poland
Gabriel Wurzer* - TU Wien, Austria
Claudia Yamu - University of Groningen,
Netherlands
Tadeja Zupancic - University of Ljubljana, Faculty
of Architecture, Slovenia
List of Reviewers - Volume 1 - eCAADe 34 | 13

Contents
5 Theme
7 Acknowledgements
11 List of Reviewers
21 KEYNOTES
23 Keynote Speakers
25 Simplexity (and Complicity) in Architecture
Branko Kolarevic
33 Complex Modelling
Mette Ramsgaard Thomsen
43 Structural Design
Toni Kotnik
51 WORKSHOPS
53 The Wave of Autonomous Mobility
Jaehwan Kim, Mathew Schwartz, Andrzej Zarzycki
59 Contextualized Digital Heritage Workshop - Oulu
Danilo Di Mascio, Anetta Kepczynska-Walczak, Nicholas Webb
63 Digital Leadership
Tadeja Zupancic, Johan Verbeke, Henri Achten, Aulikki Herneoja
69 Folding The Unfoldable
Ergun Akleman, Negar Kalantar, Alireza Borhani
73 Transformational Intelligent Systems
Pablo Baquero, Nelson Montas, Effimia Giannopoulou
77 Multi User Online Design Environment for Mass Housing
Lo Tian Tian, Marc Aurel Schnabel
81 FABRICATION | Design & Application

83 Adaptive Fabrication
Zubin Khabazi, Michael Budig
93 Negotiate My Force Flow
Oliver Tessmann, Moritz Rumpf, Philipp Eisenbach, Manfred Grohmann, Topi Äikäs
103 Additive Manufacturing and Value Creation
Heidi Turunen
Contents - Volume 1 - eCAADe 34 | 15

113 Groove Light
Shinya Okuda, Lin Zhenyi
121 FABRICATION | Robotics: Design & Assembling

123 Building Traditions with Digital Research
Rui Oliveira, Jose Pedro Sousa
133 Biomimetic Robotic Construction Process
Chi-Li Cheng, June-Hao Hou
143 Discrete Assembly and Digital Materials in Architecture
Gilles Retsin
153 The Robotic Production of the GRC Panels in the CorkCrete Arch Project
José Pedro Sousa, Pedro Filipe Martins
161 FABRICATION | Robotics: Generative Design

163 Calculated Matter
Elif Erdine, Alexandros Kallegias
169 Optimization Process Towards Robotic Manufacturing in Actual Scale
Odysseas Kontovourkis, Panagiota Konatzii
179 AMA - Additive Merged Appliance
Alina Rejepava, Mario Haring, Anna Gruber
185 Generating an Anamorphic Image on a Curved Surface Utilizing Robotic
Fabrication Process
Marko Jovanovic, Vesna Stojakovic, Bojan Tepavcevic, Dejan Mitov, Ivana Bajsanski
193 CAAD EDUCATION | Design Concepts & Strategies

195 Digital Architectural Composition - 30 years of experience and
experimentation
Aleksander Asanowicz
205 Flying Bricks
Wolfgang E. Lorenz, Gabriel Wurzer
213 Magnet-based Interactive Kinetic Bricks
Mohammad Kouchaki, Mohammadjavad Mahdavinejad, Parastoo Zali, Shahab
Ahmadi
219 Aesthetics of Decision
Alessio Erioli
16 | eCAADe 34 - Contents - Volume 1

229 Complexity and Simplicity
A. Benjamin Spaeth, Theodoros Dounas, Joachim Kieferle
237 From Designing Buildings from Systems to Designing Systems for Buildings
Heike Matcha
241 Ubiquitous Digital Repositories In the Design Studio
Theodoros Dounas, Benjamin Spaeth
251 CAAD EDUCATION | Design Tools

253 Computational Thinking and the Architectural Curriculum
Guzden Varinlioglu, Suheyla Muge Halici, Sema Alacam
261 Teaching Programming and Urban Complexity to Architecture Students
Lukas Treyer, Daniel Zünd
269 Experience in CAAD Education Using a MOOC System
Ádám Tamás Kovács, Mihály Szoboszlai
275 Dynamic Light as a Transformational Tool in Computer-aided Design
Ellen Kathrine Hansen, Michael Finbarr Mullins, Georgios Triantafyllidis
283 IT Driven Architectural Design for All?
Sebastian Białkowski, Anetta Kępczyńska-Walczak
291 Implementing Biomorphic Design
Asterios Agkathidis
299 Building Information Modelling - the Quest for Simplicity Within Complexity
Anetta Kepczynska-Walczak
309 CAAD EDUCATION | Fabrication

311 Crafting as Inquiry into Computation
Vernelle A. A. Noel
321 FlexiMold: Teaching Numeric Control through a Hybrid Device
Serdar Aşut, Winfried Meijer
329 Fabricating Semi Predictable Surfaces
Joachim B. Kieferle, George Katodrytis
335 Nip and Tuck: A Simple Approach to Fabricate Double-Curved Surfaces with
2D Cutting
Negar Kalantar, Alireza Borhani, Ergun Akleman
345 Digital Materialization: Additive and Robotical Manufacturing with Clay and
Silicone
Cristina Nan, Charlie Patterson, Remo Pedreschi

355 CAAD EDUCATION | Creative Practices
357 Computation in Early Design Education as Investment in Attitudes
Desantila Hysa, Mine Özkar
367 Objects-to-sense-with
Athina Papadopoulou
377 'Starting with Nothing' and 'Ending up with Something'
Jeremy Ham, Marc Aurel Schnabel, Lawrence Harvey, Daniel Prohasky
387 Understanding the Verbal Concepts Appropriated by the Students in the
Architectural Design Studio
Burak Pak, Zeynep Aydemir
395 GENERATIVE DESIGN | Design Concepts & Strategies

397 Coding as Creative Practice
Corneel Cannaerts
405 Using Processing with Architectural 3D Modelling
Inês Caetano, António Leitão
413 Using the Phase Space to Design Complexity
Mary Katherine Heinrich, Phil Ayres
423 Evolving Parametric Models using Genetic Programming with Artificial
Selection
John Harding
433 GENERATIVE DESIGN | Applications

435 Homeorhetic Assemblies
Roberto Monesi, Alessio Erioli
445 Monolith: The Biomedical Paradigm and the Inner Complexity of Hierarchical
Material Design
Panagiotis Michalatos, Andrew Payne
455 Design Exploring Complexity in Architectural Shells
David Gerber, Evangelos Pantazis
465 DrAFT: an Algorithmic Framework for Facade Design
Inês Caetano, António Leitão
475 Elemental Intricacy
Alexandros Kallegias, Eleni Pattichi

483 COLLABORATION AND PARTICIPATION
485 CumInCAD 2.0: A Redesigned Scalable Cloud Deployment
Tomo Cerovsek, Bob Martens
493 Collaborative Design with Mobile Augmented Reality
Leman Figen Gül, Süheyla Müge Halıcı
501 De Humani Corporis Fabrica - Fabricating Emotions through Architecture
Maria da Piedade Ferreira, Andreas Kretzer, José Pinto Duarte, Didier Stricker,
Benjamin Schenkenberger, Markus Weber, Takumi Toyama
509 SMART AND RESPONSIVE DESIGN

511 Responsive Systems: Foundations and Application
Goncalo Castro Henriques
521 Material Agency and Physical Boundaries
Christina Doumpioti
529 Using Tensegrity and Folding to Generate Soft Responsive Architectural Skins
Sherif Abdelmohsen, Passaint Massoud, Ahmed Elshafei
537 Sensitive Shelters: Poetics of Interaction
Andrés Martin Passaro, Gonçalo Castro Henriques, Rodrigo Cury Paraizo
549 Adaptive Pneumatic Shell Structures
Paul Poinet, Ehsan Baharlou, Tobias Schwinn, Achim Menges
559 MATERIAL STUDIES | Concepts

561 Model Translations
Ashish Mohite, Toni Kotnik
571 Knitflatable Architecture
Yuliya Baranovskaya, Marshall Prado, Moritz Dörstelmann, Achim Menges
581 Natural Complexity
Arielle Blonder, Yasha Jacob Grobman
591 "Free Skin" Collaboration
Timo Carl, Frank Stepper
601 MATERIAL STUDIES | Applications

603 Bio-Enabled Façade Systems
Martina Decker, George Hahn, Libertad M. Harris
613 From Interactivity Towards Ambience Through a Bottle-brick
Kateřina Nováková, Henri Achten

621 Bending-active Structures
Eleanna Panagoulia, Simon Schleicher
631 rolyPOLY
Andrew Wit, Simon Kim
639 Material Active Geometry
Alireza Borhani, Negar Kalantar
649 TOWARDS SMARTER CITIES | Concepts and Strategies

651 Smart City: Adding to the Complexity of Cities
Emine Mine Thompson
661 From Moving Cube to Urban Interactive Structures
Vasilija Abramovic, Henri Achten
669 SenCity - Piloting Intelligent Lighting and User-Oriented Services in Complex
Smart City Environments
Henrika Pihlajaniemi, Eveliina Juntunen, Anna Luusua, Mirva Tarkka-Salin, Johan
Juntunen
681 Adaptive Designs with Distributed Intelligent Systems
Andrzej Zarzycki
691 Animating the Static. Case Study of The Project "Urbanimals"
Sebastian Dobiesz, Anna Grajper

KEYNOTES
Keynote Speakers
Branko Kolarevic
Branko Kolarevic holds the Chair in Integrated Design and co-directs the Laboratory for Integrative De-
sign (LID). Prior to his appointment at the University of Calgary, he was the Irving Distinguished Visiting
Professor at Ball State University in Indiana. He has taught architecture at several universities in North
America, most recently at the University of Pennsylvania, and in Asia, in Hong Kong.
He has lectured worldwide on the use of digital technologies in design and production and has au-
thored, edited or co-edited several books, including the recently published “Manufacturing Material Ef-
fects: Rethinking Design and Making in Architecture” (with Kevin Klinger), “Performative Architecture:
Beyond Instrumentality” (with Ali Malkawi) and “Architecture in the Digital Age: Design and Manufac-
turing.” He is the past president of the Association for Computer Aided Design in Architecture (ACADIA)
and is the recipient of the ACADIA 2007 Award for Innovative Research.
Mette Ramsgaard Thomsen

Mette Ramsgaard Thomsen’s research centers on the intersection between architecture and computer
science. During the last 15 years her focus has been on the profound changes that digital technologies
instigate in the way architecture is thought, designed and built. In 2005 she founded the Centre for IT
and Architecture research group (CITA) at the Royal Academy of Fine Arts, School of Architecture, Design
and Conservation, Denmark. During the last 6 years she has successfully built up a strong international
group. In 2010 she became full Professor in Architecture and Digital Technologies.
In CITA she has piloted a special research focus on the new digital-material relations that digital tech-
nologies bring forth. Investigating advanced computer modelling, digital fabrication and material spec-
ification CITA has been central in the forming of an international research field examining the changes
to material practice in architecture. This has been led by a series of research investigations developing
concepts and technologies as well as strategic projects such as the international Digital Crafting Network
that fosters interdisciplinary sharing and dissemination of expertise and supports new collaborations in
the fields of architecture, engineering and design.
Toni Kotnik
Toni Kotnik is Professor of Design of Structures at Department of Architecture in Aalto ARTS. Toni Kot-
nik (born 1969, Germany) received his Ph.D. degree (Dr. sc. Nat.) in mathematics from the Univ. of
Zurich, Switzerland, in 1999. After completion of his second degree in architecture he worked as se-
nior researcher at the Swiss Federal Institute of Technology (ETH) in Zurich and as studio master at the
Architectural Association (AA) in London. Dr. Kotnik was Ass. Professor at the Institute for Experimen-
tal Architecture at the University of Innsbruck and Associate Professor at the Singapore University of
Technology and Design. He is principal of d'HKL, a Zurich-based office focusing on experimental and
research-oriented architectural design, and has been lecturing worldwide including Harvard University,
Copenhagen Royal Academy of Fine Arts, Princeton University, The Bartlett School of Architecture, Oslo
School of Architecture and Design, Sydney University of Technology, the Guggenheim Museum Bilbao
Keynote Speakers - Volume 1 - eCAADe 34 | 23

and MoMA New York. His practice and research work has been published and exhibited internation-
ally, including the Venice Biennale, and is centered on the integration of knowledge from science and
engineering into architectural thinking and the design process.
24 | eCAADe 34 - Keynote Speakers - Volume 1

Simplexity (and Complicity) in Architecture
Branko Kolarevic1
1
University of Calgary
1
branko.kolarevic@ucalgary.ca
The paper examines the interplay between simple and complex in contemporary
architecture. It discusses how complex-looking designs could arise out of simple
design and production strategies (i.e. through "simplexity") and likewise, how
simple-looking designs can be based on rather complex conceptual and tectonic
solutions (i.e. "complicity"). The paper focuses on simplexity as a conceptual
framework for architects to embrace complexity in design.
Keywords: Simplexity, Complicity
INTRODUCTION ily complicated designs and tectonic details; how-

The broad adoption of parametric design and digi- ever, complexity - whether formal, spatial, program-
tal fabrication in architecture schools and offices in matic, or tectonic - need not be complicated. Simple
the first decade of this century led to a prolifera- rules and processes can produce remarkably com-
tion of complex-looking designs, often with dou- plex looking outcomes. As argued later in this paper,
bly curved surfaces and highly differentiated com- simplexity - defined as simple complexity - is seen as
ponents in series, each with a unique shape. This a promising way to use effectively digital design and
triumph of differentiation over repetition marked a production technologies.
dramatic departure from mass production as one of In the sections that follow, complexity and sim-
the basic tenets of Modernism, which premised eco- plicity are briefly discussed, with their manifestations
nomic production of buildings on an endless rep- within the disciplinary context of architecture. Sim-
etition of standardized, identical components. The plexity is introduced as a paradoxical notion that syn-
emergence of geometric modeling software based thesizes simplicity and complexity in a productive,
on NURBS meant that there were no limits to the conceptually compelling duality, resulting in simple
kinds of complex geometries that could be accu- design strategies that may lead to the emergence
rately represented. The computer controlled ma- of complex looking formal outcomes. Such thinking
chinery for subtractive and additive fabrication made is in line with the complexity theory (Gribbin 2004),
complex shapes relatively easily attainable in a vari- but is still at the periphery of interest in contempo-
ety of materials. Geometric complexity was no longer rary architecture. The principal argument is that new
a barrier, either representationally or production- design insights may result from exploring how com-
wise. plexity and simplicity could coexist and co-evolve in
Many saw the resulting formal exuberance as the design process.
an aberration, an unnecessary distraction that would
quickly fade out. In some projects, complexity be-
came an end in itself, often leading to unnecessar-
Keynotes - Volume 1 - eCAADe 34 | 25

EXPLAINING COMPLEXITY the form of hierarchy, as a principal structural scheme
According to the Merriam-Webster dictionary, com- in which a complex system is decomposed into sev-
plex means "a group of things that are connected eral subsystems, which in turn have subsystems, and
in complicated ways;" complex and complicated are so on. Simon offers additional insights, describing
considered as synonyms, with "having confusingly the dynamic properties of hierarchical systems and
interrelated parts" provided as a definition. The dis- how their behavior could be analyzed.
cussion of synonyms, however, does make a distinc- The focus on structure is at a center of the aptly
tion later on between the complex and the compli- named structuralism, which has emerged in 1960s to-
cated: complex "suggests the unavoidable result of gether with reductionism as a parallel development.
a necessary combining and does not imply a fault The systems theory was developed at that time too,
or failure," whereas complicated "applies to what of- with its discussion of open and closed systems, and
fers great difficulty in understanding, solving, or ex- the emphasis on interactions between the system
plaining." Clemens Bellut (2008) examines the distinc- and its environment (in open systems). Certain char-
tion between the "complex" and the "complicated" acteristic properties of complex systems were identi-
in a similar vein when discussing the concept of fied early on: the importance of feedback loops, the
complexity from a philosophical perspective: stating notion that a small change can cause dramatic ef-
that something is complicated means that "the finite fects (the so-called "threshold effect"), and that emer-
number of its determinations cannot be grasped di- gence and unpredictability are essential.
rectly;" by contrast, saying that something is complex There is a shared understanding in various fields
means that "the number of its determinations is sim- that complexity is dynamic in nature rather than
ply infinite." He also observes that "something com- static, which then raised the questions of whether
plex [...] can assume the appearance of simplicity, there is a continuum between simplicity and com-
while the complicated always excludes everything plexity and whether that continuum is linear or non-
simple from itself." linear. Furthermore, there is also a question of mea-
Complexity as a concept is present in many do- suring complexity, as it is rather obvious that some
mains, fields and disciplines, carrying a different phenomena are less and some are more complex.
meaning depending on the context in which it is ap- Kolmogorov has proposed that measuring complex-
plied. As Mitchell (2009) points out, however, neither ity is rather straightforward: if some system could be
a single science of complexity nor a single complex- described in a concise manner then it is less complex.
ity theory exists yet. According to Erdi (2008), in most Kolmogorov complexity, also known as K-complexity
contexts the discussion of complexity is related to the and algorithmic complexity, measures the complex-
structure of a system. Herbert Simon in his seminal ity of a given system or a problem by the length of
article "Architecture of complexity" published in 1962 the algorithm that describes it. Kolmogorov com-
defined the principal challenge as finding the right plexity is particularly apt in the context of architec-
representation to describe complex structures: ture (and visual arts) as it distinguishes between vi-
"How complex or simple a structure is depends sual and structural complexity. As observed by Je-
critically upon the way in which we describe it. Most sus Mosterin (2002), "regardless of the complexity in-
of the complex structures found in the world are volved in the appearance of a pattern, complexity is
enormously redundant, and we can use this redun- by definition based on the reproducing algorithm,
dancy to simplify their description. But to use it, to that is, a series of instructions that will generate the
achieve the simplification, we must find the right rep- visual pattern."
resentation." For some, the essence of complexity is that the
Simon argues that complexity frequently takes outcome of the system's operation should not be ob-
26 | eCAADe 34 - Keynotes - Volume 1

vious from its constituent parts and their interactions complexities present in the projects by Peter Eisen-
- all of which could be rather simple. Such struc- man, Frank Gehry, and Daniel Libeskind, for example,
tural simplicity seems to be at the core of research are manifestations of broader cultural influences of
aimed at understanding complex phenomena. Ac- the emerging sciences of complexity. Jencks makes
cording to Gribbin (2004), "complexity theory is unit clear that architectural form is where complexity
covering the deep simplicity from which complexity manifests itself.
arises: simple laws, non-linearity, and sensitivity to Arguably, Modernism was about simplification in
initial conditions." It is this "deep" structural simplicity architecture, succinctly articulated as "less is more"
in design systems capable of generating complex for- by Mies van der Rohe. The resulting monotonic sim-
mal and spatial outcomes that is the focus of this pa- plicity in urban contexts is what Robert Venturi's was
per. The topological definition of the design's struc- reacting to. Venturi was interested in the effects that
ture, i.e. the definition of the relationships between buildings have on the perceptions; Modern architec-
its constituent parts plays an essential role in this pro- ture with its stripped down articulation of surfaces
cess. The complexity of a design could be expressed and simple forms didn't offer much to the eye. It was
as the length of the algorithmic description of its gen- the absence of surface ornamentation and complex
esis. What is interesting to note here is that the algo- forms that arguably made the minimalist aesthetics
rithmic definition of the design is independent of its of Modernism less affectionate (Kolarevic and Klinger
form; it operates at the structural level. 2008), contributing in part to its demise.
According to Ernst Gombrich (1979), the human
COMPLEXITY AND SIMPLICITY IN ARCHI- mind has an intrinsic need for "careful balance" be-
tween complexity and order. The mind has no trou-
TECTURE
ble deconstructing a simple, regular grid (i.e. recog-
There is no single definition of complexity in the con-
nizing the monotonous); it quickly "disconnects" in
text of architecture. We talk about complexity as
reading complex configurations if it cannot recog-
being social, cultural, programmatic, functional, for-
nize an underlying structure. Gombrich argues that
mal, spatial, tectonic, performative, phenomenolog-
a "careful balance" between these two conditions,
ical... It was Robert Venturi (1966) who famously
i.e. between monotony and complexity, is what the
re-introduced the subject of complexity into archi-
mind looks for in its constant processing of the sur-
tecture; Venturi, however, wasn't specific about the
rounding environments. In that vein, the infatua-
kinds of complexity he was interested in, making
tion with complex geometry in mid-1990s soon was
references both to historically present complexities
replaced by the exploration of highly crafted, non-
in architecture and the complexities associated with
uniform surface effects based on complex pattern-
the design itself. Around the same time, György
ing, texturing, or relief (Kolarevic and Klinger 2008).
Kepes (1965), a Hungarian-American art theoreti-
cian, heralded the emergence of movement from the
"classical sciences of simplicity toward a modern sci- COMPLICITY AND SIMPLEXITY
ence of ordered complexity." For Kepes, complexity In the history or architecture complexity and simplic-
offered an integrative structural ordering that would ity have generally been considered as contrary no-
interconnect arts, architecture, science and technol- tions. The complexity science demonstrates other-
ogy. Three decades later, Charles Jencks in his "Archi- wise, that simplicity and complexity are interdepen-
tecture of the Jumping Universe" (1995) juxtaposes dent and mutually coexisting. As noted by complex-
what he refers to as the "Post-Modern sciences of ity theorists, there is a dialectical interplay between
complexity" to the "Modern sciences of simplicity," simplicity and complexity (Gribbin 2004). Complex-
with the former sublating the latter. For Jencks, the ity can arise out of simplicity, and likewise, simplic-

ity can arise out of complexity, giving rise to sim- careful deliberation and selection of influences that
plexity and complicity, respectively, as new concepts would affect the designs that could emerge from the
that capture the interdependence and coexistence of operation of the system.
complexity and simplicity and interaction and emer-
gence that is implied. The exact origins of these ECONOMY OF METHOD
terms are not clear. Jack Cohen and Ian Stewart After seeing many students hopelessly entangled in
(1995) define complicity and simplexity interchange- complicated complexities of their own making, it
ably as "interaction between coevolving systems that became apparent rather quickly that teaching (or
supports a tendency toward complexity" and a "sim- "preaching") of complexity requires a pedagogic re-
pler order to emerge from complexity". For Jeffrey finement. As argued earlier, complexity - whether
Kruger (2008), simplexity is about a "complemen- formal, spatial, programmatic, or tectonic - need not
tary relationship between complexity and simplicity". be complicated. Simple rules and processes can pro-
Obviously, complicity and simplexity are hybrid, reduce remarkably complex looking outcomes. A par-
lated concepts framed by a synergetic relationship ticularly potent example of a complex-looking de-
between simplicity and complexity. In my view, sim- sign based on simple rules is the design for the Ser-
plexity can be defined as simple complexity and, like- pentine Pavilion in London (2002, figure 1) by Cecil
wise, complicity as complex simplicity. Balmond and Toyo Ito. The apparently random pat-
Simplexity is compelling because it offers a way terning that wraps the entire pavilion is produced
of describing seemingly complex outcomes by rel- by incremental scaling and rotation of a series of in-
atively few simple rules and interactions. Simple, scribed squares, whose edges were extended and
straightforward rules may facilitate the emergence of trimmed by the pavilion's unfolded box shape (figure
highly complex designs by defining a direction with- 2) to create a beautiful, seemingly irregular-looking
out confining it. This conditioned emergence can pattern of alternating voids and solids.
lead to shapes, forms, and spatial organizations that There are numerous examples of designers using
look surprisingly complex, yet, at the core, are gen- design strategies based on simple rules to produce
erated by a system based on simple rules. (Cellular complex looking visual, surface, spatial or formal out-
automata are an example of simplexity, where sim- comes (Kolarevic and Klinger, 2008). Such complex-
ple rules can lead to startlingly complex outcomes.) ity based on simplicity - that is, simplexity - implies a
Intentional simplicity, however, can be problem- certain economy of method that became the essence
atic, because a single goal, single focus, single pre- of my design teaching over the past decade.
occupation can preclude consideration of any oth- To explore the relationship between complex-
ers. Simplicity may thus diminish the consideration ity and simplicity and to understand their interplay,
of other, often more promising options that exist I have asked students to produce relatively quickly,
on the periphery and favor "exploitation" over ex- in an iterative fashion over five weeks, a series of self-
ploration. A reductive spiral of simplicity can be of- similar, complex looking objects by defining simple
ten counterproductive in design. But, as Obendorf parametric and production processes. An important
(2009) observes, "reduction is the path to simplicity, design and production dimension of the exercise was
and minimalism describes paths to approach reduc- a certain "economy of method," introduced as "less
tion." He argues for minimalism as a "tool to think effort, less machine time, less material, less waste,"
about the simple and to discover and instantiate pat- and summed up in the end as "less for more" - a
terns for designing simplicity." Generating visual, spa- thinly veiled reference to Mies van der Rohe's famous
tial, or formal complexity with minimal means, mini- motto, but with an entirely different connotation.
mal structure, is actually not that simple. It requires This design/production dimension was an attempt

Figure 1
The Serpentine
Pavilion in London
(2002), designed by
Cecil Balmond and
Toyo Ito.
Figure 2
Serpentine Pavilion:
the
irregular-looking
pattern is based on
incremental scaling
and rotation of a
series of inscribed
squares.

Figure 3
Expanded
Topographies, by
Dustin Headley
(2006).
to introduce resource economy (time-, material-, and Figure 4

energy-wise) into the design and production pro- Expanded
cesses. Complex effects were to be achieved through Topographies: a
simple means; the underlying ethos being that com- simple parametric
plexity need not be synonymous with complicated, procedure
i.e. that conceptual and production simplicity can automatically
produce a perception of complexity in the outcome. generates different
Expanded Topographies (figure 3), a project by cutting patterns.
Dustin Headley completed in 2006 at Ball State Uni-
The project's premise was that variegated surface
versity, offers a particularly successful demonstra-
patterns, i.e. apertures of gradually increasing or de-
tion of such a resource economy approach to design
creasing sizes, could be produced by simply vary-
and production. It was inspired by research into ex-
ing the values of expansion parameters including the
panded metal meshes, which are produced by simul-
length of cut, aligned spacing between the cuts, and
taneous slitting and stretching of a flat sheet of metal,
spacing between the successive lines of cuts. Using
resulting in a regular, repetitive pattern of diamond-
scripting in Rhinoceros, a simple parametric proce-
shaped holes. What is interesting about this process
dure automatically generated different cutting pat-
is its geometric and production simplicity, and that
terns (figure 4), which could be directly transmitted
nearly zero metal waste is generated during the pro-
to a digitally controlled cutting machine. Various pro-
cess; in addition, the final product - the expanded
totypes were produced by laser-cutting flat, rectan-
mesh - is stronger (by kilogram) and lighter (by me-
gular sheets of acrylic, which were then heated and
ter) than the original sheet.
expanded by applying equal force (in opposite di-

rections) to the two shorter sides of the sheet. The REFERENCES
sheets would deform in the process, depending on Bellut, C 2008, 'The Gordian Knot of Complexity', in
the density and the lengths of the cuts, producing to- Gleiniger, A and Vrachliotis, G (eds) 2008, Complex-
pographic surfaces, with apertures that vary in size ity: Design Strategy and World View, Birkhauser, Basel
Cohen, J and Stewart, I 1995, The Collapse Of Chaos:
across the length of the surface. Precise topogra-
Discovering Simplicity In A Complex World, Penguin
phies were produced by controlling the length of Books, London, UK
each cut and X and Y spacing between the adja- Érdi, P 2008, Complexity explained, Springer, Berlin
cent cuts. In addition, by making non-parallel cuts, Gombrich, EH 1979, The Sense of Order: A Study in the Psy-
i.e. by introducing angle as an additional parameter, chology of Decorative Art, Cornell University Press,
further possibilities for surface articulation opened Ithaca, NY
Gribbin, JR 2004, Deep Simplicity: Bringing Order To Chaos
up. The design and production processes were sim-
And Complexity, Random House, New York
ple and straightforward, with nearly zero material Jencks, C 1995, Architecture of the Jumping Universe: A
waste, resulting in an artifact with intricate surface ef- Polemic: How Complexity Science is Changing Archi-
fects, subtle undulations and series of apertures that tecture and Culture, Wiley, New York
change in size across the length of the panel. Kepes, G (eds) 1965, Structure in Art and Science, George
Braziller, New York
Kolarevic, B and Klinger, K 2008, 'Manufacturing / Ma-
CONCLUSION terial / Effects', in Kolarevic, K and Klinger, K (eds)
The multiplicity of opposite meanings associated 2008, Manufacturing Material Effects: Rethinking De-
with simplicity and complexity in architecture and sign and Making in Architecture, Spon Press (Taylor &
the implied paradoxical relationship between the Francis), London, UK
Kruger, J 2008, Simplexity: Why Simple Things Become
two actually open up interesting opportunities to ex-
Complex (And How Complex Things Can Be Made Sim-
plore them in a non-dichotomous way. The dialecti- ple), Hyperion, New York
cal relationship between the two could be very pro- Mitchell, M 2009, Complexity: A Guided Tour, Oxford Uni-
ductive, as demonstrated by a number of authors versity Press, Oxford, UK
from a very broad range of disciplines and fields, who Mosterin, J 2002 'Kolmogorov Complexity', in Agazzi,
argue that complexity and simplicity coexist and im- Evandro and Montecucco, Luisa (eds.), Complexity and
Emergence, Proceedings of the Annual Meeting of the
ply each other in so many different contexts. Think-
International Academy of the Philosophy of Science,
ing of them as diametrical opposites diminishes the Bergamo, Italy
productive role of simplicity in understanding com- Obendorf, H 2009, Minimalism: Designing Simplicity,
plexity (and vice-versa). The principal argument is Springer, New York
that simplicity is an elegant, efficient, way to produce Simon, HA 1962 'The architecture of complexity', Pro-
complexity in architecture. The resulting simplexity ceedings of the American Philosophical Society, Vol.
106, No. 6. (Dec. 12, 1962), pp. 467-482
could indeed be the source of new conceptual "min-
Venturi, R 1966, Complexity and Contradiction in Architec-
imalism" in architecture. ture, Museum of Modern Art, New York

Complex Modelling
Questioning the infrastructures of information modelling
Mette Ramsgaard Thomsen1

1
The Royal Danish Academy of Fine Arts
1
Mette.Thomsen@kadk.dk
This paper outlines the keynote given at eCAADe 2016, Oulu.
Keywords: Simulation, Adaptive Parametrisation, Open Topologies, Multi Scale

Modelling
INTRODUCTION engineered materials from the artificial silk of nylon

We are finding ourselves in a time of rethinking our stockings to high modulus polymers, ceramic com-
material practices. As research into design compu- posites to Nano-materials (Beukers 1999, pp. 14-
tation matures, its fundamental links to fabrication 15). Now, architecture and building culture is enter-
are challenging industrialised paradigms of building ing this way of thinking. The link between design
construction. With origin in the many explorations of and direct material manufacture and the ability to
mass customisation and material investigation, our programme advanced steering mechanisms, has al-
practice is entering a new era in which architects lowed architects to prototype material systems that
become the designers of materials as well as arte- grade intensity or structure in response to design in-
facts. The interest in designing directly for fabrication tent, optimising material usage and employing in-
has transcended the scales of materialisation mov- herent material performances. In architecture, the
ing from the fabrication of individualised building el- ability to use materials in smarter and less intense
ements to the direct production and specification of manner are fundamental building blocks for the con-
the materials themselves. This extension of the dig- ception of lighter building culture; a culture in which
ital design chain integrating the design of not only lighter means less material, less transportation and
material systems but also the composition of their altogether a lighter impact on our environments.
components is now an overarching paradigm inform- This paper discusses the modelling paradigms
ing research and finding its way into practice. needed to engage with this new material practices.
In this paradigm materials are no longer stan- If traditional architectural drafting have successfully
dardised to a measure of global uniformity, but in- transitioned into parametric design paradigm such
stead locally tuned and optimised. This thinking is as Building Information Modelling (BIM) and other
linked to interdisciplinary efforts in the broader fields more bespoke computational design practices, the
of engineering and material science that are rethink- question remains how we transform these systems to
ing how we materialise our world. We are living in a engage the profoundly inter-disciplinary and inter-
time of extreme material inventiveness and are sur- scalar design environment of the extended digital
rounded by highly engineered materials. The 20th chain. When we design across the scales of the build-
century has brought with it an era of synthetic and ing system, the element and the material, we en-

counter multiple scales of engagement all of which Figure 1
has its own discipline specific modelling paradigms. 3D printing auxetic
This paper questions how these models can be inte- material, Norbert
grated so as to enable strategic exchange between Palz, CITA 2009.
them, and what underlying infrastructures we need
to develop in order to work successfully with these
concepts. The claim is that these new practice ne-
cessitate new modelling practices that fundamen-
tally exceeds the boundaries of existing modelling
paradigms and that it is this ground research ques-
tion that is the core enquiry of our time.
CONSTRAINTS IN CURRENT MODELLING

PRACTICES
At present, the development of digital design tools Figure 2
in architecture is structured around large-scale indus- Stressed Skin, Paul
try led efforts that have sought to standardise infor- Nicholas, David
mation and develop shared protocols between inter- Stasiuk, Esben
disciplinary partners (Jernigan 2008). However, core Clausen Nørgaard,
efforts such as BIM have proved inadequate in tack- CITA 2015.
ling the high degrees of complexity of current build-
ing practice while at the same time not being able to
support the needs for flexible, intuitive and commu-
nicable design processes (Salim 2010, Holzer 2007).
Practice is aware of these embedded limitations of
current modelling practice and the industry devel-
opment has therefore been paralleled with a series
of profession-led research and development efforts
creating bespoke modelling methods allowing com-
plex design solutions and creating links to fabrica-
tion (Burry 2011, p. 28). The practice of architects
building their own information tools, encoding their
models and engaging directly with model interfacing
is therefore an embedded part of existing practice.
However, this effort is project-led, practice-specific
and rarely shared.
As building culture enters a rethinking of its ma-
terial practices, we need to future-proof our rep-
resentations. Rather than building common stan-
dards and libraries for known practices, we need to
develop the fundamental infrastructures for yet un-
known practices. This position fundamentally chal-
lenges some of the cornerstones of present mod-
elling paradigms.

Figure 3 A persistence of hierarchy: Building practice is tra-
The Rise, Martin ditionally organised hierarchically as a sequence of
Tamke, Mette subsystems each with their own scale of engagement
Ramsgaard and team of specialised professions. This division
Thomsen, David persists within modern design practice and is mir-
Stasiuk, Hollie rored by legislative according of responsibility within
Gibbons, CITA 2013. the design chain. However, this understanding of de-
In The Rise the sign as a progression through the scales limits the po-
growth algorithm tential for design innovation as it excludes our ability
and consecutive to understand how the small scales - material or de-
sprouting of tail - can affect large scales - environment or struc-
branches means, ture. To support a real implementation of feedback
that structural in the design chain we need to develop mechanisms
performance of the for multi-phased and multi-scalar feedback in which
artefact is cyclical interdependencies can be investigated and
continually assessed.
changing.
Figure 4
Tower, Mette
Ramsgaard
Thomsen, Martin
Tamke, Anders
Holden Deleuran,
Ida Tinning Friis,
CITA and Christoph
Gengnagel, Michel
Schmeck, UDK The ideal of the unified model: The ambition to inte-
2015. In Tower grate all design phases and practices has proven dif-
highly ficult as different practices use different kinds of tools
interdependent to analyse and represent knowledge. BIM models are
system of active breaking their own modelling frameworks in terms
bent GFRP rods and of pure scale becoming bigger, wider and deeper The expectation of known design parameters: De-
tensile membranes in the sense that they encompass more information sign is process of discovery. Unknown opportunities
interact to create from more disciplines (bigger), they include more and limitation appear through the collective invest-
overall structural phases or design (wider) and they expand into new ment into the design phase. However, current infor-
performance. These scales of design concern (deeper). This is creating a mation modelling necessitates a priori understand-
interactions are bottle-neck that is impeding innovation and creativ- ing of key design parameters. When new parame-
simulated ity in architectural design practice. Instead of inte- ters appear, design models break and either need re-
employing light grating information into one containing model, we programming or become messy hacks (Davis 2013,
weight goal based need to build networked models that pass informa- pp 37-47). A central premise for the future of the in-
methods and tion between discreet part models that are dedicated formation model is therefore that it engenders open
correlated to FE to particular tasks and can be continuously tuned and topologies by which adaptive parametrisation can
analysis. changed. take place so that we can build models in which pa-

rameters can appear or conversely dissipate into con- and material, our aim is to prototype future meth-
stants without breaking the design logic. ods for information modelling. The project builds
Geometry as information interface: Finally, we on CITA's central investigation into material perfor-
need to reconsider our reliance on geometry as a mance. Here, material performance is understood as
prime interface for information. Simulations and a core resource for design innovation being closely
generative design processes include numerical mod- tied to material optimisation. Positioning feedback
elling occurring across discreet time steps though as a central concern cascading through all scales of
which solutions emerge. We therefore need to ex- engagement, the project asks how dynamic modes
pand our perception of modelling as something that of organisation can introduce new logics into the de-
occurs in time and through event (Nicholas 2014). sign of architectural information models.
Furthermore, as we start implementing more com- Where an initial interest in Complex Modelling
plex n-dimensional data analysis algorithms such lay with the abstraction of the high dimensionality
as k-means clustering, the description of three- of complex solution spaces and the construction of
dimensional extension becomes a limitation in our methods by which to strategically steer these, it has
means of visualisation of a design space. Finally, become clear that the challenge is not how to de-
the expansion of digital design chain, making mate- sign in n-dimensions but rather how to understand
rial design part of architectural design, also includes and capture interdependency. In Complex Mod-
the encounter of new discipline- and scale-specific elling interdependency occurs at multiple levels. Be-
models that do not fit with geometric representation. tween material systems, between scales of material
Where building culture is used to other means of de- manipulation, between modes representation and
scription such as numeric tables for processes such modes analysis and simulation and between explicit
as cost estimation or stress strain graphs, these are design strategies and those that are generative and
understood as secondary to the actual 3D model. As optimised. As such each project engenders their
new modelling methods mature, we need to under- own distinct landscape of networked models. Where
stand the inherent heterogeneity of the networked concepts and tools are ported across projects, each
model while retaining the creative and intuitive prop- project finds their own particular way of orchestrat-
erties of our own methods. ing these into cohesive wholes.
The following goes through some of the cen-
RESEARCHING COMPLEX MODELLING tral enquiries of Complex Modelling. These are ex-
In CITA, the enquiry into how these modelling emplified through brief project presentations. Com-
paradigms are changing have shaped the last five plex Modelling follows an experimental design led
years of research. Currently this is framed by the re- method in which a series of linked research projects
search project Complex Modelling, a framing project as well as 'sketch probes' act as material experiments.
supported by the Danish Sapere Aude Advanced The emphasis on the design and implementation al-
Research Programme. Here, we are exploring fu- lows the project to engage directly with the inves-
ture modelling paradigms and how methods from tigated techniques and technologies moving along
parallel disciplines including engineering and com- the digital chain from design and analysis to spec-
puter science can broaden our practices and trans- ification and fabrication. This integrated approach
fer central concepts and tools. With special focus positions the research inquiries within a network of
on systems that integrate material performance, en- interconnected expertise and practice. The results,
gage high degrees of interdependency and allow the the design probes and the demonstrators generate
emergence of design agency and feedback between shared empirical data that can be further tested, anal-
the multiple scales of the structure, the element ysed and evaluated. The projects are therefore to

be seen more as vehicles by which the research en- FE simulations allowing for continual feedback be-
quiry can be undertaken than as independent re- tween early stage design exploration and later stage
search goals. The discussion presented here there- design refinement. Similarly, simulations are not lim-
fore moves across multiple projects exposing the ited to material or structural simulation, but incorpo-
methods by which different concepts are investi- rated into a larger framework of simulation particular
gated. to each individual project. These have included light
simulation (Nicholas 2014), computational fluid dy-
Integrating Simulation namics or agent based simulations (Tamke 2010) by
CITA's central interest in material performance has led which design performance, structural performance
to a fundamental exploration into integrating simu- and material assembly are negotiated.
lation. The emergence of the shared digital design This practice of conceiving the information
platform has brought with it an opening up of tradi- model as comprising multiple design integrated sim-
tional disciplinary boundaries and a merging of core ulations only intensifies as we work in a multi scalar
modelling concepts. The ability to interface complex design space. Here, scale specific simulations at the
analysis tools for the simulation of force and flow, scale of the structure, the element and the material
such as Finite Element (FE) analysis that discretise are interfaced, passing information between individ-
complex problems into finite numbers of interrelated ual part models. As such, simulation is not singular,
nodes to compute their force-relations, is influencing but instead recurrent and distributed across the land-
the thinking of structural design enabling the realisa- scape networked model.
tion of buildings with higher degree of formal free- Simulations introduce important concepts to de-
dom and structural complexity (Clough 1999). sign modelling. Occurring in time and through event,
In CITA, our interest lies especially with the in- simulation necessitate time steps thus making us un-
tegration of simulation into the early design phases. derstand design as the incremental movement to-
By allowing simulation to not only be employed as a wards solutions in difference to the explicit shaping
concluding evaluation of an already mature design, of absolute form. This new temporality of our compu-
but also a tool for exploration and form finding, sim- tational design space is exploded and discontinuous.
ulation becomes part of a larger set of dynamic mod- Rather, discreet events can be called, occur in parallel
elling tools for exploring and creating design possi- or enact in isolation triggering new design decisions.
bilities. This has meant a particular interest in light-
weight spring-based simulation tools including first Adaptive Parametrisation and Open
Nucleus (Holden Deleuran 2011) and later Kangaroo Topologies
but also self-scripted physics engines (Tamke 2014). A second focus in Complex Modelling is the explo-
In our extensive exploration of active bending (Rams- ration of methods for adaptive parametrisation in the
gaard Thomsen 2011, Tamke 2012, Nicholas 2013) information model. To preserve design control, para-
and highly interdependent material systems (Rams- metric models necessarily operate with a reduced
gaard Thomsen 2015, we examine light-weight sim- number of design parameters thereby reducing the
ulation as a means of understanding and formalising complexity of the model. Where this is practical, it
material behaviour, calibrating the bending of tim- also limits our way of understanding the potential of
ber lamellas or GFRP rods and calculating its interac- design synthesis and leads to an inherently reductive
tion with interconnected tensile members. In these design process.
investigations, simulation is not seen in isolation, but The field of computational design is currently en-
rather as intrinsically correlated. Light-weight simu- gaged in a push to build new models of for under-
lations are both tested against and testing more solid standing and capturing the emergent effects of high-

Figure 5
Leaning to be a
vault, David Stasiuk,
Mette Ramsgaard
Thomsen, CITA
2014. In Leaning to
be a vault
generative
algorithm are
coupled with
machine learning
algorithm to
explore the
classification of a
large design space
employing open
topologies. The
model optimises
the active bending
of rattan elements
as well as multiple
other design
parameters.
order parametric design. Here, new tools are being cess, allow an inherent flexibility in design by en-
prototyped for understanding multi objective opti- abling the activation of new - or the neglecting of
misation allowing architects to navigate the design obsolete - parameters. We need to build methods by
space along the Pareto-front and "...giving feedback which these can be steered and controlled so that we
on the best trade-offs that were found so far" (Vier- can move beyond random mutation and into much
linger 2014). These new tools support existing con- more deliberate progressions.
cepts for understanding design as 'optioneering' or In CITA recent work is exploring these concepts
'versioning', in which the singular design object, is through dynamic modelling tools including growth
seen as part of a larger space of possibility, that were algorithms and machine learning tools. Growth algo-
introduced already in early digital design processes. rithms introduce interesting ways of understanding
However, they often entail a fundamental fixing of the model as actively evolving through the design
the underlying topology onto which variation is pro- process. This firstly establishes a temporal dimen-
cessed. sion to the design model but also allow us to think of
To retain the flexibility of this new design space, topologies with changing body plans (Tamke 2013).
we need to embed the possibility for change of the This enquiry into open and adaptive modelling
topology of the design model. Open topologies, strategies has led to an interest in machine learn-
in which the dependencies between parameters are ing. Here, generative and evolutionary design strate-
emergent and open to change during the design pro- gies are coupled with tools for emergent classifica-

Figure 6
Leaning to be a
vault, David Stasiuk,
Mette Ramsgaard
Thomsen, CITA
2014. The
unsupervised
learning algorithm
classifies the design
space into
intuitively
understood
differences.

Figure 7
The multi scale
modelling levels of
Stressed Skin, Paul
Nicholas, David
Stasiuk, Esben
Clausen Nørgaard,
CITA 2015.
tion. In sketch projects such as "Learning to be a processes of expansion, increasing in number and in
vault" we have explore supervised as well as unsuper- complexity by each time step of evolution. The de-
vised learning methods for clustering large scale designer becomes part of a design cycle in which classi-
sign spaces into formal classification systems, guided fication, querying and management of data sets be-
both by the designer and allowed to emerge from come new design concepts.
a design model's underlying data structure (Stasiuk
2014). Multi Scale Modelling
Topological thinking has furthermore led to an In CITA, the interest in these new modelling
interest in graph modelling as a means to repre- paradigms is how they impact on the way we work
sent and manage the underlying interconnectivity with material systems. Where the above outlines
of design parameters and in new projects we are strategies for developing model landscapes com-
exploring neural networks methods such as NEAT posed of multiple dynamic and interacting part mod-
(NeuroEvolution of Augmenting Topologies) (Stanley els operating across different design phases and mul-
2002) by which to create evolutionary processes that tiple scales, the question becomes how inter-model
can change and optimise design topologies. communication is facilitated. The aim is feedback.
These processes present alternative strategies Rather than reducing the digital chain to adhere to
to the embedded reductionism of parametric mod- a traditional perception of design as a one-way pro-
elling. Here, models exist in multiples - in thou- cess of refinement consecutively encountering ever
sands of models - that are spawned by the generative smaller scales, our interest is in supporting inter scale
system to then be analysed by the learning system. relationships, in which design at the small scale is
Models are no longer singular end points but belong linked to design at the large scale.

This is explored through concepts for multi scale modelling methods is presenting. Where projects
modelling (Nicholas 2015). Multi scale modelling is tend to foreground the particular structural, material
interesting because of its broad interdisciplinary ap- or performative advancement it is investigating, the
plication in problems that aim to analyse and repre- new modes of information modelling developed to
sent large scale problems with high degrees of com- do so are equally important. Our field is criticised
plexity and where it becomes intractable to model for being speculative or overly interested in formal
the problem within one unified model. Multi scale aspects of design, both by more conventional prac-
modelling presents methods for coupling and inter- tice and by our neighbouring disciplines. However,
facing models so that meta scale models can be in- I believe that we need to think of these multiple ex-
formed and parametrised through lower scale mod- periments as the prototyping of methodologies. As
els but also conversely parse information back down such, we are embedded into a culture of ground re-
through the design chain thus profiting from both search exploration, which will profoundly change the
the macroscopic models as well as the accuracy of the modalities of architectural representation. This is im-
microscopic models (Weinan 2011). portant because, as all architects know, the means of
In CITA we employ multi scale modelling meth- representation are our means of conception and en-
ods using both nested strategies in which lower scale action.
models neatly reference the uniform subdivision of Furthermore, it is a field in which we are con-
a larger scale model as well as more tactical strate- tributing quite singularly. Where design computa-
gies in local variation in the scale subdivisions corre- tion is fundamentally interdisciplinary, and core re-
late to areas of higher degrees of complexity. In these search engagements are achieved because of the
strategies, the mesh becomes a shared interface al- broad disciplinary project teams, then the evolution
lowing the translation of information between dif- of the design model tends to be in the hand of the
ferent scales of design engagement (Nicholas 2015). architects. Perhaps this is because geometric repre-
The mesh is no longer only a geometric descriptor sentation traditionally is the hands of the architect or
but instead a dynamically changing infrastructure because it supports the architect's role of orchestrat-
coarsening or refining so as to keep information at ing collaboration. Despite the actual methods being
the right resolution level. highly interdisciplinary coming from computer sci-
ence, engineering, material science or biology, then
CONCLUSION it is architects that are standing for these transfers of
Complex Modelling is one out the many projects knowledge.
across our community to explore these paradigms. We are transforming our practice and present-
Although perhaps unique in its focus on the under- ing a new consideration of design as something in-
lying infrastructure of advanced digital modelling, it herently dynamic, interconnected and incongruous.
is part of larger push to expand and evolve our mod- Here, complexity is not an extravagance but rather
elling practices. Our community is maturing. With the probing of a future to come.
large scale research projects such as the Digital Fab-
rication NCCR at ETH, the Biological Design and In- REFERENCES
tegrative Structures at University of Stuttgart and In- Beukers, A and van Hinte, E 2001, Lightness: the inevitable
nochain at CITA and across 6 European institutions, renaissance of minimum energy structures, 010 Pub-
our field is contributing with real solutions for future lishers, Rotterdam
Burry, M 2011, Scripting Cultures - Architectural design and
building culture.
Programming, Architectural Design Primers, John
However, we need to be much more conscious Wileys and Sons Ltd, London
of the research contribution that the evolution of our Clough, RW and Wilson, EL 1999 'Early finite element re-

search at Berkeley', Proceedings of the 5th US National Tamke, M, Stasiuk, D and Ramsgaard Thomsen, M 2013
Conference Computational Mechanics, Boulder, Col- 'The Rise – Material Behaviour in Generative Design',
orado Proceedings of the 33rd Annual Conference of the As-
Davis, D 2013, Modelled on Software Engineering: Flexi- sociation for Computer Aided Design in Architecture,
ble Parametric Models in the Practice of Architecture, Cambridge, Ontario, pp. 379-388
Ph.D. Thesis, chool of Architecture and Design, Col- Tamke, M, Stasiuk, D and Ramsgard Thomsen, M 2014,
lege of Design and Social context, RMIT University 'The Rise – Building with Fibrous Systems', in Gra-
Holden Deleuran, A, Tamke, M and Ramsgaard Thom- mazio, F, Kohler, M and Langenberg, S (eds) 2014,
sen, M 2011 'Designing with Deformation - Sketch- Fabricate: Negotiating Design and Making, GTA Pub-
ing material and aggregate behaviour of actively lishers
deforming structures', Proceedings of SimAUD 2011, Ramsgaard Thomsen, M, Bech, K and Tamke, M 2011,
Boston, pp. 5-12 'Thaw – Imaging a soft tectonics', in Glynn, R and
Holzer, D 2007 'Are you talking to me? Why BIM is not the Sheil, B (eds) 2011, Fabricate: Making Digital Archi-
answer', Association of Architecture Schools Australa- tecture, Riverside Architectural Press, Canada
sia Conference, University of Technology Sydney Ramsgaard Thomsen, M, Tamke, M, Holden Deleuran,
Jernigan, FE and Onuma, KG 2008, Big Bim Little Bim, 4site A, Friis Tinning, I, Leander Evers, H, Gengnagel, C
Press, Salisbury and Schmeck, M 2015 'Hybrid Tower, Designing Soft
Nicholas, P 2013, Designing Material Materialising Design, Structures', Design Modelling Symposium: Modelling
1st Ed., Riverside Architectural Press Behaviour, Springer, Berlin Heidelberg
Nicholas, P, Stasiuk, D, Clausen Nørgaard, E and Hutchin- Vierlinger, R and Bollinger, K 2014 'Accommodating
son, C 2015 'A Multiscale Adaptive Mesh Refinement change in parametric design', Proceedings of the
Approach to Architectured Steel Specification in the 34th Annual Conference of the Association for Com-
Design of a Frameless Stressed Skin Structure, in puter Aided Design in Architecture (ACADIA) ISBN
Modelling Behaviour', Design Modelling Symposium: 9781926724478, Los Angeles, pp. 609-618
Modelling Behaviour, Springer, Berlin Heidelberg Weinan, E 2011, Principles of Multiscale Modelling, Cam-
Nicholas, P, Tamke, M and Riiber, J 2014 'The Agency of bridge University Press
Event: Event based simulation for architectural de-
sign', Proceedings of the 34th Annual Conference of the
Association for Computer Aided Design in Architecture
(ACADIA) ISBN 9781926724478, Los Angeles, pp. 63-
74
Stanley, KO and Miikkulainen, R 2002, 'Evolving Neural
Networks Through Augmenting Topologies', Evolu-
tionary Computation, 10(2), p. 99–127
Stasiuk, D and Ramsgaard Thomsen, M 2014 'Learning
to be a Vault - Implementing learning strategies for
design exploration in inter-scalar systems', Proceed-
ings of the 32nd International Conference on Educa-
tion and research in Computer Aided Architectural De-
sign in Europe, Volume 1, Northumbria University,
Newcastle upon Tyne, UK, pp. 381-390
Tamke, M, Lafuente Hernández, E, Holden Deleuran, A,
Gengnagel, C, Burry, M and Ramsgaard Thomsen,
M 2012 'A new material practice - Integrating de-
sign and material behaviour', Proceedings of SimAUD
2012, Orlando, FL, USA, pp. 5-13
Tamke, M, Riiber, J, Jungjohann, H and Ramsgard Thom-
sen, M 2010, 'Lamella Flock', in Ceccato, C, Hes-
selgren, L, Pauly, M, Pottmann, H and Wallner, J
(eds) 2010, Advances in Architectural Geometry 2010,
Springer Verlag, Vienna, pp. 37-48

Structural Design
A Systemic Approach
Toni Kotnik1
1
Aalto University
1
toni.kotnik@aalto.fi
The paper sketches out the idea of a systemic approach to structural design.
Starting with the notion of complexity as interweaving of feedback loops it
introduces a thermodynamic understanding of the problem of form-finding and
discusses some implications for the design of building structures.
Keywords: structural design, systemic thinking, form-finding, thermodynamics,

complexity
INTRODUCTION This paper is examining the notion of complexity

Over the past twenty years, the introduction and in- and its relation to computation more carefully and is
tegration of computational methods and techniques sketching out some of its implication for architectural
into architectural design and discourse has been par- thinking, exemplified by changes to structural design
alleled by the almost ubiquitous presence of the no- as a sub-discipline.
tion of complexity in connection with applications
of the digital: complex architectural geometry, com- WOVEN COMPLEXITY
plex fabrication, complex assembly, complex envi- The etymological roots of the word complex can be
ronmental performance, or complex urban analysis traced back to the Latin complecti, from com- (with,
to name but a few. In this context, complexity is used together) and plectere (to weave, to twine). In its
synonymously as indication of an advanced level ap- original meaning, therefore, complexity refers to an
plication to an otherwise well-known topic within ar- activity of weaving together independent threads
chitecture. into a new organized whole. This bringing together,
Such a connotation of the notion of complex- however, is not a simple additive process as the ac-
ity as advanced level application, however, stabilizes tual weaving of material illustrates: the weft and the
our current understanding of architecture. It veils the warp interact with each other, both threads deform
mind-changing potential of the digital and limits the and connect to one another by friction.
conscious recognition of computation as a paradig- It is the coordinated accumulation of this active
matic shift within the discipline. Architecture is tak- interaction that results in new properties of the over-
ing part in an "intellectual revolution [that] is happen- all configuration that is not apparent in the threads
ing all around us, but few people are remarking on themselves. Such emergent properties like a specific
it. Computational thinking is influencing research in spatial morphology or structural rigidity are used for
nearly all disciplines, both in the sciences and the hu- example by the artist Joe Hogan in the creation of
manities. . . . It is changing the way we think" (Bundy, sculptures based on traditional basket weaving tech-
2007).

Figure 1
Principle of
weaving based on
embedding of weft
thread into
collection of warp
threads and
examples of
possible weaving
patterns; From
Earth And Sky,
woven sculpture,
2012, by Joe Hogan,
Co Galway, Ireland
(Photo: Rory Moore,
Belfast, Northern
Ireland).
niques (Figure 1). Here, especially the resulting over- Figure 2

all shape of the sculpture is often hard to predict as Translation of a
the bending stiffness of the willow changes along the simple warp-weft
thread as well as from thread to thread. configuration into
In general, the weaving pattern introduces an in- the related
terdependency between neighboring threads which connectivity matrix
causes a feedback mechanism that influences the and according
With this in mind, the notion of complexity has to
precise formation of each individual thread (Figure 2). directed graph.
be understood essentially as a systemic notion. It
That is, weaving can be seen as the construction of a relates to pattern of interwoven simple relationship
network of threads with a large number of overlap- between a larger set of agents that allow for feed-
ping and interacting feedback loops. It is this feedback loops as a form of exchange of information be-
back mechanism that guides the formation process tween the agents. Complexity describes the ability
as visible expression of the regulatory behavior and of a set of agents to organize itself into a new whole,
distribution effect caused by the circular interaction. of phase change, of re-organization into a new unity.
An effect, explored in depth first by Norbert Wiener in Complexity, therefore, fundamentally is a transitional
his seminal book Cybernetics or Control and Commu- phenomena at the transformational stage between
nication in the Animal and the Machine published in hierarchies of organization. It refers to the ability in-
1948 (Wiener). Together with General Systems The- herent in a system for organized construction, a con-
ory, established by Ludwig von Bertanlanffy in the struction of morphology or structural stability like in
1940s, Wiener's Cybernetics provided the foundation the example of the woven sculpture by Joe Hogan.
for the development of systemic thinking and sys-
tems theory up to the present (Ståhle, 2009).

Figure 3
Paradigmatic shift
in science from
reductionism to
systemic thinking
and related
concepts of
exploration that
have been taken up
in digital
architectural design
with some temporal
delay.
SYSTEMIC BEHAVIOR due to potentially new methods of spatial organi-

Systemic thinking is a holistic thinking that caused zation and form-generation provided by computers
a paradigmatic shift in science by questioning the and appropriate software (Weinstock, 2010). As a re-
universality of reductionism as scientific method of sult, over the past decade, systemic notions and con-
exploration. Since the 1950s, this shift has been cepts from science have diffused into architectural
triggered by the availability of computational means discourse and are currently being explored for design
which made is possible to handle and explore a large purposes.
set of interaction between various quantifiable enti- One of the main applications of the self-
ties (Mussmann, 1995). The possibility to explore in regulating capacity of systems has been in the devel-
more depth the self-regulating effect of feedback has opment of computational tools for form-finding of
resulted in a successive modification or even replace- long-span roof structures and lightweight structures,
ment of reductionism as the predominant paradigm a classical problem in structural design, at the con-
of research thinking. That is, the mechanistic under- ceptual stage of the design process. Such structures
standing of nature and the continuous top-down re- transfer their loads purely through axial or in-plane
duction of the whole into parts has been exchanged forces and the shape is determined primarily by the
from patterns of local interaction to the overall global flow of forces in space. This distribution of forces in
arrangement of the parts as an emergent bottom- space is typically not know in advance and therefore
up property of the system (Figure 3). Starting in a formation process is required (Veenendaal & Block,
the late 1980s it is not surprising that architects be- 2012).
came interested in these systemic models of nature

Figure 4
Material system
with multiple stable
equilibrium
positions. The
material system is
based on a
membrane covered
by
triangular-shaped
panels resulting in a
hinged plate
Before the availability of computational tools, physi- The reason for this limitation is that particle-spring system with limited
cal experiments were used to determine the shape. systems are highly ordered systems: the relation be- angular range.
Probably the best-known examples are the hang- tween neighboring particles is very restrictive and Through the
ing chain model by Antonio Gaudi for the Colònia aims for balancing out of uneven force distribution lightning effect at
Güell in Barcelona or Heinz Isler's membrane models resulting in pure compression or tension within each the interior zones of
for thin-walled shell buildings (Kotnik, 2011). These neighborhood of a particle and no rapid change of compression and
experimental form-finding methods can be approx- the surface geometry. This means all neighborhoods tension can be
imated computationally by particle-spring systems, of the surface have to be similar in their geometric identified visually
a multibody system method (Kilian & Ochsendorf, quality. If, on the other hand, compression and ten- (Aalto University,
2005). One of the first tools created to explore sion are available within the same neighborhood of Creative Design in
particle-spring systems as a method of form-finding a particle then rapid change of the surface geometry Engineering,
was CADenary, a simulation tool built mainly by Axel is possible and various equilibrium solutions of the Students: Iulia
Kilian in 2002. It gained popularity due to the intu- system can exist (Figure 4). This means, the systemic Radion, Joonas
itive understandability and the interactive character behavior of structures in general is much more intri- Jaaranen, Umar
of the particle-spring system but lacked features to cate than the existing methods of form-finding are Riaz).
make it into a versatile design tool. Some of these de- able to capture. Compression and tension function
sign limitations have been overcome by Kangoroo, a as pushing-and-pulling effect in the formation pro-
popular physics engine for Rhinoceros developed by cess of structural systems and it is well-known from
Daniel Piker that allows an interactive simulation and the study of dynamic systems in physics and biol-
form-finding (Piker, 2013). ogy that it is the simultaneity of pushing and pulling
One of the big shortcomings of the particle- within a system which is responsible for the emer-
spring approach is that the described method of gence of new phenomena and configurations (May-
form-finding is only valid for tension-only respec- field, 2013).
tively compression-only structures like membranes
or shells. In more general cases, the underlying dy- THERMODYNAMIC FORM-FINDING
namic relaxation process does not converge. The In order to overcome the limitations of the existing
same limitation is true for other existing methods of methods of form-finding it is, therefore, suggested
form-finding in structural design like the Thrust Net- that a new systemic approach to structural design is
work Analysis (Block & Ochsendorf, 2007) embedded required for a better understanding of formation pro-
into Rhinovault, a form-finding software by Philippe cess in structural systems.
Block. Starting point of such a new approach is the

Figure 5
The transformed
free body diagram
of static equilibrium
of the hanging
chain model as
open system
diagram and
alternative force
flow within system observation that in engineering the understanding structural systems.
which result in of the system under consideration is defined by the As mentioned above: one of the essential pre-
bending. free body diagram, an isolated description of the requisites for the existing form-finding methods is
part of the structural system and all the forces act- the assumption of flow of forces through the sys-
ing upon it (Muttoni, 2006). The primary goal of this tem purely through axial or in-plane forces. On one
diagram is the detailed examination of the force dis- hand, this enables an efficient use but on the other
tribution within the structural system at static equi- hand it reduces the typology of shapes to a mini-
librium. Hence, it describes a fixed situation, the sit- mum. If more material is introduced into the sys-
uation at the end of the formation process. Form- tem then other types of inner force flow get possi-
finding, however, is a dynamic process and is driven ble like in the case of a beam where non-axial forces
by constant change of the configuration. In the case lead to the simultaneity of compression and tension
of the particle-spring system this change is provoked in the same material section and consequentially to
by a residual force at each of the particles, a lumped bending of the beam (Figure 5). In the case of the
mass, that introduces some kinetic energy into the particle-spring system residual forces dissipate into
system (Figure 5). The process of change continues movement, in the case of the beam energy does not
till all the energy has dissipated and structural equi- dissipate out of the system but remains within, stored
librium is achieved. By introducing this residual force as elastic deformation respectively bending stress.
and the initial configuration into the diagram, the Thermodynamically, bending can be under-
free body diagram transforms into a diagram of an stood as dissipation of energy within the system that
open system with the external forces as exchange of is the resulting deformation is a morphological re-
energy of the system with the surrounding and the action of the system to the flow of energy through
residual forces as internal energy flow. This reading of space. This interpretation seem to be farfetched.
the diagram introduces a thermodynamic perspec- However, locally compression and tension define op-
tive into structural design and with it the interpreta- posite directionalities. In a material section with com-
tion of the above mentioned form-finding methods pression and tension like for example in the redirec-
as near-equilibrium systems. tion of the force flow around a cut the resulting bend-
Such systems tend to evolve towards equilib- ing can be seen as a local turbulence pattern around
rium, a special state that has been the focus of multi- the cut. This pattern of energy dissipation resembles
body research for a century. Yet much of the richness the flow of wind around an obstacle like for example
of the world arises from conditions far from equilib- a wall (Figure 6).
rium (Schneider & Sagan, 2005). Phenomena such as Consequently, the inner flow of forces as phe-
turbulence, earthquakes, fracture, and life itself oc- nomena of mechanics shares some similarity with the
cur as phenomena far from equilibrium. Subjecting flow of energy as phenomena of thermodynamics.
materials to conditions far from equilibrium leads to An insight that forms the basis for constructal the-
otherwise unattainable properties. This is also true in ory, a general theory on the formation of pattern

Figure 6
Based on plasticity
theory the flow of
forces can be
manipulated
following simple
geometric
operations like the
redirection of
forces: The direct
flow of inner forces
between two
external loads (a)
within flow systems (Bejan & Zane, 2012). But it also primarily along linear elements. This has fostered a can be redirected
an insight that should motivate the interpretation of move from surface to line that is the replacement of within the material
structural systems as distributed systems of energy massive walls by linear structure and the develop- by adding an
in space. Such a perspective would relate the prob- ment of building elements and fixed typologies as internal force that
lem of form-finding much closer to research in sci- structural solutions. The implied tendency towards pushes the flow out
ence like physics and biology and especially with re- lightness and dematerialization of structure has di- of its preferred line
search into pattern development in far from equi- minished the potential of building structures to func- and compensating
librium open system. On the scale of urban design tion outside of the mono-functional purpose of pro- the additional force
such a perspective already has been put forward al- viding structural stability. to ensure
ready (Pulselli & Tiezzi, 2009; Wilson, 2009). On the Within a systemic paradigm building structures equilibrium in the
building scale a thermodynamic perspective within are perceived differently. Based on a thermodynamic system (b). The
research and design is only slowly emerging (Bra- ambition building structure should be spread into resulting
ham, 2016; Moe, 2013) but not specifically targeted space in order to maximize the possibility for the configuration of
towards structural design. even distribution of loads. Animal architecture like forces can be
for example bird nests exemplify this very well (Figure understood as
7): out of the interweaving of soft and flexible mate- pattern of turbulent
CONCLUSION: DESIGN OF STRUCTURE
rial structural stability emerges as complex property dissipation of
Clearly, the implied similarity of flow pattern still re-
like in the initial example of the basket weaving. At energy (c). The
quires a lot of in depth research before it can be made
the same time the weaving enables a flexible adapta- discrete turbulence
operative and a thermodynamic approach to struc-
tion of the material system to the environmental con- pattern in force
tural design is only sketched out in a very rough way.
ditions and the spatial requirements. This step-by- flows resembles in
But it already starts to point towards a possible direct
step construction is a formation process that results an abstract way the
link between ongoing research within science and ar-
in a structural system that not only fulfils a mono- patterns of energy
chitectural design based on formation process. And
functional purpose but is also the generator of the dissipation - visible
even at this infant state of development some con-
space as well as shading device and thermal regula- in the local increase
clusion can made regarding the implications for our
tion. of wind speed - that
current understanding of structural design.
A system design of structural systems aims at occur in the flow of
At the moment, teaching and design of building
such kind of polyvalent use of the building structure. wind around an
structures still is based on a reductionistic paradigm
Systemic structural design, therefore, has to be non- obstacle like a wall
of hierarchy of structural systems governed by the
hierarchical and cannot be based on typologies but (d).
idea of accumulation of forces from top to bottom

rather has to be based on first principles as genera- Energy, Routledge
tive driver of the design. Systemic thinking in struc- Mussmann, F 1995, Komplexe Natur - Komplexe Wisssen-
tural design is the move away from the static iconog- schaft: Selbstorganisation, Chaos, Komplexität und
der Durchbruch des Systemdenkens in den Naturwis-
raphy of form towards the dynamic and procedural of
senschaften, Leske + Budrich
formation. It is the move away from structural design Muttoni, A 2006, The Art of Structures, EPFL Press
towards design of structures! Piker, D 2013, 'Kangoroo: Form-Finding with Compu-
tational Physics', AD Architectural design, 83(2), pp.
Figure 7 132-133
Nest of a Baya bird Pulselli, RM and Tiezzi, E 2009, City out of Chaos: Urban
in Southeast Asia Self-organization and Sustainability, WIT Press
and nest construct Schneider, ED 2005, Into the Cool: Energy Flow, Thermo-
dynamics and Life, University of Chicago Press
of Social Weaver in Ståhle, Pirjo 2009, 'The dynamics of self-renewal', in A,
South Africa. Bounfour (eds) 2009, Organisational Capital – Mod-
elling, measuring, and contextualizing, Routledge,
pp. 119-145
Veenendaal, D and Block, P 2012, 'An overview and
comparison of structural form finding methods for
general networks', International Journal of Solids and
REFERENCES Structures, 49, p. 3741–3753
Bejan, A and Zane, JP 2012, Design in Nature: How the Weinstock, M 2010, The Architecture of Emergence: The
Constructal Law Governs Evolution in Biology, Physics, Evolution of Form in Nature and Civilisation, Wiley &
Technology, and Social Organization, Doubleday Sons, Oxford
Block, P and Ochsendorf, J 2007, 'Thrust Network Anal- Wiener, N 1948, Cybernetics or Control and Communica-
ysis: A New Methodology for Three-Dimensional tion in the Animal and the Machine, MIT Press
Equilibrium', Journal of the IASS, 48(3), pp. 167-173 Wilson, A 2009, 'The “Thermodynamics” of the City: Evo-
Braham, WW 2016, Architecture and Systems Ecology: lution and Complexity Science in Urban Modelling',
Thermodynamic Principles of Environmental Building in Reggiani, A and Nijkamp, P (eds) 2009, Complexity
Design, Routledge and Spatial Networks, Springer
Bundy, A 2007, 'Computational Thinking is Pervasive',
Journal of Scientific and Practical Computing, 1(2), pp.
67-69
Hogan, J 2011, Bare Branches, Blue Black Sky, Musgrove
Willows
Kilian, A and Ochsendorf, J 2005, 'Particle-Spring Sys-
tems for Structural Form-Finding', Journal of the IASS,
46(2), pp. 77-84
Kotnik, T 2016, 'Geometric Diagrams of Energy Flow', in
García-Germán, J (eds) 2016, Thermodynamic Inter-
actions, Actar, pp. 125-132
Kotnik, T 2011, 'Experiment as design method: On the
possibility of the integration of the methodology of
the natural sciences in architecture', in Moravansky,
A and Kirchengast, A (eds) 2011, Experiments in Ar-
chitecture, Jowis, pp. 24-53
Kotnik, T and Schwartz, J 2011, 'The Architecture of Heinz
Isler', Journal of the IASS, 52(3), pp. 185-190
Mayfield, JE 2013, The Engine of Complexity: Evolution as
Computation, Columbia University Press
Moe, K 2013, Convergence: An Architectural Agenda for

WORKSHOPS
The Wave of Autonomous Mobility:
Architecture Facilitating Indoor Autonomous Navigation
Jaehwan Kim1 , Mathew Schwartz2 , Andrzej Zarzycki3

1,2
Advanced Institutes of Convergence Technology, South Korea 3 New Jersey In-
stitute of Technology, USA
1
jhkim13@snu.ac.kr 2 umcadop@gmail.com 3 andrzej.zarzycki@njit.edu
When considering architectural and urban responses to autonomous mobility, it

becomes evident that the future strategies will have to include a significant
transformation to the built environment, particularly the ways it operates and
interacts with inhabitants. Designers will not only need to rethink formal and
functional arrangements but also, and perhaps primarily, consider the
environment--buildings and cities--as active and equal actors with adaptive and
autonomous behaviors similarly to those people or self-driving cars manifest.
This paper discusses initial planning and design strategies for the integration of
autonomous vehicles and other forms of autonomous mobility into the built
environment. Specifically, it looks into necessary steps required to develop
infrastructure to a level of autonomy that can facilitate a next generation of
wayfinding and mobility. A growing research area into smaller personal mobility
vehicles that would revolutionize elderly and disabled mobility brings to the light
the major technical challenges present in current building infrastructure.
Keywords: Autonomous Vehicle, Navigation, Localization, Smart Buildings,

Smart Infrastructure
INTRODUCTION tects are able to formulate a design-informed and

With the rapid and technology-driven implemen- user-centered vision? The most immediate and di-
tation of autonomous mobility, from assisted rect technology in the relation to architecture is an
wheelchairs to drones and autonomous vehicles, introduction of the autonomous personal mobility
the question of what kind of cities and buildings vehicles (PMVs).
are necessary to facilitate this transformation takes Autonomous and connected vehicles, given
front stage. Are architects and urban planners ready the issues surrounding their implementation, must
to embrace this imminent intervention into the built be tested within real-world conditions (Jiménez-
environment they have stewarded in the past? Will González et al. 2013; Martinez et al. 2013). How-
the future direction and shape of autonomous- ever, these tests often rely on limited-scope devel-
vehicle-enabled cities be established before archi- opment as a necessary step in optimizing known
Workshops - Volume 1 - eCAADe 34 | 53

technologies and coping with unknown limitations, PMV deployment. This would require changes
as is the case with Masdar City, Abu Dhabi.[1] How- to existing urban and building systems, mate-
ever, full integration into current infrastructure re- rials used, and control technologies.
mains a challenge. For this reason, places such as the 3. Broader building design issues, such as a
Mcity research facility in Ann Arbor, Michigan, have building access and navigation for larger
opened (in 2016) to serve as research test beds for amounts of wheeled mobility will require spa-
autonomous and connected cars. Additional testing tial adjustments to horizontal and vertical cir-
grounds are appearing throughout the world, such as culation including the question whether the
in Singapore [2], GoMentum and Willow Run. While continued reliance on elevators for indoor
testbeds are being created to research the city wide movement in appropriate.
aspect of autonomous vehicles, the work done in
PMVs is still in the early stages without significant THE PAST MODES OF QUASI-AUTONOMOUS
precedents or case studies. INFRASTRUCTURE
The history of human civilization is filled with
machines and technologies that assisted people. Figure 1
Some argue that technological development is what Personal mobility
shapes or even defines the progress as well as the vehicles designed
very nature of our culture (Ornstein and Burke 1995). for an in-building
The idea of autonomous, or at least human- and use. This split
vehicle-guiding infrastructure is certainly prevalent image shows a PMV
today, but it has already been around for some time. entering and
For example, much of the navigation on canals in exiting an elevator
Europe and the United States was guided and con- autonomously. The
trolled by outside agents. These could have been vehicle was
While PMVs benefit from many technology transfers horses pulling barges or, during the Industrial Era, a developed in the
coming from self-driving cars and test beds discussed series of elaborate locks, ramps, and rail platforms, Advanced Institutes
earlier, there is a need for the research and case stud- such as those still deployed in the Morris Canal [3], of Convergence
ies specifically addressing the scale of architectural in Great Britain, or the Elbląg Canal in Poland (Figure Technology.
spaces and their users. Consequently, this paper ad- 2) [4]. From early beginnings, railways relied on the
dresses three important urban- and building-design infrastructure of switches to direct train navigation,
strategies for the (future) autonomous personal mo- and more recently, autonomous subway systems in-
bility technology: troduced sensors for speed guidance and train car lo-
calization in respect to each other or to a platform. A
1. The role of infrastructure in supporting mo- number of metropolitan cities, such as Copenhagen
bility has expanded beyond its original appli- in Denmark, operate fully automated car parking fa-
cations in transportation networks. This ex- cilities [5], in which cars are self-parked by a con-
pansion will continue affecting other aspects trolling robot-like structure with no human presence
of physical environments and it will closely or participation. A similar approach has spread into
interconnect transportation and building in- the airplane industry, where on-ground infrastruc-
frastructures. ture significantly assists pilot-controlled landings,
2. The advancement of PMV technology (Figure particularly in limited-visibility situations when auto-
1) is highly contingent on the availability of landing may be preferable [6]. A question is; to what
supporting infrastructure that would be able point and to what extent similar autonomy-assisted
to communicate with or, at least facilitate, the functionalities applicable to buildings should be con-
54 | eCAADe 34 - Workshops - Volume 1

sidered? Considering the impact of safety eleva- accompany pedestrian crossings on busy streets to
tors on the development of high-rise buildings in the compensate for the static nature of tactile paving.
late 19th-century America [7] and the following high- In these cases, an outside building and city infras-
density urbanization in many cities around the world, tructure provides a framework for enhancing mobil-
an introduction of PMVs could similarly redefine what ity and safety. Similar strategies, such as guiding
is possible in cities and architecture. floor patterns, could easily be applied to autonomous
forms of mobility, from wheelchairs and personal
Figure 2 mobility vehicles to construction and maintenance
A boat transfer robots and self-driving cars (Jeon et al. 2015). Au-
between two levels tonomous vehicles navigating a city or following a
of the canal. path, even when the environment is known, may
benefit from these forms of guidance in the same way
vision-impaired people may know the general area
of where they are but need an extra tactile path to
precisely locate an elevator door, water fountain, or
building exit. The significant difference between tac-
tile guides (Figure 3) and embedded sensory guid-
ance systems is that the latter can change real-time
AUTOMATED BUILDINGS and adapt to various occupancy scenarios. For exam-
In addition to elevators, there are many other in- ple, elevators that are not in service would have deac-
tegrated automated systems present in contempo- tivated guides to avoid possible "signage" confusion.
Figure 3 rary buildings, such as motion-sensors (pyroelectric

Tactile paving infrared, or PIR), actuated lights, automatic doors
provides guides for with PIR, vision, floor pressure sensors, or sensor-
vision-impaired controlled water flow in public showers, where users
building occupants. often demonstrate lower resource conservancy. All
of these examples speak to the increased role that
building infrastructure plays in defining occupancy
patterns and safety. Occupancy and assets tracking
becomes a critical factor in operations of large-scale
infrastructure, such as airports, with a close integra-
tion between data BIM models and real-time moni-
toring (Kim et al. 2015). These models could be fur-
Furthermore, PMV guidance systems could also be
ther extended into future guidance systems for au-
used by visually impaired in conjunction with mo-
tonomous mobility and real-time resource and user
bile devices to more effectively and effortlessly move
management.
around the building.
Similarly, an increased focus on inclusive envi-
ronments transforms many aspects of urban infras- PMV NAVIGATION STRATEGIES
tructure and technologies used in buildings (Williams
Materials
et al. 2016; Parkin et al. 2012). Tactile paving provides
Methods for a PMV to 'see' the world largely focused
guidance to vision-impaired building occupants (Fig-
around 3-dimensional data, 2-dimensional cameras,
ure 3) [8] by demarcating travel paths and access
and magnetic sensors. While an autonomous car
points. For the same reason, auditory guides often
may rely heavily on GPS data, the indoor nature of

a PMV makes common GPS unusable. While other model to complete a wayfinding task. As buildings
methods for indoor localization exist, such as motion continue to be made out of glass facades and floor
capture or UWB networks, both of these require high to ceiling glass walls, the use of a LIDAR for naviga-
costs for installation throughout an environment. As tion becomes increasingly difficult, and must rely on
more Internet of Things (IoT) devices and connectiv- other sensors. On the other hand, if a standard can
ity between users and infrastructure develop, tech- be decided to which all glass walls or facades have
nology based infrastructure will provide more use a specific band allowing LIDAR to track these reflec-
than for only a PMV. Until then, a few considerations tive or clear objects, the autonomous vehicles can be
during the design process can help alleviate the chal- designed to expect this information. This could be
lenges and delays in deploying PMVs into the build- integrated with current code requirements that re-
ing. quire marking large glass surfaces with visually dis-
tinct markers to avoid human-to-glass collisions.
To give a PMV a path to follow, two methods, in- Figure 4
dependent of each other, can be used. In the first Diagram of a tiling
case, a magnetic sensor on the vehicle can detect system in which
floor embedded magnets. The strength of the mag- colored tiles
net is directly relational to the height of the PMV sen- provide a path to
sor off the floor, and the materials between them. follow. Intersecting
The cost of high powered magnets can become pro- paths blend colors
hibitive for implementing under flooring, while low and can be thought
powered magnets may pose an aesthetics issue due of as transfer
to a limited number of cover materials. Addition- stations. The map
ally, implementation of this could be similar to metal can be stored as a
tracks embedded in asphalt, and conceptually would 2D representation
work similar as a fixed track with intersections of- of the building
fering an option for direction. As lightweight and space without
durable floorings become available, architects and requiring a full 3D
engineers can test the efficacy of using these mag- map.
One of the most relied on sensors is the LIght nets in a building for the expectation of autonomous
Detection And Range Sensor (LIDAR) to receive vehicles and PMVs in the future.
the 3-dimensional data. This sensor captures 3- Finally, one of the simplest methods for naviga-
dimensional points based on the distance a point is tion is line tracking. The initial problem with a cam-
from the device. Through multiple channels and a era based system is the effect of ambient light on
sweeping method, a cloud of points can be gener- color. However, built-in under body lights of a PMV
ated quickly, giving a rough 3-dimensional map of can alleviate this issue to create a consistently lit line.
the surrounding space. This can be used by itself, This method can be created cheaply by painting lines
or more commonly as test dataset to compare to a on flooring, similar to how road lines are painted, or
known 3-dimensional model of the building in order more creatively and long lasting, by embedding the
to find the relative position. The real-time and real- information into the flooring materials itself.
location aspect of LIDAR allows for tracking of peo- One possibility is to make a specific line for a PMV
ple or other obstacles that are not within the building to follow. Using colored tiles, lines can represent a
model. The downsides of a LIDAR system are the poor path for the PMV to take, and can be reduced to spe-
quality in detecting reflective or glass surfaces, and cific locations that the PMV can travel to. A simple
the need for matching to an existing 3-dimensional

Figure 5 the primary access point for various floors become a
In an alternative to hindrance, and a ramp similar to those in a parking
a direct line structure become the logical choice. A simple calcu-
representing a path, lation based on general values provides initial insight
a coded system to the benefits of using the ramp type structures used
could be in car parks. Using a minimum turning radius of 6 me-
implemented in ters, a ramp able to accommodate a two-way flow of
which a given PMVs is evaluated. This type of ramp would have a
section represent a 16m diameter with a 4.2m width, compared to esti-
unique pattern mates of current infrastructure in Table 1.
within the overall While the ramp model requires significantly
map. Autonomous more space, this number should be considered
vehicles would have against the speed at which mobility exists within the
the ability to freely space, especially as general elevators rated for higher
travel throughout weight will travel slower. In this example case, a
the space while multi-floor building is assumed, where the average
maintaining a example is shown in Figure 4.
PMV travels to the 5th floor (Figure 6).
localization To give a PMV more freedom in movement while
This calculation, however, does not take into ac-
accuracy equal to maintaining localization, a coded floor can be cre-
count multi-directional movement that will often oc-
the unique pattern ated. Similar to the unique combinations of barcodes
cur between the floors, in which case the ramp ef-
size. or QR codes, a tiled floor can be generated to give
ficiency would further increase. Beyond the effi-
unique groupings a unique position within the space
ciency for transportation, while elevators are safe due
(Figure 5).
to physical stops, an elevator may be "out of com-
Table 1 mission" for structural or electrical issues, whereas a
Required Space for ramp is solely reliant on its structural integrity.
Infrastructural
Transportation
DISCUSSION AND CONCLUSION
Autonomous vehicle-enabled cities can be seen as
Methods.
the next level of human cohabitation that considers
a diverse range of users (human and nonhuman) and
provides a higher level of services and efficiencies.
Rather than matching LIDAR data to a 3-dimensional For some, it will be a more sustainable and resilient
model, a simple algorithm for detecting a cluster of environment that will optimize resources; for others,
tiles within the larger map can give a PMV ultimate a more inclusive and democratic environment that
flexibility within the space while remaining localized. will allow them a greater participation in urban life,
The size of the tile, resolution of the camera, and den- as in case of mobility-impaired or elderly residents.
sity of the coded pattern can determine the variety of For others yet, it will be a place for new interactions
designs. with similar transformative qualities as social net-
works have. The inclusion, and possibly even eman-
Indoor Mobility cipation, of nonhuman agents as new urban inhabi-
As PMVs begin integration with indoor infrastructure, tants will definitely reposition our understanding of
the mobility methods of car parks and multi-level cities and the way architects design them. There will
garages become of interest. Specifically, one must also be a change to what we consider privacy, iden-
consider at what point does the use of an elevator as tity, and control.

REFERENCES Figure 6
Jeon, D, Kim, Ck and Kim, J 2015 'Dynamic Object Track- Speed ratio of
ing System', 2015 12th International Conference on elevator to ramp
Ubiquitous Robots and Ambient Intelligence (URAI), per square meter.
pp. 512-515 At roughly 14
Jiménez-González, A, Martinez-de Dios, JR and Ollero, A vehicles the size to
2013, 'Testbeds for Ubiquitous Robotics: A Survey',
Robotics and Autonomous Systems, 61(12), pp. 1487- speed ratio of a
1501 ramp system
Kim, E, Yun, J and Cho, S 2015 'Integrated Space and becomes more
Asset Management System for Large Scale Airport', efficient, and
Emerging Experience in Past, Present and Future of continues for each
Digital Architecture, Proceedings of the 20th Interna- extra vehicle.
tional Conference of the Association for Computer-
Aided Architectural Design Research in Asia (CAADRIA
2015), Daegu, pp. 807-816
Martinez, DSH, Gonzalez, M, Huang, X, Irvine, B, Hsieh,
CH, Huang, YR, Short, MB and Bertozzi, AL 2013,
Further, the increased connectivity between au- 'An Economical Testbed for Cooperative Control and
tonomous agents and the built environment open Sensing Strategies of Robotic Micro-vehicles', in Fer-
new possibilities for design workflows. As BIM ex- rier et al., J.-L. (eds) 2013, Informatics in Control, Au-
tomation and Robotics, Springer Berlin Heidelberg,
pands from building planning to lifecycle mainte-
pp. 65-75
nance, the connectivity provided by autonomous ve- Ornstein, R and Burke, J 1995, The Axemaker's Gift, G.P.
hicles can expand to the construction phase. While Putnam’s Sons, New York
researchers are tackling issues of robotic fabrication Williams, MA, Dubin, B, Amaefule, C, Nguyen, L, Ab-
and pre-fab, the use of ramps in the final construction dolrahmani, A, Galbraith, C, Hurst, A and Kane, SK
can just as well provide a new industry for on-site fab- 2016, 'Better Supporting Blind Pedestrians and Blind
Navigation Technologies Through Accessible Archi-
rication in which robotics can travel throughout the
tecture', in Langdon et al., P (eds) 2016, Design-
space through similarly using the ramp. ing Around People, Springer International Publish-
There is a need for continued research in the area ing, Cham, pp. 237-246
of the PMV integration within buildings as they pro- [1] http://www.crankandpiston.com/on-the-road/all-a
vide an extremely accessible mode of transportation board-the-future-self-driving-car-abu-dhabi/
for a variety of users. Beyond people with disabili- [2] http://www.jtc.gov.sg/industrial-land-and-space
/pages/one-north.aspx
ties, larger buildings such as airports, complex set-
[3] https://en.wikipedia.org/wiki/Morris_Canal
tings such as shopping malls or grocery stores, and [4] https://en.wikipedia.org/wiki/Elbl%C4%85g_Canal
other structures in which a user needs fast access to [5] http://www.alectia.com/en/projects/copenhagen-o
a small portion of a building can enhance user inter- pens-its-first-fully-automated-car-park-en/
action with the use of PMVs. [6] https://www.quora.com/How-often-are-commercial
-flights-landed-using-autopilot
[7] http://nypost.com/2014/02/08/how-elevators-tran
sformed-nycs-social-landscape/
[8] https://en.wikipedia.org/wiki/Tactile_paving#/m
edia/File:Tactile_pavers_Japan_yellow.jpg

Contextualized Digital Heritage Workshop - Oulu
Danilo Di Mascio1 , Anetta Kepczynska-Walczak2 , Nicholas Webb3
1
Northumbria University 2 Lodz University of Technology 3 University of Liverpool
1
https://www.northumbria.ac.uk/ 2 http://www.p.lodz.pl/en
3
https://www.liverpool.ac.uk/
1
danilo.dimascio@northumbria.ac.uk
2
anetta.kepczynska-walczak@p.lodz.pl 3 N.Webb1@liverpool.ac.uk
Constructions pertaining to built heritage represent precious material and

cultural resources to be studied and preserved for present and future generations.
In recent years, this built heritage is increasingly understood, documented,
managed, analysed and disseminated through the application of several digital
technologies; this trend has created one of the main research areas in
architecture, called Digital Heritage. Digital technologies open up alternative
and new possibilities in the study of tangible and intangible features of built
heritage. This workshop aims to create a space to allow scholars to discuss, share
and apply their knowledge in a digital heritage exercise contextualized in Oulu.
Keywords: Digital Heritage, Context, Built Heritage, Digital Technologies
BACKGROUND AND MOTIVATION and disseminated through the application of several

Built heritage represents a precious material and digital technologies; this trend has created one of
cultural resource to be studied and preserved for the main research areas in architecture, called Digi-
present and future generations. Italo Calvino, the fa- tal Heritage. Digital technologies open up alternative
mous Italian writer, wrote that literature classics have and new possibilities in the study of tangible and in-
always something new to say (Calvino 2013), this is tangible features of built heritage. They don't substi-
why we are still studying them. Remarkable build- tute traditional methods and tools but complement
ings pertaining to built heritage can be compared to and enrich them; for example, freehand analytical
the literature classics (Di Mascio 2015); they continue sketches can represent an invaluable support during
to inspire people's projects and lives also after cen- 3D digital reconstructions because they allow to in-
turies. Built heritage is a key element in defining and vestigate and clarify complex or missing parts. The
understanding local identity. It reflects not only the huge variety of constructions pertaining to cultural
importance of tangible objects but also the intangi- heritage is studied using a likewise wide range of dig-
ble image of a society and its attitude towards pro- ital technologies and methods such as: laser scan-
tecting the past (Kepczynska-Walczak and Walczak, ning, CAAD, 3D modellers, game engines, BIM and
2013). In recent years, constructions (buildings, mon- Semantic Web technologies. The use of these digi-
uments) pertaining to cultural heritage are increas- tal technologies is always supported by the collection
ingly understood, documented, managed, analysed and analysis of historical information and by theoret-

ical and methodological frameworks, influenced by a heritage workshop to a local context in a multidisci-
large number of architectural theories and concepts. plinary way.To summarize, we seek to:
One of the main points that characterizes built her-
itage compared to most of the contemporary con-
structions is the strong connection with a specific • Create a space to allow scholars to discuss,
physical and cultural context. For this reason, un- share and apply their knowledge on case
like most of the traditional workshops on this topic, studies located in Oulu;
the main idea that underlies this workshop proposal • Build a community of scholars interested in
is linked with a contextualized practical exercise in digital heritage;
Oulu, Finland. Every tangible and intangible feature • Plan future activities on digital heritage.
of a heritage construction can be documented and
analysed in a multitude of ways using several con-
cepts, theories and digital tools. The participants will A poster will be prepared to show to the conference
be asked to creatively use their diverse knowledge attendees the different methodologies and tools
and experiences in the local context to deal with spe- elaborated by the workshop participants for each
cific buildings' features. The workshop could also case study. Concepts and methodologies elaborated
produce a project for the local historic heritage com- during the workshop will be presented in a confer-
munity. This event is thought to represent the first ence paper or journal article.
of a series of future contextualized digital heritage
workshops and hence, the name Oulu interchange- WORKSHOP PLAN
able with the name of any other city or place. The workshop will provide a balance of on-site vis-
its, plenary sessions, breakout discussions and prac-
GOALS AND OUTCOMES tical exercise activities. The event will be split in two
The workshop will represent a place to share con- days. During the first day, the selected constructions
structive and polyvalent discussions on digital her- pertaining to Oulu's built heritage will be presented
itage; participation in a previous digital heritage pro- with the support of a member of the local organizing
ject/research is mandatory.This workshop aims to committee. After the presentations, there will be on-
create a space to allow scholars to discuss, share and site visits to have a direct experience of the building
apply their knowledge in a digital heritage exercise and collect information (such as notes, sketches, pic-
contextualized in Oulu. Moreover, this event also tures, videos) useful for the exercise. During the after-
aims to build a community of scholars interested in noon and the morning of the day after, participants
digital heritage, with a specific focus on architecture. will work in groups of two or three to elaborate their
With the support of the organizing committee, two speculations using their specific knowledge and ap-
different built heritage in Oulu (pertaining to diverse proaches. During the rest of the day, participants will
building typologies) will be selected as case studies present their ideas and receive feedback and ques-
for the exercise. Participants will be asked to use their tions from the other scholars. A final discussion led by
expertise to speculate how to document, manage, the workshop organizers will attempt to synthesize
analyse or disseminate any specific feature of the se- the key points touched during the event and to iden-
lected constructions. In this way, they will have the tify themes and issues for future research and collab-
opportunity to share their different point of views oration, and to develop concrete actions to progress
and approaches on the same built heritage and learn future work in this area (e.g. developing networks, fu-
from each other. As far as we are aware, this event ture workshops). Finally, attendees will be invited to
may represent the first attempt to connect a digital propose new and emerging themes identifying po-
tential plans for publications.

Figure 1
Workshop's
Programme.
PRESENTERS Heritage in the Council for the Implementation of IN-

Danilo Di Mascio, Architect and PhD, is currently SPIRE Directive in Poland. Recent monograph "Cul-
a Postdoctoral Research Fellow at Northumbria Uni- tural heritage management in the knowledge-based
versity (Newcastle, UK). He has undertaken research society". She will be hosting the 2018 eCAADe con-
projects on digital technologies applied in the com- ference.
prehension, analysis and documentation of tangible Nicholas Webb is a PhD, a qualified architect and
and intangible characteristics of cultural heritage, in: lecturer at the Liverpool School of Architecture. He is
Belgium, England, Italy, Scotland and United States. researching the application of contemporary digital
The research include a wide variety of buildings from techniques as a method of analysing historic works
vernacular architecture to masterpieces such as the of architecture, providing new and enhanced infor-
Glasgow School of Art designed by C.R. Mackintosh. mation that would not have been possible in a pre-
He participated and published at various interna- digital era. These applications are investigated in re-
tional conferences in Europe, Middle and Far East; at lation to designs that were destroyed or unbuilt, for
the Chinese University of Hong Kong (2010), he re- example using immersive VR to augment our under-
ceived the Young CAADRIA Award. standing of Jeremy Bentham's Panopticon prison, as
Anetta Kepczynska-Walczak, Associate Profes- well as buildings that still exist; such as studying me-
sor and Head of Department of Digital Technologies dieval vault design and construction in the British
in Architecture and Urban Planning at Lodz Univer- Isles for the Tracing the Past project.
sity of Technology, Poland. M.Sc. in Architecture;
Pg.Dip. in Computer Aided Building Design and Ph.D. REFERENCES
at the University of Strathclyde in Glasgow, UK; 2014 Calvino, I 2013, Why Read the Classics, Penguin, London,
habilitation. Acted as Director Commissioner for the United Kingdom
implementation of the National Heritage Database at Kepczynska-Walczak, A 2013 'Performing the Past and
the National Heritage Board of Poland and as Rep- the Present for the Knowledge of the Future', Pro-
resentative of the Ministry of Culture and National ceedings of the 31st eCAADe Conference – Volume 2,

Faculty of Architecture, Delft University of Technol-
ogy, Delft, The Netherlands, 18-20 September 2013,
pp. 453-462
Kepczynska-Walczak, A and Walczak, B 2013 'Visualis-
ing »genius loci« of built heritage', Proceedings of the
11th Conference of the European Architectural Envi-
sioning Association, Envisioning Architecture: Design,
evaluation, communication, 23-28 September 2013,
Politecnico di Milano, Milano, Italy, pp. 451-457
Di Mascio, D 2015 'Analytical drawings of architec-
tural built heritage', Proceedings of the 12th Confer-
ence of the European Architectural Envisioning Associ-
ation (EAEA), Envisioning Architecture: Imag, Percep-
tion and Communication of Heritage, Lodz University
of Technology, Lodz, Poland, pp. 197-206

Digital Leadership
Tadeja Zupancic1 , Johan Verbeke2 , Henri Achten3 ,
Aulikki Herneoja4
1
University of Ljubljana, Faculty of Architecture 2 KU Leuven, Faculty of Archi-
tecture Sint Lucas / Aarhus School of Architecture 3 Czech Technical University in
Prague, Faculty of Architecture 4 University of Oulu, Oulu School of Architecture
1
http://www.fa.uni-lj.si/ 2 http://arch.kuleuven.be/
3
http://www.fa.cvut.cz/ 4 http://www.oulu.fi/architecture/
1
tadeja.zupancic@fa.uni-lj.si 2 johan.verbeke@kuleuven.be
3
achten@fa.cvut.cz 4 aulikki.herneoja@oulu.fi
Leadership is an important quality in organisations. Leadership is needed to

introduce change and innovation. In our opinion, in architectural and design
practices, the role of leadership has not yet been sufficiently studied, especially
when it comes to the role of digital tools and media. With this paper we intend to
initiate a discussion in the eCAADe community to reflect and develop ideas in
order to develop digital leadership skills amongst the membership. This paper
introduces some important aspects, which may be valuable to look into when
developing digital leadership skills.
Keywords: digital leadership, research by design, creative practice, design

research, impact
INTRODUCTION In this position paper we outline a number of relevant

Digital tools are part and parcel of contemporary aspects related to "Digital leadership." We hope to de-
practice in architecture and in the academic envi- rive from this a first attempt to define digital leader-
ronment. Everyone is aware of the potential of ship, but we are well aware that this very new con-
tools such as parametric design, Building Information cept can only be developed through collaborative ef-
Modelling (BIM), generative design, and so on. High- fort between academic and practice partners.
profile projects such as the British Museum Great
Court, Smithsonian Institution Courtyard roof, Bei- IMPACT
jing New Airport Terminal Building, Arnhem Central This section is focused to the digital leadership im-
Transfer Hall, and so on could not have been realized pact potentials at diverse levels and scales. The lead-
without such technologies. Characteristic for leading ership position, when achieved, seeks for and offers
offices that are pushing the envelope of architectural a high level of impact on the society - in its essence.
design is that they have an integral view of the archi- This impact includes a better transfer of knowledge
tectural design and the digital tools that enable such between sectors and disciplines, stronger networks.
design - we would call this "digital leadership." Which leadership styles are most effective in this

sense, appropriate for the contemporary society and and how do we seek recognition (Hatleskog 2015)?
in diverse socio-cultural settings? How do we identify recognition in relation to the
The shift from the creative applications of digi- topic discussed? How can we change the commu-
tal tools and communication interfaces towards lead- nity where we are able to identify recognition? How
ing the development of the digital initiative and sup- is this related to the media we use for communica-
port of the design process, from responding to driv- tion? Who can we address if we change the commu-
ing, changes the impact flow essentially. The impact nication mode?
from the digital is shifted to the impact on the dig-
ital realm - and far beyond. The digital realm may MEDIA
become design-led but still and even more powerful. How do we communicate the design ideas with our-
The integrated power of the creative research/prac- selves and with others? Who are these others - in the
tices within the diverse digital domains seems to po- case of digital leadership? How can we achieve the
tentially create an important impact on the society as leadership position in relation to the flexible and re-
a whole. sponsive media use and development?
The knowledge collected in a digital database is The studies of design media have already shifted
highly specialized and still distributed (Kocaturk and from the obsession with the digital to more hetero-
Medjdoub 2011) rather than integrated. How can geneous inclusiveness of the diverse analog options.
the idea of the digital leadership impact the essential On one side the development of the BIM modeling
knowledge integration? It offers a wider view, miss- (intended for professional collaboration) and the po-
ing in the last decades of architectural computing. tential of rapid prototyping both clearly lead to the
This wider view is also a potential for a wider impact. sculptural way of thinking about architecture. Archi-
To the communities of practice and research that we tecture becomes a pure object, a sculpture, a mon-
cannot even imagine. ument. The dimensions of the spatial contextualisa-
However, the identification of the potential and tion often move to other areas, to landscape architec-
actual impact of digital leadership is not as easy ture, to painting. This is the consequence in the cases
as it seems on the first inspection. What is share- where the cognitive visualisation is used more for the
able/transferrable from/to design research through discussion with clients and general public (in urban
practice/practice based research (Rendell 2004, Os- design) than for the self-reflection within the design
terle and Otto 2010, Koskinen et.al. 2011, Schaik process. How can we overcome these tendencies?
and Johnson 2011, Verbeke and Zupancic 2014) and Some further questions: How the use of media
how? How/where can we find the initial and also the influences the design thinking? How do we choose
longer-term evidence of the added value of what we and develop the media in relation to the way we
are doing for diverse communities? How to find and think? What drives the main media flows, what are
monitor the evidence of the quality of the impact? the triggers of the changes (Holder 2015)? How can
The shift to social science research is essential here. we start the leading position in this process, being
We are far from the potentially misleading calcula- aware of the potential shifts within the design think-
tions of the 'impact factor' from the research dissem- ing, that originate from or lead to the media devel-
ination databases. opment? How can we lead the public behaviour of
The initial questions to identify the (potential) creative practices through the media development?
impact are: Who/how/with what do we want to ad-
dress? How can we develop the arguments about
the relevance of what we are doing, especially in the
RESEARCH BY DESIGN
Based on the work of Henk Borgdorff (2010), EAAE
case of a high level of research singularity? Where
produced the following definition of research by de-

sign: knowledge processes need to be facilitated and stim-
"In architecture, design is the essential feature. ulated in order to develop knowledge form the de-
Any kind of inquiry in which design is a substan- sign activities? Which skills and competencies are
tial constituent of the research process is referred to needed for a digital leader?
as research by design.In research by design, the ar-
chitectural design process forms a pathway through INTEGRATIVE KNOWLEDGE
which new insights, knowledge, practices or prod- As described in Verbeke (2013) it is explained there
ucts come into being. It generates critical inquiry are many forms of knowledge which are all important
through design work. Therefore research results are for our actions and understanding of the world. In
obtained by, and consistent with experience in prac- Verbeke and Glanville (2006) it is stated that knowl-
tice."(www.eaae.be) edge is much more than the traditional understand-
Following the Frascati manual, research and ex- ing of explicit written-down knowledge. Tacit knowl-
perimental development is creative work undertaken edge is discussed since Polanyi (1966). The distinc-
systematically to increase the stock of knowledge- tion between mode 1 and mode 2 knowledge was in-
including knowledge of humanity, culture, and troduced in Gibbons et al. (1994). Effable, embedded
society-and the use of this stock of knowledge to de- and embodied knowledge are frequently used to in-
vise new applications (OECD). dicate specific types of knowledge.
The crucial aspect is that knowledge can be Integrative knowledge is developed while iden-
emerging or produced through designing. This is tifying, connecting, synthesizing and demonstrating
in line with the work of Donald Schön (1983) who knowledge and skills that are gaining from all areas of
stimulated practitioners (including but beyond de- life, in our case more specifically in the field of digital
sign) to actively reflect on their activities (design or technology and architectural design.
other) and to consolidate these reflections into more In Peet et al. (2011) it is concluded that focus
explicit knowledge. There reflections can be on two on integrative knowledge leads to increased capac-
levels: a) on the use of digital tools and environments ity to: 1. identify, demonstrate and adapt knowledge
to contribute to a better design and b) on digital tools gained within/across different contexts; 2. adapt to
to support the development and sharing of design differences (i.e. in people and situations) in order to
knowledge. create solutions; 3. understand and direct oneself as
Within digital leadership, it seems beneficial that a learner; 4. become a reflexive, accountable, and re-
digital leaders act on three levels: a) stimulate mem- lational learner; 5. identify and discern one's own and
bers of the organization to consolidate knowledge others' perspectives; and 6. develop a professional
from their individual activities; b) consolidate and digital identity.
share knowledge within the team or group to de- While designing and while directing develop-
velop a deeper understanding and c) brokerage ments, it is crucial to combine and integrated dif-
knowledge between members of the organization ferent types of knowledge. It is well known that
but also bring external knowledge into the organiza- a field developed when implicit and explicit knowl-
tion. edge both develops and interact with each other.
The main questions, which are then arising, in- For a digital leader, it then becomes important
clude the following: which (digital) tools can help to be able to manage these different types of knowl-
members of the organization to develop and share edge and to be able to integrate them in activities,
knowledge? Which knowledge processes need to innovation and development. But, how to do this?
be facilitated and stimulated in order to learn from Are there heuristics? Are there specific ways for doing
the use of digital tools for designing? Which Which this? For facilitating that knowledge get integrated.

That knowledge form different fields and disciplines sign; BIM has yet to reach maturity; reliable costing
gets combined and integrated into new ideas and de- is still far way; and sustainable and safe designs and
signs? What skills are needed to facilitate such pro- processes have not yet been reached.
cesses? Which role can a digital leader play in archi- The most dominant technology to integrate de-
tectural offices? And how can the variety of experi- sign in a comprehensive format is Building Informa-
ences by brought together to develop a deeper un- tion Model. BIM is based on the long-standing vi-
derstanding and contribute to better designs? sion of a shared data and process model formulated
as early as the 1970'ies (Eastman 1980, Kalay 2004,
DIGITAL DOMAINS 66-68). Whereas early visions included all phases of
Leadership in design originates from a keen under- the design process, BIM today is geared towards pro-
standing of all aspects of design. Digital leadership duction of final phase documentation - a task which
is not different in this respect. The complex nature of is of high complexity but occurs after most design
design makes it impossible that any advanced level of work has been carried out. Pushing design and dig-
design can be obtained through digital means only. ital technologies forward then, does not rely solely
Therefore, a "digital leader" not only has knowledge on BIM but on a wide range of additional software
and mastery of digital techniques, but also of pro- and techniques. It is in particular in the selection of
cess management, materials, analogue techniques, relevant techniques within a proper framing (Schön
and so on that we can find in design leaders as well. 1983) that the most interesting advances are made.
What stresses the "digital" in digital leadership how- Within Schön's understanding of design as a se-
ever, is an acute understanding how "digital" can be ries of naming - framing - moving - evaluating, the de-
a unifying set of methodologies and technologies to sign process takes on a guided bottom-up approach.
bring skills and knowledge together. By naming the architect sets the boundaries of the
Computers have been used in architecture for current issue in design; framing sets the actual prob-
about six decades. Throughout this period there lem and way of thinking of the problem - these two
have been a lot of developments, not only in com- steps influence the choice of (digital) design technol-
puter aided architectural design itself, but also in ar- ogy. By moving the architects creates one or more
chitecture (the application area of CAAD), informa- design solutions - with creating also meaning gener-
tion & communication technology (the base tech- ating through some digital tool if necessary, followed
nology for CAAD), and society itself (the context in by evaluation to check whether the process is going
which architecture unfolds). Major advances have in the right direction. This process is not completely
occurred in Human-Computer Interface (the ease by bottom-up as this would imply ad-hoc reaction to
which architects can use CAAD), visualization, com- problems as they become apparent, but it fueled by
plexity and size (the degree to which comprehensive experience and top-down set goals.
models can be built), data exchange (to support in- There is not one single selection of digital do-
formation exchange between experts), reality cap- mains that gets the job done. The skill of the digital
ture, adapting software (through scripting languages leader lies in a confident and fluent command of sev-
like Grasshopper) and output to reality (rapid proto- eral design techniques (digital or not) that best fit the
typing and design-to-production pipelines). current and overall design strategy.
Despite these advances there are also still areas
where progress is lacking. Architects are still slow to LEADERSHIP
pick up simulation technology in design; the concep- Leadership is a term that originates mainly from or-
tual design process and hence it's support remains ganizational sciences. There is a great amount of re-
largely unclear; the same applies for collaborative de- search on the aspect how leadership is compounded

of interpersonal skills and managerial skills (de Vries sign team or organization - therefore on the organi-
et al. 2010), but the amount of research on leadership zational level - being a digital leader by team effort.
in a design context is almost non-existent. Pahl's et
al. (1999) review of 12 years interdisciplinary empiri- CONCLUDING TEXT
cal studies of engineering design in Germany shows This paper initiates a discussion on the role of digital
that leadership in design is not well understood. leadership in current architectural practice and de-
Leadership is often viewed as a negative thing, sign. It introduces 6 key elements: impact, media, re-
where a person is perceived as dominating and set- search by design, integrative knowledge, digital do-
ting a course without consideration of others. In de- mains and leadership itself. These elements should
sign, where often results are achieved through team- be seen as a start of further research on skill, compe-
work, such attitudes may have an adverse effect on tences related to digital leadership and how we can
the quality. Lee and Cassidy (2007) discuss leader- understand and use these skills in practice.
ship in industrial design - they identify among oth-
ers "the leader as a catalyst of organization change"
which seems to be the closest to a "digital leader."
REFERENCES
Adams, RS, Daly, SR, Mann, LM and Dall'Alba, G 2011, 'Be-
The list of good leadership traits at the end however,
ing a Professional: Three Lenses Into Design Think-
includes mostly interpersonal skills (personal char- ing, Acting, and Being', Design Studies, 32, pp. 588-
acteristics, maintain friendship, attitudes and values, 607
leading styles, and proper roles). This list stresses the Borgdorff, H 2010, 'The Production of Knowledge in
team player aspect of good leadership. Artistic Research', in Biggs, M and Karlsson, H (eds)
Adams et al. (2011) stress strategic leadership 2010, Routledge Companion to Research in the Arts,
Routledge, Abingdon, Oxon
in particular in cross-disciplinary team. More specifi-
Buhse, W 2012, 'Changing the Mindset: Die Bedeu-
cally, among others they identify the ability to "mak- tung des Digital Leadership für die Enterprise 2.0-
ing or enabling conceptual connections" and "fa- Strategieentwicklung', in Lembke, G and Soyez,
cilitating systems-oriented strategies or frameworks N (eds) 2012, Digitale Medien im Unternehmen,
that leverage diverse perspectives" as important fac- Springer, Berlin, Heidelberg, pp. 237-252
tors for leadership success. It is precisely in this area Eastman, C 1980 'System Facilities for CAD Databases',
ACM, New York, pp. 50-56
of speculative and multi-facetted work that digital
Gibbons, M, Limoges, C, Novotny, H, Schwarymann, S,
leadership operates as well. Scott, P and Trow, M 1994, The New Production of
In a managerial context the only example we can Knowledge: The Dynamics of Science and Research
find is described in Buhse (2012). Buhse stresses agile in Contemporary Societies, SAGE, London, Thousand
management as an important building block for suc- Oaks, New Delhi
cessful Enterprise 2.0 business model, including as- Hatleskog, E 2015, Public Behaviours: Compilation and
analysis of combined behaviours (report), ADAPT-r
pects like team-based formulation of agenda, goals,
Holder, A 2015, Transformative Triggers: Compilation and
and strategies. He sees the digital leader as mod- analysis of combined triggers (report), ADAPT-r
erator, bridge builder, and network organizer rather Kalay, YE 2004, Architecture’s New Media: Principles, The-
than classical top-down manager. ories, and Methods of Computer-Aided Design, MIT,
We define in our workshop "digital leadership" Cambridge, Massachusetts
on two levels: (a) a particular skill set of a designer Kocatűrk, T and Medjdoub, B 2011, Distributed Intelli-
gence in Design, Willey & Blackwell, Chichester
that allows him or her to push the frontiers of design
Koskinen, I, Zimmerman, J, Binder, T, Redstorm, J and
- therefore on the individual level - being a digital Wensveen, S 2011, Design Research Through Practice,
leader by example; (b) a particular skill set of a per- Morgan Kaufmann, Waltham
son to advance game-changing technologies in a de- Lee, KCK and Cassidy, T 2010, 'Principles of Design Lead-
ership for Industrial Design Teams in Taiwan', Design

Studies, 28, pp. 437-462
Osterle, H and Otto, B 2010, 'Consortium Research, A
Method for Researcher-Practitioner Collaboration in
Design Oriented IS Research', Business & Information
Systems Engineering, 5, pp. 283-293
Pahl, G, Badke-Schaub, P and Frankenberger, E 1999, 'Ré-
sumé of 12 Years Interdisciplinary Empirical Studies
of Engineering Design in Germany', Design Studies,
20, pp. 481-494
Peet, M, Lonn, S, Gurin, P, Page Boyer, K, Matney, B, Marra,
T, Himbeault Taylor, S and Daley, A 2011, 'Fostering
Integrative Knowledge through ePortfolios', Interna-
tional Journal of ePortfolio, 1, pp. 11-31
Polany, M 1966, The Tacit Dimension, Doubleday, Garden
City, NY
Rendell, J 2004, 'Architectural Research and Disciplinar-
ity', Architectural Research Quarterly, 8(2), pp. 141-
147
Schaik, LV and Johnson, A (eds) 2011, By Practice, By
Invitation: Design Practice Research at RMIT, One-
pointsixone, Melbourne
Schön, DA 1983, The Reflective Practitioner: How Profes-
sionals Think in Action, Basic Books (2003: Ashgate),
London (2003: Aldershot)
Verbeke, J 2013, 'This is Research by Design', in Frazer,
M (eds) 2013, Design Research in Architecture, An
Overview, Ashgate, Burlington, pp. 137-159
Verbeke, J and Glanville, R 2006 'Knowledge Creation
and Research in Design and Architecture', EURAU 04.
European Symposium on Research in Architecture and
Urban Design, Marseille 2004�: Considering the Imple-
mentation of Doctoral Thesis in Architecture, Marseille
Verbeke, J and Zupancic, T 2014, 'Adapting to and
adapted by ADAPT-r = Prilagajanje projektu in pri-
lagajanje projekta ADAPT-r : architecture, design
and art practice training-research', AR Arhitektura,
raziskave / AR Architecture, Research, 15(2), pp. 49-52
Vries, RE de, Bakker-Pieper, A and Oostenveld, W 2010,
'Leadership = Communication? The Relations of
Leaders’ Communication Styles with Leadership
Styles, Knowledge Sharing and Leadership Out-
comes', Journal of Business Psychology, 25, pp. 367-
380

Folding The Unfoldable
A Method For Constructing Complex-Curved Geometry With Quad Edge
Panels
Ergun Akleman1 , Negar Kalantar2 , Alireza Borhani3

1
Department of Visualization, Texas A&M University 2,3 Department of Architec-
ture, Texas A&M University
1
http://www.viz.tamu.edu/faculty/ergun 2,3 www.thetranslab.com
1
ergun.akleman@tamu.edu 2,3 {kalantar|borhani}@tamu.edu
This paper explains a method will be used during a workshop for constructing
complex-curved geometry with quad edge panels. In this workshop, we
demonstrate that quad-edge mesh data structure can efficiently be used to
construct complex large shapes. With hands-on experiments, we will show a vast
variety of shapes can be constructed using square, rectangular, parallelogram
and extruded-line shaped panels. In addition, using a system we have recently
developed to unfold polygonal mesh, we will demonstrate how desired shapes can
be constructed by using laser-cut quadrilateral panels. This approach is
particularly suitable to construct complicated sculptural and architectural shapes
from anisotropic materials that can only be bended in one direction.
Keywords: Shape Modeling, Physical Construction, Complex-Curved Geometry,

Digital Fabrication
INTRODUCTION sume that we can use isotropic materials such as pa-

Currently, the most readily available digital fabrica- per or thin metal that can provide forming any devel-
tion tool is flat cutting technology such as laser cut- opable shape.
ters or 3-axis CNC routers. In general, this workshop On the other hand, not all materials are isotropic
introduces a simple method of fabrication, allowing and there is now a strong interest in construction
complex forms with positive or negative Gaussian with anisotropic active materials that can change
curvatures to be constructed from planar surfaces. their shapes when they are activated (Gandhi and
In this workshop, we demonstrate our current Thompson 1992, Otsuka and Wayman 1999). These
progress for the realization of complex 3D surfaces materials have now been applied to medical and
and structures that are assembled from 2D planar aerospace and we predict that they will be common
shapes (Akleman et al. 2016). Our goal is to construct in architectural construction.
physical structures that can be raised and formed in There is one significant restriction with some of
3-space using minimal amount of planar materials. these active materials such as smart laminates made
In current construction processes with panels, we as- from shape memory alloys (SMAs). They cannot pro-

Figure 1
At Texas A&M
University, we had a
workshop for
architecture
students. Each
individual designed
and fabricated a
complex geometry
with quad-edge
panels.
vide all possible developable structures since they present such a framework for modeling the geome-
are anisotropic; i.e. they can bend only one (or few try of planar panels that can self-fold into the desired
directions) direction and almost rigid in the perpen- 3D shape and we demonstrate that these elements
dicular direction (or other directions) (Saunders et al. can be folded to create desired shapes.
2014). In other words, they can provide only gen-
eralized cylinders and cannot provide general devel- METHODOLOGY
opable structures such as cones. In this workshop, we consider physical shapes that
Therefore, the panels we can use can have only are constructed with panels corresponding to edges
a single bending direction. Moreover, since such ac- of the planar meshes. We assume that each edge of
tive materials are much more expensive than paper, initial planar mesh unfolded into a special type of flat
it is also important to reduce waste materials. In con- shaped panel, which we call quad-edge panels (See
clusion, there is now a need for a general framework Figure 3). Using these quad-edge panels, it is possible
that can facilitate economic construction with such to unfold any given polygonal mesh surface into a set
anisotropic planar materials. In this workshop, we of planar pieces that can efficiently be packed since

Figure 2
A Stanford Bunny
structure
constructed with
quad-edge panels.
The construction
took 12 hours to
assemble for 4
people. There are
approximately 600
quad-edge panels
and1200 fasteners.
they consist of a series of trapezoids without even us- area. Therefore, it is possible to obtain efficient pack-
ing simple rectangular packing algorithms. ing even with anisotropic materials.
Another important problem is handling and as-
Figure 3 sembling the large number planar panels. This prob-
Examples of lem is like putting pieces of a large puzzle together.
Quad-edge panels However, unlike puzzles we do not want construction
that can allow process to be challenging to the construction peo-
anisotropic ple. Instead, we want to simplify the construction
materials to be used process in such a way that the construction people
in construction. In fact, one of the biggest expenses for construction who may not have extensive experience can assem-
Square, rectangle, of large shapes comes from packing large number ble the components with a minimum instruction. The
rhombus, planar panels. If the panels have irregular shapes, it quad-edge data structure is particularly suitable for
parallelogram and will be harder to pack them efficiently. If expensive construction since it provides a well-defined assem-
trapezoids are anisotropic materials, which have a preferred bend- bly process. The construction workers can simply as-
shown here. ing direction, need to be used, packing become a semble quad-edge panels checking identical corners,
bigger issue since we cannot freely rotate shapes for which are given a pair of face and vertex identifica-
more efficient packing. This restriction can cause sig- tion numbers.
nificant amount waste material. We have developed a multi-panel unfolding soft-
Quad-edge panels provide efficient packing with ware (Akleman et al. 2016). In our framework, we
anisotropic materials since they only need to be start with a polygonal mesh approximation of the
bended along parallel fold lines -- called scores desired 3D shape. Then, we unfold it into multiple
in laser cutting terminology -- forming generalized planar panels, which are simply 2D polygons. We
cylinders in the construction of large shapes. This have used our unfolding software in one architec-
property of quad-edge panels makes them suitable tural studio at Texas A&M University [1]. Students (de-
to be printed using anisotropic materials, which al- signed and) physically constructed a wide variety of
ready have a preferred bending direction. The quad- shapes that closely resemble the shapes of their orig-
edge panels can be packed efficiently along their pre- inal designs by demonstrating the feasibility of this
ferred bending direction that is given by a set of par- approach (Figure 1& 4 & 5).
allel lines. Moreover, a single quad-edge panel can fit
inside of a single rectangle without significant waste

Figure 4
Top: Original mesh,
Folded quad-edge
panels and
Unfolded
quad-edge panels
and details. Bottom:
Physical structure
constructed by
quad-edge panels.
possible to form desired shapes by using quad-edge Figure 5

panels. The quad-edge panels are guaranteed to be Assembly process,
developable and have a preferred bending direction. Freshman design
Therefore, they can easily be unfolded and folded. studio at Texas A&M
Once folded, quad-edge panels become stronger in University.
the perpendicular direction.
REFERENCES
Akleman, E, Ke, S, Wu, Y, Kalantar, N and Borhani, A 2016,
'Construction of 2-Manifold Surfaces Using Physical
Version of Quad-Edge Data Data Structure', Shape
Modeling International, 58, p. 172–183
Eigensatz, M.V., Kilian, M, Schiftner, A, Mitra, N.J.,
Pottmann, H. and Pauly, M 2010 'Paneling architec-
CONCLUSION tural freeform surfaces', ACM Transactions on Graph-
With the design and construction of more and more ics (TOG), p. 45
unusually shaped buildings, the computer graphics Gandhi, M.V. and Thompson, B. 1992, Smart materials
community has started to explore new methods to and structures, Springer Science & Business Media,
Netherlands
reduce the cost of the physical construction for large
Otsuka, K and Wayman, C.M. 1999, Shape memory mate-
shapes. Most of the currently suggested methods ei- rials, Cambridge university press, United Kingdom
ther focus on reduction of the number of differently Saunders, R, Hartl, D, Malak, R and Lagoudas, D 2014
shaped components to reduce fabrication cost using 'Design and analysis of a self-folding sma-smp com-
mass production (Eigensatz et al. 2010) or identify posite laminate', ASME 2014 International Design En-
shapes that can be built using components that are gineering Technical Conferences and Computers and
Information in Engineering Conference, Buffalo, p. 8
already mass produced. In this workshop, by using
[1] www.youtube.com/watch?v=xpR_3_szq5g&
our custom software, we will demonstrate that it is nohtml5=False

Transformational Intelligent Systems
Parametric Simulation Workshop
Pablo Baquero1 , Nelson Montas2 , Effimia Giannopoulou3

1,3
Faberarium 2 Universitat International de Catalunya
1,3
www.faberarium.org
1,3
{paniba|efeminno}@faberarium.org 2 arq.montas@gmail.com
This workshop intends to propose a specific kinetic design application, to define

an architectural component and to simulate Shape Memory Alloy material
behavior. The objective of the workshop is to simulate the experiment before the
fabrication, itself aimed to model and anticipate the application's physical
behavior traits as a means to preview assembly.
Keywords: digital simulation, muscle wires, programmable matter, responsive

components, kinetic architectural systems
INTRODUCTION the gap between those two workflow processes (de-

Contemporary design methods and processes are sign and fabrication).
shaped by the tools and industry related topics "Kinetic systems with embedded intelligence
such as materiality, mechanics and organic assembly. will expose new programmes and forms as this
These factors influence, if not determine, not just the technology is incorporated into our everyday
way the final product is achieved, but also how we lives." (Fox, 2014)
think about our emotional attachment to products
themselves. METHOD STEP ORDER SWAPPING
In that framework, our proposal is that if we can In a conventional research method, the steps
develop integrated design, simulation as a mean to to achieve the design of self assembly and Pro-
preview fabrication, not only can we achieve a more grammable Matter, according to Raviv et al., are:
fluid road between ideas and component/products, Design, Fabrication and Simulation, because before
but also upgrade its potential influence in the pub- a certain material property are not mapped to be
lic through an adaptability that can be far more cus- computed in artificial simulation systems, it is im-
tomizable than anything we've done or even imag- possible to carry out the simulation cause of lack of
ined before. In this mindset, intelligent kinetic sys- data input. After the necessary data is derived from
tems are at the forefront regarding intelligence em- physical material testing, then it is possible to build
bedded, dynamic systems. a reliable simulation model. What were are propos-
Material behavior simulation (specifically Shape ing to achieve is, using laboratory tested and proved
Memory Alloys or SMA, in this case Ti-Ni) could shape data, to swap the last two stages in the decision-
the decision making processes in design and pose making process to make the laboratory testing step
specific kinetic challenges that arise from bridging

Figure 1
Proposed
simulation to
fabrication swap
method, taken from
(Montás, 2016)
an optional one, which is a more expensive one to conceptual frameworks used to address the design
accomplish. (see Figure 1) and decision-making processes towards the imple-
mentation and definition of kinetic, programmable
A FRINGE PARADIGM and shape shifting component applications. Which
As Skylar Tibbits et al. have stated, "if today we themselves will be drawn from a joint, intertwined
program computers and machines, tomorrow we design process regarding a Symbolic, top-down
will program matter itself" (Tibbits et al. 2014). logic, which is based on a combination of higher
This domain in material science has taken over re- level, ad hoc application and problem solving logics
search institutions over the last 5 years, with the and a Sub-symbolic, bottom-up logic which is de-
Self-Assembly Lab (SAL) at the Massachusetts Insti- rived from lower level traits and features that evolve,
tute of Technology (MIT) being one of its world leadin open ended processes, as these systems grow or
ers. Programmable Matter (PM), a term coined by adapt to different multi-level situations and contexts.
the staff at SAL and Thomas Campbell on a project This suggests that material functionality and aesthet-
commissioned by the United States Defense Ad- ics, which from now on will be conflated together,
vanced Research Projects Agency (DARPA-with none will be developed from a small (millimeters) to an ar-
of them taking direct credit for the term's coining) is chitectural construction scale (meters) to meet spe-
what is called a "fringe science", sitting at the cross- cific building and environmental criteria acting as
disciplinary convergence of, at a general level, three design guidelines to be followed.
main research and design fields: material science, This article's definition of Material Design (MD),
computational design and digital fabrication. while aware that there are a number of authors that
Material Design, Programmable Matter and have coined slightly different ones on the subject,
Computational Design will be the three convergent will be the one that this paper's authors have coined,

which itself is influenced by Liat Margolis's definition, thus influencing and, in a larger extent, potentially
in which "material" becomes a notion that is simulta- determining the overall building design. The pro-
neously matter (an alchemy of the periodic table) and posed Ni-Ti alloys in this paper have been selected
meaning (that which accrues behavioral descriptors from a commercially available in Robotshop.com and
according to their ever-changing empirical or theo- hold a transition temperature of 70C° to 90C° so, un-
retical contexts) ( Margolis et al. 2011), therefore: der normal thermal conditions, they have to be ac-
tuated by electricity (very low voltages ranging be-
Material Design (MD) tween 20 and 4000 Milliamps, for more information
MD as this research defines it is, within material sci- please refer to: [1] ).
ence, the process of configuration, reconfiguration or
Figure 2 modification, through programming or other meth-
Comparison of both ods like physical form finding, of any material to meet
(original and a certain set of criteria ranging from macroscopic
replicated) manifestation to behavior properties, thus enhanc-
simulations of the ing its capabilities to meet a specific set of design
deformation of a 2D needs, all together from the molecular scale, chem-
grid into a ically designing purpose oriented matter and thus
sinusoidal wave, developing unobtanium materials or multi-material
still-frames matrices and cast combinations that own up to such
sequence objectives (Montás, 2015).
(perspective view)
Nelson Montás Programmable Matter (PM)
(2015) based on PM is in science, engineering, and design of physi-
Raviv et al. (2014). cal matter that has the ability to change form and/or
Here the double function (shape, density, moduli, conductivity, color,
curvature etc.) in an intentional, programmable fashion[iv].
deformation Numerous authors, such as Michael Hensel, Achim
compared to that of Menges, Skylar Tibbits, Dan Raviv, Manuel Costoya,
the Raviv et al. Emilio Otero, Michael Fox et al., seem to agree on the
simulations and fact that this "procedural" way, as some have called It,
physical models is of designing architectural components straight from
an astonishingly the properties of materials and addressing design PHYSICS SIMULATIONS
close from their inner structures provides a new field of Kinematics is the part of mechanics that studies the
approximation to work whose unpredictable results mark a new stage movement of particles and systems without taking
the real models in of man's relationship with matter and nature itself in account the forces that produce it or, more simply
the original (Montás, 2015). put, the geometry of movement and which is used to
experiments.
relate displacement, velocity, acceleration and time,
Material Functionality and Design without making reference or the cause of the move-
Shape Memory Materials (SMM), in this case, Shape
ment. Then based on the notion that Kinetics is the
Memory Alloys (SMA), in the form of Nickel-Titanium
part of dynamics that studies the vector systems that
(Ni-Ti) alloys, pave the way for actuation driven, ar-
produce movement.
chitectural component systems that embed low level
The simulation will be done with Kangaroo, a real
intelligence (on-off, open-close logic) into their envi-
time, interactive physics engine that allows to ma-
ronmental mediation technology and programming,

nipulate time parameters and possible external fac- will search for application possibilities and formal lan-
tors such as thermal conduction which can be de- guage expression concerning the morphological re-
rived into structural load transmission. The forces sults.
derived from the alloys' programmed behavior are Digital materialization and the behavior of a ki-
used as data inputs to digitally model the transforma- netic system could give various values to design pro-
tions as springs with different stiffness and and rest cess and its final product. The simulation on a digital
length factors, according to the shape-shifting inher- platform helps designers to optimize geometric com-
ently designed for the kinetic material system, these positions according to a configuration of multiple pa-
spring particle systems are based in Hooke's elasticity rameters. Using SMA systems and parametrically op-
law (unary forces or springs). (see Figures 2 and 3) timizing the composition it is possible to reduce over-
charged mechanical stress and energy consumption
hence improving environmental performance. Hav- Figure 3
ing a digital simulation of a kinetic project can inform Comparison of both
future users about the behavior and can be used as (original and
visual demonstration and proof of concept. replicated)
simulations of the
deformation of a 2D
REFERENCES grid into a double
Baquero, Pablo and Giannopoulou, Effimia 2015 'Strate-
gies for Metallic Vault Structures', Proceedings of curvature surface
eCAADe 2015, Vienna (convex and
Fox, Michael A 2001 'Sustainable applications of intel- concave)
ligent kinetic systems.', Second International Confer- still-frames
ence on Transportable Environments. sequence (left side
Margolis, Liat 2011, 'Encoding Digital & Analogue Taxon- one-point
avigation', in Schröpfer, TS (eds) 2011, Material De-
SCOPE AND CONTRIBUTION sign Informing Architecture by Materiality, Birkhäuser perspective) Nelson
Our workshop exercise consists of building a spring GmbH, Basel, p. 148 Montás (2015)
mass, particle system approximation (Grasshopper + Montás, Nelson 2015, Performance Software Approaches based on Raviv et al.
kangaroo) simulation of an imploding structure that for Kinetic Architecture: Programmable Matter Based (2014). The graphic
will have the ability to shapeshift from a rigid poly- Simulations., Ph.D. Thesis, Universitat Internacional animations show a
de Catalunya strong resemblance
hedron shape (shape A) into "collapsed" like mor-
Tibbits, Skylar 2014, 'THE NEXT WAVE: 4D PRINTING PRO-
phology (shape B), derived from modeling the actu- GRAMMING THE MATERIAL WORLD', Atlantic Council of the replication
ation patterns comprised of flexible joints and rigid BRENT SCOWCROFT CENTER ON INTERNATIONAL SE- simulation with the
bars that react, as a whole, to external stimuli, in this CURITY, may 2014, p. 12 original one.
case heat. One of the objectives is to simulate the [1] http://www.robotshop.com/media/files/pdf/flexin
material's programmed behavior in order to properly ol-technical-data.pdf
predict critical function, actuation and physical prop-
erties. Grasshopper + Kangaroo bridge the design-
simulation tool's workflow in a single stream to opti-
mize and protocolize a smoother and fluent decision
making process.
These mechanisms will be studied from a macro-
scopic behavioral standpoint nonetheless being de-
rived from, lower level, molecular phenomena. Dur-
ing the workshop, together with the students, we

Multi User Online Design Environment for Mass Housing
A collaborative decision-support tool
Lo Tian Tian1 , Marc Aurel Schnabel2

1,2
Victoria University of Wellington
1
skyduo@gmail.com 2 marcaurel.schnabel@vuw.ac.nz
Participatory design is not new, various design projects have included public
opinions, ideas, and suggestions to help architects to produce designs that fit the
desires and needs of the users or community. In the context of mass housing,
participation is seldom applied as the primary focus is on efficiency. Designs are
generated mainly based on architects' experiences or through data from
developers or government sectors that claim to reflect the needs and desires of the
occupants. In this workshop a digital platform called 'ModRule' is introduced to
provide a means for a participatory process in the connection with a virtual
environment software, 'Fuzor'. ModRule plays the role of a collaborative design
platform while Fuzor provides real-time visualisation and building information.
The aim of this workshop is to explore how a user participatory design set-up
allows for an active participation of stakeholders in the initial design phase of
mass housing.
Keywords: participatory design, mass housing, digital system
INTRODUCTION MODRULE
This workshop aims to achieve design collaborations 'ModRule' (Figure 1) is a system [1] developed to fa-
between end users and architects to jointly generate cilitate the collaboration between the architects and
a mass housing design solution. Using a computa- the end users during the preliminary stage of mass
tional support system, the process can be facilitated housing design (Lo et al., 2015). This system takes ref-
and 'simplified' into design and interaction rules for erences from game design logic to develop a design
participants to engage with the process effortlessly. flow that allows architects to prepare a framework for
The participatory design experience of mass housing collaborative design with end users. The interface is
is enriched by an interactive gamified virtual environ- a step-by-step workflow that is very easy for partic-
ment that offers incentives for social design interac- ipants to get actively involved and be able to con-
tions. tribute to the process in an informed manner.

WORKSHOP SET UP Figure 1
In this workshop, participants shall experience a real- Top: Screenshot of
time online design collaboration. The workshop fo- ModRule
cuses on mass-housing designs. The aim is that preparation
every participant engages actively in the design of interface. Bottom:
their apartment within one mass-housing while at Screenshot of
the same time negotiating with their neighbours. ModRule
Obviously a higher number of participants will in- collaboration
crease the communication and negotiation possibili- interface.
ties testing both the capability of the system and the
design methodology. Every participant has the op-
portunity to encounter both perspectives, the one
of the architect and the one of the end user, allow-
ing participants to understand system setups and the
participatory processes.
Performing the role of the architects, the partici-
pants will have to set up their overall design and lay-
out within ModRule to establish the overall design
framework of the proposed mass housing building.
They have to decide how much design freedom they
are willing to offer to their end users. They define fix
points, such as vertical access, etc. and sub-divide
the plan in a meaningful grid for end users to build
upon. Since ModRule is developed based on game
FUZOR design logics, the architects may have the impres-
'Fuzor' (Figure 2) is a virtual environment BIM soft- sion that they are designing an architectural game for
ware [2] that allows users to simulate, evaluate and their end users. This process makes them aware of
collaborate design through real-time a cutting edge the various issues that are involved in participatory
visualisation and communication system. Fuzor has design set ups.
the capability of incorporating and linking data that Performing the role of the end-users, partici-
derive from a BIM software (e.g. Autodesk Revit) into pants explore the many folded possibilities the sys-
its system. tem offers to them reflecting a variety of users' needs.
Fuzor offers an open source API for other pro- Thus exploring the how mass housing allows for indi-
grams to connect like a plugin. ModRule take this vidual and collective design options. They can role-
opportunity making use of Fuzor's strong visualisa- play a certain family type or simply try to design an
tion and BIM capability to provide the users with a apartment to their own liking. In this process, the
more realistic virtual environment for their design en- participants will use both ModRule and Fuzor to en-
gagements. As ModRule is a prototype that uses we- gage with their design processes. They can experi-
bGL as its source code, the interface is limited to geo- ence a virtual simulation of their design within Fuzor
metric planes. Since the target audiences of users of and make their design decision based on these expe-
the above described systems are laypersons without riences.
much architecture knowledge, a better visualisation
environment is necessary for a more intuitive interac-
tion.

Figure 2
Top: Screenshot of
Fuzor with design
model and site
environment.
Bottom: Screenshot
of Fuzor with BIM
data and
component library
for user to choose
and design their
spaces.

WORKSHOP EXPLORATIONS REFERENCES
The workshop explores and discusses with the partic- Lo, TT, Schnabel, MA and Gao, Y 2015, 'ModRule, A User-
ipants the follow points: Centric Mass Housing Design Platform', in Celani, G,
Sperling, DM and Franco, JMS (eds) 2015, Computer-
Aided Architectural Design Futures. The Next City -
• Co-designing living space: How can layper- New Technologies and the Future of the Built Envi-
ronment, Springer Berlin Heidelberg, Sao Paolo, pp.
sons without any knowledge of how to design
236-254
architectural spaces play a significant role in [1] http://modrule.com/session/login
the design development and to what extent [2] https://www.kalloctech.com
can they insert their own visions into this pro-
cess?
• Presence of the architect: How can an effec-
tive communication between architects and
laypersons be achieved?
• Mass-housing: How can design of mass
housing be modified such that participants
can be included without much complica-
tions?
• Computational tools: What are the elements
that allow the design of mass-housing to pro-
vide high flexibility yet still within the con-
straints of an architect's design?
• Virtual environment: What are the level of
details needed to attain a comprehensive de-
sign collaboration process?
The participants are invited to provide constructive

feedback to the authors of how to simplify or en-
hance the participatory design process of mass hous-
ing.
ANTICIPATED OUTCOMES
At the end, participants are able to generate an de-
sign outcome of a mass housing building that caters
for the need for all involved stakeholders. Partici-
pants gain a deeper understanding about the prepa-
rations required for participatory designs in general
and how technology can assist with this process.
ACKNOWLEDGEMENTS
ModRule has been supported by Kalloc Tech provid-
ing technical knowledge to develop the interoper-
ability of ModRule and Fuzor.

FABRICATION | Design & Application
Adaptive Fabrication
Cellular Concrete Casting Using Digital Moulds
Zubin Khabazi1 , Michael Budig2

1
PhD Candidate at Singapore University of Technology and Design 2 Assistant Pro-
fessor, Singapore University of Technology and Design
1
morphogenesism.com
2
http://asd.sutd.edu.sg/faculty/michael-budig/
1
zubin@morphogenesism.com 2 michael_budig@sutd.edu.sg
Computational design and digital fabrication have expanded the use of digital
manufacturing machineries for the realization of architecture, yet they have their
own limitations of material use. These limitations caused some materials like
cement, plaster and clay become marginal in this new digital context, despite
their vast use in the building industry. In this context, this paper will present a
research, focusing on the use of concrete through the development of a
custom-designed device, which is an adjustable digital mould. This digital mould
has been designed specifically for a project called Procrystalline Wall and has
been 'adapted' to the conditions of its agenda in terms of size, shape, typology,
and even technical matters. However, this adaptability means that the device is
not aimed to work for any other project and remain exclusive to this particular
design only. This paper will further discuss the validity and obstacles of the
presented method in a more global context.
Keywords: Concrete Fabrication, Digital Casting, Digital Adjustable Mould,

Cellular Concrete Casting, Cellular Solid Morphologies
INTRODUCTION so-called 'paradigm shift' in its agendas, education,

Access to affordable hi-tech digital facilities and dom- practice and even theory (Schumacher 2014). There
inance of information technologies, have gradually are methods and tools for design and fabrication that
integrated into our daily lives, with the aim of im- have been established already, and have been used
proving our lifestyle, and fluency and easiness of for a while, yet new researches, especially for the
workflows in everyday life. Architecture, as one of realization of design products, would be needed to
the most physical phenomenon in human culture, widen the spectrum of applications and possibilities.
also utilizes digital media, to incorporate new de- Extensive use of computation in architecture,
sign tools and new fabrication technologies in its pro- created a global tendency towards complexity in de-
cesses. Such incorporation caused it to experience a sign, yet the question of materialization of such de-
FABRICATION | Design & Application - Volume 1 - eCAADe 34 | 83

sign products in the real world has engaged lots of Category 1: 'Concrete Casting' by Digitally
practitioners and researchers (Menges 2015). Search- Fabricated Moulds/Formworks
ing to answer this question resulted in the use of Three types of methods are studied in this category:
digital machineries (i.e. CNC routers, industrial laser - Mass-Customized Moulds/Formworks/Flexible
cutters, robotic arms, 3D printers), for the fabrica- Formworks/Lost Formworks. Mass-Customized
tion of complex products. But these machines have Moulds/Formworks are group of similar, but non-
brought their own limitations when used with certain identical moulds which are made for casting every
materials such as cement, plaster, clay, and caused single component of a project, with differentiated
them become marginal in this new digital context, parts. Here, instead of one mould for a singular
despite their vast use in building industry. Because module, being casted repeatedly for several times,
of the lack of proper machines to work with these there would be several customized moulds that each
substances (especially in the paste form), they have would be used once. The use of mass-customized,
not been used more than some academic researches digitally fabricated moulds/formworks is now easily
and experimentations. It seems that because of the achievable in terms of technicality. Digital design
integration of machine and material in this digital platforms are capable of providing facilities to model
realm, proper machines (and their respective meth- differentiated geometries and it is just couple of steps
ods) should be developed to facilitate the utilization further to convert them into moulds or formworks to
of such substances. be fabricated. The main concern here is not the tech-
In this context, this paper will present a research nique or process, but material use.
about using concrete through digital media of design The very first negative aspect of customized
and fabrication. In order to have a comprehensive technique is material waste. Moulds and Formworks
understating of the context, in the first part, research provided by these techniques are unique for any sin-
background is studied and common achievements gular object and they are not intended to be used for
and problems in the area are reviewed. This would any other piece/project. The waste that they provide,
help to establish the main research questions, which show more drastic negative impacts because of the
are addressed in part two. To study the methodol- materials these techniques usually use: PVC, foam,
ogy of the work in second part, experimentation on polyurethane, and other similar materials are 'smart'
a cellular concrete casting project with its relevant selections when considering the cost, but leaving
custom-designed mould is described. Final evalua- them as (mostly non-recyclable) waste in large quan-
tions, main obstacles in front of the system devel- tities, is noticeable. Another important issue that
opment, and conclusions are discussed in part three should also be noticed is the machining time needed
where it would also address future developments for this processes (especially for methods like engrav-
and agendas. ing) which is time/manpower expensive.
- Re-Useable Moulds. Implementation of re-useable
RESEARCH BACKGROUND moulds in concrete construction seems promising to
Research background would investigate the most es- solve the problem of material waste. The intention
tablished techniques of digital fabrication with con- is to use a flexible material for moulding, getting it
crete. Methods that are cited here are mostly from formed each time for a new piece, and re-form it for
academia and research projects to observe develop- another use after casting (Gramazio et al. 2014b).
ments, and to further address the achievements as Clay and Wax currently exist as tried-and-tested de-
well as shortcomings and technical difficulties they formable materials (McGee et al. 2014). In order to
face. This has been addressed in two categories: prepare the set up, clay/wax should be formed inside
a container to make the negative geometry of cast-
84 | eCAADe 34 - FABRICATION | Design & Application - Volume 1

ing piece. This could be done using a robotic arm or termines the maximum envelope of a structure that
a customized CNC machine. While this process shows could be printed. So far, large printers have been
smarter material use, there are certain shortcomings developed to print in the scale of houses, but any
as well. larger structure does not mean to enlarge the printer
Using flexibility as the initial property might (Skyscraper problem), and definitely other strategies
cause deformations during casting. This would cause should be developed (Tibbits 2015). There still exist
inaccuracies in the piece. It would be more drastic the problem of making inclined geometries, curved
when the size of the piece becomes bigger and hy- surfaces and so on, which without support for the al-
draulic forces of the fresh concrete become higher. ready printed concrete, is not easily possible. There is
These inaccuracies would affect the homogeneity of the question of consistency and cohesiveness of the
the product, and if complex assembly is also included concrete, suitable for printing purposes. The ques-
in the final process, would be a critical point. Mate- tion of reinforcement and how to implement it in the
rial specific problems exist as well. For example in the concrete is also valid.
choice of wax, the amount of energy needed to form - Concrete Deposition using Mobile Robotic de-
it, and then melt it back, should be calculated. The vices. The main idea behind mobile robots is to use
extra use of energy should not sacrifice the material small robotic fabricators, which can deposit concrete
re-use. while moving. They could move on top of the lay-
- Adjustable Moulds. Using a mechanically ad- ers that they have laid by themselves, so using the
justable mould, either being controlled by electrical already printed structure as the pathway (i.e. IaaC
motors or actuated with an external robotic arm (Gra- Minibuilder project) (Gramazio et al. 2014a) (Male-
mazio et al. 2014b), could potentially solve the prob- Alemany and Portell 2014). If working effectively in
lems of material and energy use. In this scenario, a groups through (mobile) communication, it is pos-
mould could get various forms and thus could pro- sible to use multiple robotic agents that are print-
vide various outputs. But it should be noted that ing a building in a collaborative manner (an image
these moulds are limited in terms of geometrical out- of agent-based fabrication). The idea of using small
put, mostly to one family of forms with capability of robots to do building jobs, now gets more attention
differentiation in some features like length, angle or in academic fields. Using multiple drones to assem-
curvature. This means that they are not generic soluble a building, follows the same logic. While most
tions and should be designed for any specific project of the problems from the previous section (concrete
separately. The cost of such operation, and time deposition) would remain the same, there are certain
needed to do so is considerable and sometimes a real technical issues that should be researched in the filed
obstruction. of robotic devices, which is mostly related to engi-
neering, computer science, robotics and other inter-
Category 2: 'Additive Manufacturing' of disciplinary fields.
Concrete Elements - Cement 3D Printing using Powder-Based Print-
Three main strategies are cited in this category: ing Technology. While the earlier experimentation
- Concrete 3D Printing (Deposition). The idea of 3D and mainstream of concrete 3D printing focused
Printing concrete like 'Contour Crafting' (Khoshnevis on the deposition of fresh concrete, there exist a
2004), or 'Deposition' technique, use this strategy to strategy of using cement in powder-based print-
deposit fresh concrete with a nozzle that is driven ers to print concrete. This idea has been tested in
by a computer-controlled machine. Inherent to the the academic fields, but not through printer-making
idea are certain questions that are still under devel- companies to see the actual working capacity of
opment. There is the problem of size and that dean industrial machine. The quality of the output

product, its density, strength, modulus of elasticity, would be shifted to design a digital mould. This
etc. needs lots of laboratory tests and proves to fabrication device could get the data immediately
be considered as standard concrete. But in general, from design and would facilitate casting a family of
this technique would be an interesting solution for forms by changing some geometrical features (im-
cellular/component-based 3D printing of differenti- plemented in the design of both mould and pieces),
ated building blocks. so the device as adapted to the typology of forms
which is generated in computational design pro-
METHODOLOGY AND IMPLEMENTATION cess. In order to test this new technique, a proto-
type project, the 2.2 meters high 'Procrystalline Wall',
Research Questions
with more than 400 differentiated concrete cells have
While investigating how digital media could form
been designed and fabricated.
concrete in two categories, it should be noticed that
the idea of 3D printing is out of the scope of this pa- Figure 1
per. On the other hand, casting techniques have their Digital Mould with
own benefits, but need to meet the requirements of adaptable
digital technologies and comply with its machines to geometry which
develop further. Casting concrete with digital me- can communicate
dia has different variables like mould/formwork/lost- with design
formwork approach, pre-fab or in-situ strategy, con- platform through
tinuous or discrete elements, solid or adjustable digital code.
mould, etc., where the focus of this work is on pre-
casting of discrete elements. This would be possible
through customized or adjustable/reusable moulds,
and as already discussed, those techniques are deal- Parallel Design Process
ing with issues like waste, inaccuracy, time/manpow- The design process includes designing the mould,
er/cost, and so on. The enquiry of this project is about and the wall, simultaneously, both affecting and up-
reducing the waste, time and cost of casting differen- dating each other. Challenge of the mould devel-
tiated pieces, and the focus is on the development of opment was more into solving technical issues (Elec-
a 'custom-designed moulding device' as an alterna- tronic, Mechanic, Programing), studying geometri-
tive to 'customized moulds'. The methodology and cal possibilities, improving performance (to be water-
procedure will be discussed in this part. tight, to maneuver easily, to be precise) and to reach
the point where data transfers seamlessly between
Adaptive vs. Customized Moulding: 'Pro- digital platform and machine. Challenge of the wall
crystalline Wall' Project design was to demonstrate differentiations, while
In response to the problems of customized mould meeting the limitations, needs and necessities of the
making, a new strategy has been developed to use simultaneously developing digital mould.
an adjustable mould, with digitally controllable ge- Cell geometry is the shared property between
ometry. In this scenario, digital mould could adapt the wall and the mould. While they should be ag-
its geometry to a digital object, through direct com- gregated and geometrically adapted to their specific
munication of data with computer (design platform). positions to build up the wall, each of them should
So instead of making one mould for each piece, there follow some guidelines to be able to cast by the vari-
is one digital mould, plus data for each piece, that ation of the same mould. Cells have been designed
adjusts the mould for casting (Figure 1). In this way, with a kite geometry in their plan, and extruded with
the demanding task of fabricating multiple moulds constant height. Among various geometrical param-

Figure 2 the kite wider or longer, yet the geometry would re-
Cell differentiation, main symmetric. As a result, the design of the wall
based on the initial was limited to some trace lines, upon which, cells
state in the middle, would be generated and arranged to fit automati-
Top: Angle change cally (Figure 3).
and Bottom:
Length change. Digital Adaptive Mould
Digital mould was developed as in integral part of
the design process. The aim was to keep mechani-
cal/electrical parts as simple as possible and the cost
of production as minimum as it could be. The oper-
ation of the device was observed through animated
model to make sure all pieces are working seamlessly
together. After modeling, all pieces have been de-
rived (Figure 4), either went through the fabrication
process to be cut, or ordered. Pieces include side
walls of the cells' negative geometry, gears, supports,
flaps, base and also electronics, which were a micro-
controller (an Arduino board), two Servo motors, Se-
rial communication cable and power supply. After as-
sembly, micro-controller was set in direct serial com-
munication with computer, so it could receive its data
from the model in order to run servomotors immedi-
ately. One servo used for the angle variation and an-
other one for the length. The steps of workflow (Fig-
Figure 3 ure 5) comprised of:
Final design of the
wall, where cells 1. Select a cell
being generated 2. Read its Angle/Length data
with differentiated 3. Remap data into the numeric range that is
angles and lengths suitable for servo motors
to fit into each 4. Send the data to the Arduino board to run
other. servo motors
5. Motors would adjust mould to adapt to the
geometry of the selected cell
eters to make variations, two of them have been cho- 6. Mould cavity should be lubricated for casting
sen, in order to make all cells' differentiations: one
was the 'angle' between the first two equal-length In order to work faster, based on the number of
sides, and the other one was the 'length' of the other pieces but limited with the budget, three separate
two equal-length sides (Figure 2). So both changes digital moulds have been made. They were slightly
would happen across the diagonal of symmetry of different from each other because each one of them
the kite geometry, either changing the angle at one was capable of maneuvering for a different range of
end or changing the length at the other end, making values (for angle and length). This has been decided
in order to make their movement more accurate by

Figure 4
Extruded diagram
of mould parts.
Figure 5
Data flow in the
process of
fabrication.

limiting the angular range that each servomotor op- mould faster. Since the method of using concrete
erates on. So all cells have been categorized based was moulding, especially with sharp edges and small
on their angle/length values (Figures 6 and 7), each cavities, plasticizers also added to simplify the move-
group associated with one of the three moulds (Fig- ment and forming of concrete. Finally, expansion
ure 8). The process has been carried on by labeling polymers have been added to the mixture to reduce
all the pieces, setting the moulds, and casting pieces the weight of the cells. All material strategies, per-
accordingly, letting them get their initial shape for centage of ingredients in the mixture, and quality of
about 7 minutes, taking them out for drying and cur- the output were consulted with experts in order to
ing process, cleaning the moulds and starting every- have the final product lightweight with sharp, clean
thing again. It took up to 25 minutes for each batch edges (Figure 9).
of three moulds/cells to be set and cast.
Figure 6 DISCUSSION
Cells with smaller Ideas of using adjustable moulds have been around
angle values but for some years, but they did not become widespread.
larger length (sharp There are various reasons like cost, feasibility, validity,
types). and technicality that might describe some obstacles.
This section will broaden the discussion on pros and
cons of the digital mould project for further evalua-
tions. The fact is that studies and observations show
three main concerns: project-device specificity, tech-
nical complexity and procedure difficulty, which are
discussed in this part.
Being 'specific' to one project only, devices that
are made for such non-universal solutions become
Figure 7 economically inefficient. The digital mould of this
Cells with larger project has been made by using simple electron-
angle values (wide ics and low quality parts and pieces to remain cost-
types) efficient. But it showed the problem of accuracy and
vulnerability in high-demand jobs, because the qual-
ity of its parts is not assured in this instance. It has
been proved that this strategy is affordable while us-
ing non-expensive materials/electronics, but at the
same time it should be considered that such devices
are not industrial, heavy-duty machines, suitable for
high-demand works. They will lose their consistency
after a while, especially when working with a high vis-
cosity material like concrete. However upon conver-
The concrete mixture that has been used in this
sion into a more industrial-type device, there would
process was reinforced with fibers for extra consis-
be a consequential increase in cost. Here the evo-
tency. Since cells should have remained clean and
lution from maker-quality device, to an industrial-
sharp at their edges, this fiber reinforcement was
grade machine would flaw the current discussion, as
necessary for extra cohesiveness. Chemical admix-
the assumptions are fundamentally different. The
tures have been added to the concrete to reduce the
idea of maker-quality moulds might be applicable
time of drying, in order to take the pieces out of the

Figure 8
Cell distribution
between three
different moulds
based on their
angle / length
range of values.
for a small-scale project, yet maintenance would be helped to simplify it for easier implementation. This
needed during the work and some room should be means that upon certain research and development
considered for inaccuracies and error proofing. of systems, it could be strategic to implement such
Issues related to technical complexity are less methods in the design schemes of specific goals.
problematic today, because of the availability of While elaborating on techniques of digital fabri-
open-source knowledge and instructions for similar cation with concrete, especially the case of adaptive
works. It is easy to have access to ingredients like mould, one should know that choosing any method
electronic and mechanical parts and also fab-labs to of fabrication is a multi-criteria task that needs differ-
fabricate pieces, and it is as easy to find instructions ent information to be evaluated in order to come to
and tutorials to work with, program, assemble and a decision. This information are ranging from man-
run them. Although the challenge of expensive fa- agerial and authoritative aspects, up to technological
cility is still valid, it should be noticed that affordable and detailing. All such parameters have certain im-
components and machines become more and more pacts and could affect the methods of construction.
available. It is the design-research team's responsibility to con-
Graphical interfaces and node-based- sider all possible approaches and select one that best
programing languages are among the computa- suits the project's situation.
tional improvements that help to reduce the prob-
lems of procedure difficulty. These tools/services OUTLOOK: COMPLEXITY IN DESIGN, SIM-
are helping to facilitate the use of computers, pro- PLICITY IN CONSTRUCTION
graming languages, electronics, etc. for non-expert The objective of this paper was to use all current pos-
users (like architects), and for interdisciplinary ap- sibilities and technologies available in DIY/Maker do-
plications. This type of work would be really com- mains to see how they could help to develop sim-
plicated with manual procedures but digitalization ple techniques for fabrication. The budget of each

of the moulds in this project was less than $80, yet There are two main lines of research to further
each of them managed to get more than 130 differ- progress this project. The first one would focus on
ent positions and provided quite accurate results as a the geometry of the mould/cell: while keeping it sim-
digital fabrication device. While the complex part of ple, what types of complexities could be implanted in
the process has been done in the studio, it was aimed the design phase, to get more diverse outputs. This
to make the production simple, clarify and minimize should be also accompanied by more clarification on
the interface, and complete the task without experts the platform and simplification of the interface, us-
during the casting and assembly. The idea is to work ing data visualization, visual geometry analysis and
with communities of 'non-specialized people' who so on. The other line of the research would focus
can perform such tasks, in order to build 'specialized more on the qualities of the material systems (con-
design' products, yet with simple workflows. crete/cement, plaster, clay), which were examined in
Figure 9
Final assembled
'Procrystalline Wall'
project.

the scope of this project but has not been addressed Khoshnevis, B. 2004, 'Automated construction by con-
in this paper. tour crafting-related robotics and information tech-
Although the usage of complex robotic arms, nologies', Automation in Construction, 13, pp. 5-19
Kuznetsov, S. and PAULOS, E. 2010 'Rise of the expert am-
or high-end additive manufacturing machines, indi-
ateur: DIY projects, communities, and cultures', 6th
cates one end of the digital fabrication spectrum, Nordic Conference on Human-Computer Interaction:
there would be interesting outputs via DIY fabrica- Extending Boundaries, New York, US
tion robots and custom devices at the other end. Male-Alemany, M. and PORTELL, J. 2014, 'FABbots: Re-
There are examples of the methodology of the work, search in Additive Manufacturing for architecture',
namely in the ideas of Neil Gershenfeld at MIT's in Gramazio, F., Kohler, M. and Langenberg, S. (eds)
2014, Fabricate 2014, gta verlag, Zurich
Media Lab (Gershenfeld 2011), or similar design
McGee, W., De Leon, M. P., Willette, A., Brell-Cokcan, S.
approaches by Marta Male Alemany at Fab-Robot and Braumann, J. (eds) 2014, Robotic Fabrication in
project in IaaC, Barcelona (Male-Alemany and Portell Architecture, Art and Design 2014, Springer Interna-
2014), and many more. These schemes are aiming to tional Publishing
be easily accessible, open-source, relying on commu- Menges, A. 2015 'Material Culture, Rethinking Material-
nity work, and targeting public domains (Kuznetsov ization through Computation', Taubman College of
Architecture lecture series, Taubman College of Archi-
and Paulos 2010), so not only showcasing a different
tecture, U. O. M.
type of operation, but different type of human inter- Tenorio, A. 2014, 'The Impact of Parametricism on Ar-
action with technology as well. This mixture of dig- chitecture and Society, interview with Patrik Schu-
ital with manual would ease the access to complex macher', on-line: http://www.patrikschumacher.com/
design products through simple procedures and de- Tibbits, S 2015, 'Challenges and Opportunities', 3D Print-
vices. ing and Additive Manufacturing, 2(4), p. 151
[1] http://morphogenesism.com/project/p1/p1.html
Acknowledgements
The experimental part of this project has been
done with the collaboration of A.Akbari, A.Eslami,
M.Tavafzadeh, H.Panjehpour, K.Kazemian, K.Azari,
A.Zarif, A.Mokaram, P.Heydarian, S.Vaseghi and
A.Iravani at morphogenesism, laboratory of algo-
rithmic architecture. [1] Further continuation of the
studies, review of the related works and expansion of
the discussion are part of the more broader, on-going
PhD research at Singapore University of Technology
and Design.
REFERENCES
Gershenfeld, N. 2011, Fab: The Coming Revolution on Your
Desktop-From Personal Computers to Personal Fabri-
cation, Basic Books
Gramazio, F., Kohler, M. and Langenberg, S. (eds) 2014,
Fabricate: Negotiating Design and Making, gta ver-
lag, Zurich
Gramazio, F., Kohler, M. and Willmann, J. 2014, The
robotic touch : how robots change architecture, Park
Books, Zurich

Negotiate My Force Flow
Designing With Dynamic Concrete Formwork
Oliver Tessmann1 , Moritz Rumpf2 , Philipp Eisenbach3 ,

Manfred Grohmann4 , Topi Äikäs5
1
KTH Stockholm School of Architecture 2,3,4 University of Kassel School of Archi-
tecture 5 Betoniviidakko Oy
1
oliver.tessmann@arch.kth.se
2,3,4
{rumpf|eisenbach|grohmann}@asl.uni-kassel.de
5
topi.aikas@betoniviidakko.fi
A composite system, made from layers of perforated plywood and latex film,
together with ropes and rubber seals, performs as a flexible concrete formwork.
In our research we investigated whether such a system could yield more than only
one single repetitive concrete form. We sought to reduce the amount of material
consumed by conventional formwork and we conceived formwork as part of the
design process rather than a technical means during construction. We worked as
a team of architects, structural engineers and artists and through the confluence
of computational design, digital fabrication, material simulation and prototyping.
Keywords: Dynamic formwork, concrete, flexible mould
INTRODUCTION tailored textile under tension and its boundary con-

The way we design and build with concrete lacks in- ditions (Manelius 2012, 30).
novation. Resources, energy and architectural poten-
tials are wasted because concrete processing is stuck METHODOLOGY
in the industrial logic of endless repetition and serial Our investigation on novel concrete formwork is
production. Concrete is treated as a passive material based on three major methods:
poured into rigid and massive containers that act as
formwork (Forty 2012, 170). • Decomposing and carefully exploring the dif-
In our research we challenge the notion of con- ferent functions of a formwork in order to
ventional concrete formwork as a massive and rigid find novel materials or material compositions
systems that is only capable to producing repetitive for the various functions. (Veenendaal et al.
elements. We seek for a formwork system that is not 2014)
only flexible but rather dynamic, i.e. it enables cast- • Simulating and physically exploring the forces
ing of concrete objects with various shapes using one at work during the process of becoming.
formwork. We want to overcome the formal limita- • Linking designing and making through com-
tions of flexible fabric formwork predefined by the putation and digital fabrication, with the aim

Figure 1
Concrete prototype
made with dynamic
formwork.
to conceive both formwork and its outcome plored and exceeded the limits of plywood thickness
as a design challenge. and pre-stressed the material with ropes to achieve
strength through curvature. We furthermore sim-
ulated the process of bending with particle-spring
models to quickly test the range of formal opportuni-
Decomposing Formwork
ties provided by the material system without actually
Instead of focusing on a particular form that we want
entering the cumbersome process of physical proto-
to cast, we start with a functional analysis of form-
typing.
work: What are the constituting elements and their
functional dependencies? Which functional require-
Designing with and through formwork
ments - such as withstand the hydrostatic pressure of
We reversed the conventional design process when
liquid concrete or the requirement of the formwork
working with concrete: instead of designing a form
to be watertight - can we extract from this analysis?
and subsequently transcend its ideal geometry into
Can we find new materials for certain elements? Can
concrete through whatever means, we reverse the
we re-program the dependencies?
process for the sake of research in formwork systems
focusing on the question: What are the forms we
Simulating and physically exploring form-
can achieve through the use of our formwork proto-
work
type? What forms could be found when the concrete
Our formwork is supposed to consume less mate-
in its liquid state impacts its forces on the fragile form-
rial. Instead of being massive it should rather balance
work?
the forces at work during casting and curing. We ex-

Figure 2
Formwork
components.
DYNAMIC FORMWORK stressing the strips. Consequently the concrete is

The dynamic formwork system is composed of two formed into objects with multiple structural capaci-
layers of perforated plywood, latex film, rope, and ties.
rubber seals as outlined above (Figure 2). Each of Like folded paper their in-plane tensile loads are
these elements deals with one particular function of activated and create a high stiffness. The bending
a concrete formwork. increases the effective depth of both the plywood
formwork and the cast concrete object. When grab-
Plywood bing a piece of paper we intuitively twist and bend
Planar plywood has a low stiffness while being ca- it and thereby exponentially increase the global mo-
pable to bear high loads. Through carefully placed ment of inertia, transforming a weak bending system
incisions the material gains even more elasticity so into one that carries its loads as a membrane system
that bending and twisting can be combined (Fig- (Eisenbach et al. 2013, 205).
ure 3). Slotted plywood strips gain almost textile- Prestressing the plywood strips with ropes sub-
like properties but perform structurally not only in stitutes our hands and allows for differently shaped
tension but also in bending. A wide range of forms configurations of the formwork. Through varying the
can be achieved through bending, twisting and pre- location and the tension strength of the ropes the re-

Figure 3
Formwork
assembled and
pre-stressed.
sulting forms change (Figure. 3, 7 showing different can be re-configured in different shapes. The upper
configurations), the formwork becomes dynamic. and the lower layer are either clamped or bolted to-
gether.
Latex film
The porous nature of slotted plywood requires a sec- DESIGNING FORMWORK
ond material layer to tighten the formwork. Latex Assembled from the above-described components
film proved most capable to follow the changing we explored the formal capacities of the formwork.
shapes of the plywood strips without creating wrin- The cast objects are rather results of this exploration
kles in the cast object. than a priori designed objects.
Under the hydrostatic pressure of the liquid con- Perforated plywood with its almost textile-like
crete the latex bulges into the plywood slots but ef- behaviour allows for large rotational movement
fortlessly withstands the forces. The material can be along the longitudinal axis of the strip. We tested
reused for multiple casts, however, silicon-coating the twisting shape with different rotational angles. A
wears off quickly through chemical reactions with wooden frame locks the twisted strips in precise po-
the concrete. While the first concrete objects have a sition.
glossy surface later casts are rather matte. The seal- Bolts between the two plywood layers clamp the
ing qualities of the latex are not affected. rubber seal, the latex film and the wood together and
form a precisely shaped edge of the concrete object.
Rubber seal However, the freshly cast concrete induces horizon-
Rubber profiles with rectangular cross-section seal tal bulging and reveals imbalance between the forces
off the perimeter of the formwork. The latex film is ei- at work during the process of becoming. Thus we
ther wrapped around the rubber or the different ma- dismissed the wooden frame that reduces the formal
terials are clamped together. The rubber is able to flexibility of the plywood to twisting. In the next iter-
follow the bent and twisted forms of the plywood. It ation pre-stressed ropes twist and bend the plywood.
furthermore defines the edge thickness of the con- Since ropes can be used to pull their fixing-points to-
crete object. gether but not to push them apart they have to be
All elements of the formwork are re-useable and placed on both sides of the formwork to create alter-

Figure 4
Latex, concrete,
plywood. The
incision pattern is
clearly visible in the
cast object.
Figure 5
Rubber profiles as
flexible boundary
seal.
nating curvature. To bend a strip of plywood along could be examined. A crucial aspect is the calibra-
the short axis, the rope has to be mounted parallel tion of the digital simulation with the physical pro-
to the long side and to bend the plywood along the totype. Different thicknesses of the plywood as well
long axis, parallel to the short side respectively. To as varying incision patterns produce various bend-
introduce twisting through ropes is a rather challenging behaviours that need to be analysed and the sim-
ing task but can be achieved through diagonal ropes ulation setup has to be adjusted accordingly. The
on both sides. impact of the hydrostatic pressure of liquid concrete
A computational particle-spring model sup- onto the formwork was neglected in this stage of for-
ported the investigation of the formal possibilities mal exploration.
through the simulation of the forces acting on the Curvature analyses of the digitally generated
material system and complemented the physical shapes are a first indication of their structural capaci-
testing. Proportions of the digital plywood sheets ties. The physical limitations of the actual perforated
could be adjusted and 'digital ropes' were quickly re- plywood sheets in regards to bending and twisting
configured. Thus a wide solution space of possible radii have to be kept in mind, since the 'digital ply-
formwork formats and bent and twisted geometries wood' does not break. Incorporating the sequence of

Figure 6
Formwork system
with twistable
plywood lamellas
within a wooden
frame. The
formwork was used
to produce
differently twisted
concrete lamellas.
tensioning of multiple ropes increases the complex- ously congruent drill holes do not match anymore. A
ity exponentially. The order in which two or more parametric model would obviously allow for fast re-
ropes are pulled is however relevant for the resulting verse engineering of this geometrical problem and
geometry. The effect the use of our hands - even un- provide a precise alignment of holes. However, here
consciously - has in the physical prototypes cannot we decided to have slotted holes to maintain the flex-
be implement into a digital simulation. ibility of the formwork system since every change
The incision pattern changes the geometrical of the bending radius required a new alignment of
properties of the plywood strip: The elements can be holes (Figure 9). In one formwork prototype clamps
bent into forms impossible to achieve with continu- replace the bolts in a hanging formwork and sim-
ous material. Since the material properties are un- plified its reconfiguration. Moreover the clamps fa-
changed the novel forms cannot be called double- cilitate further minor shifting while the concrete is
curved. The slots rather allow for various regions of poured into the formwork, which was meant to de-
different curvature and with it for a global form un- form during casting in this case.
achievable with non-perforated sheet material. The
curvature changes are negotiated by the small-scale CASTING
saddles (Figure 7) between the slots that twist under While casting, the uncured concrete acts upon the
the external forces induced by the ropes. Bending led formwork system through hydrostatic pressure and
to additional curvature, which - together with a re- quickly reveals every imbalance between the forces
orientation of the formwork during casting - reduced involved. Calibrating the pre-stressing forces to with-
the bulging significantly. stand the pressure proved to be a challenging task
Bending the plywood requires a revision of the that could only be tested through prototyping. We
edge fixing detail. The strips are congruent as pla- used self-levelling screed to mimic free-flowing and
nar or twisted surfaces and allow for bolting through self-compacting concrete. We needed a material that
congruent holes. When bending, the offset between flows smoothly into narrow formwork and induces
the wooden strips that forms the cavity for the con- uniform dissipation of pressure.
crete, leads to two different bending radii and previ-

Figure 7
Digital simulation
as particle spring
model and physical
prototyping of the
plywood strips
under tension. Both
approaches were
used in parallel to
inform each other.
Figure 8
Strength through
curvature.
Exploration of the
formal potential of
the formwork
systems.
HYDROSTATIC PRESSURE AS DESIGN CRI- dependent on the height of the column of liquid
TERIA above. The magnitude of that pressure is linearly de-
A major criterion in designing any kind of formwork pendant from the self-weight of the poured liquid;
for concrete elements is the effect of the hydrostatic thus the magnitude of the pressure of concrete is 2,5
pressure. As described within the term, hydrostatic times bigger than water pressure.
pressure describes a static load derived from liquids Complementary to the hydrostatic dead loads
exerting pressure normal to its contacting surfaces dynamic loads may occur from the casting process,
depending on the pouring height. Since the concrete

Figure 9
Clamping details:
Slotted holes allow
for flexible
adjustment of both
formwork layers
and the edge seal.
Figure 10
Hanging formwork
sagging under the
load of liquid
concrete.
is casted from the top of any formwork the concrete sion pattern reappears as a relief in the concrete. De-
drops down a certain amount in a way that move- pending on the orientation of the formwork during
ments are happening from the clashing. casting the hydrostatic pressure reinforces or coun-
A second component of dynamic input may oc- teracts the curvature previously induces by tension
cur by compacting the concrete, which is dependant ropes. In the hanging formwork (Figure 10) the con-
from the choice of concrete mixture and composi- crete weight leads to increasing curvature, thus the
tion. initial form is exaggerated. The twisted concrete
Obviously there is a natural contradiction when lamella in contrast suffered from horizontal bulging
developing a flexible formwork: On the one hand a of the plywood (Figure 11).
formwork surface is wanted to be as flexible as possi-
ble to be able to deform its shape; on the other hand a COUNTER PRESSURE
rigid surface is required to resist the forces originated One of the main challenges in the present research
from the casting process. project is the handling of the hydrostatic pressure of
In our research approach we tried to take the ad- the liquid concrete. It is a distinct effect that the re-
vantage of the concrete pressure as a design criteria sults of the achieved form completely change when
and to take it into account in the form finding process - for instance - changing the casting direction from
rather than to spend the effort in stringently working standing to laying.
against the pressure (Figure 10). A promising approach is to work with counter
pressure applied via a non-curing material surround-
CONCRETE OBJECTS ing the formwork. A simple way to resist the static
The cast prototypes are invaluable means to identify load from the concrete in its liquid state is the build-
the potentials and the shortcomings of the formwork up of a bed of sand below a laying formwork. In
system. Their surface and overall form clearly incor- this manner a counter pressure can be achieved from
porate and reveal its process of becoming. The inci- both sides by placing the formwork in a box and, dur-

Figure 11
Concrete
prototypes derived
from dynamic
formwork systems.
ing the casting process of the concrete element, the ity (Figure 12).
box outside of the formwork is filled up as well.
Such an approach was tested with a formwork CONCLUSION AND OUTLOOK
for double curved surfaces, which has a setup simi- The work presented in the paper is an on-going re-
lar to that of the dynamic formwork depicted above. search project. The preliminary results are rather
Only the perforated plywood was substituted with a seen as the starting point for the subsequent re-
woven formwork from strips of plastic sheet material search.
(PVC-foil with a material thickness of 0.5 mm). Lay- Further materials that act as counter pressure
ers of latex film and rubber seals were applied just as to the concrete in its liquid state will be tested.
before. Before casting the formwork was placed in a Reversible materials such as wax, soluble non-
box whereupon the gap between the formwork and hardening mortar or even water are to be considered.
the box was filled with sand. The woven formwork For weaker formworks it will be necessary to study
proved to be sound enough to withstand the pres- processes of casting concrete and fill-up material to
sure of the sand. The resulting cast - made from gyp- counter its pressure simultaneously.
sum in this case - validated the conceived procedure. Furthermore the relation between the incision
Other than the perforated plywood, the woven pattern and the deformation under pressure will be
formwork has only tiny openings between the indi- investigated. First tests have shown that a detailed
vidual strips. Therefore a further test was conducted understanding of this relationship allows exploiting
using a formwork with no latex film. Here the sand the hydrostatic pressure as a design driver in the
not only provided the counter pressure necessary, pattern generation. Such an approach will benefit
but it absorbed the minor leakages of liquid gypsum from the integration of hydrostatic pressure as a force
as well, resulting in a perfect cast without deforma- in our computational particle-spring model; another
tion of the overall geometry and a high surface qual- aspect that we will investigate further.

Figure 12
Using sand to
counter the
pressure of the
liquid concrete /
gypsum. The
genesis of casts on
the right shows
three differently
deformed results in
the top row and
two non-deformed
casts – with and
without latex film –
Significant enlargement of the geometrical spec- REFERENCES in the bottom row.
trum, and thus a further research prospect, is the de- Eisenbach, P., Vasudevan, R., Grohmann, M., Bollinger,
formation ability of formwork faces not only in bend- K. and Hauser, S. 2013 'Parapluie - Ultra Thin Con-
ing direction but also the extension and contraction crete Shell Made of UHPC by Activating Membrane
Effects', Proceedings of the International Association
in plane direction to allow fully three dimensional
for Shell and Spatial Structures (IASS) Symposium 2013
manipulations. „BEYOND THE LIMITS OF MAN”
The experiments encourage the team to fur- Forty, A. 2012, Concrete and Culture: A Material History,
ther investigate systems that overcome the notion of Reaktion Books, London
formwork as a massive and heavy means of construc- Manelius, A.-M. 2012, FABRIC FORMWORK Investigations
tion and rather make it a part of the design process. into Formwork Tectonics and Stereogeneity in Archi-
tectural Constructions, Ph.D. Thesis, Royal Academy
of Fine Arts, Copenhagen
ACKNOWLEDGEMENT Veenendaal, D., Bezbradica, M., Novák, K. and Block, P.
Supervisors: 2014 'Controlling the geometry and forces of a hy-
brid cable-net and fabric formwork for thin con-
KTH Stockholm: Prof. Oliver Tessmann, Prof. Jonas
crete shells', Proceedings of the IASS-SLTE 2014 Sym-
Runberger. posium "Shells, Membranes and Spatial Structures:
University of Kassel: Prof. Manfred Grohmann, Footprints", Brasilia
Philipp Eisenbach, Moritz Rumpf.
CONSOLIS Finland: Topi Äikäs
Participating Students:
KTH Stockholm: Lison Bader, Natalia Daukszewicz,
Uuf Edgren, Felipe Saul Franco Franco, Mimmi
Gustafsson, Benjamin Onoszko, Francesca Pernigotti
and Michelle Vallomy.
University of Kassel: Adrian Golab, Zijian Han,
Denitsa Koleva, Jannik Kratzenberg, Svenja Krieg, Flo-
rian Reichmann, Witali Suchan and Cynthia Ward.

Additive Manufacturing and Value Creation
in Architectural Design, Design Process and End-products
Heidi Turunen1
1
Aalto University
1
http://www.aalto.fi/fi/
1
heidi.turunen@aalto.fi
The objective of this paper is to clarify how value creation can be a part of
architectural design and end-products when using the new emerging technology
of additive manufacturing. Different kinds of values that have emerged from the
research material have been analysed and summarised using selected case
studies of recent building-scale projects. In applying this technique to
architecture, the result can be visually and functionally novel, smarter and more
sustainable buildings or products. A new individually manufactured or
customised architecture can be created to serve different cultural and well-being
needs cost effectively and without any waste. This new production method can
lead to unique joint structures with the use of traditionally produced new or old
building parts to enhance architecture, prevent climate change or aid
environmental issues. However, most research projects and applications done by
commercial companies are at the early stages.
Keywords: Large-scale additive manufacturing, 3D printed architecture, Digital

design, New materials, New production methods
INTRODUCTION advent of this change, many projects sought to prove

In architecture, climate change, environmental is- that this new way of constructing products can cre-
sues, digitalization and resource conciseness can be ate added value to the manufacturing process and
seen as driving forces for compressing production production. The benefits of the method have been
and simplifying structures and materials. Utilization predicted to be significant. In contrast, value cre-
of digital design and manufacturing can be seen as ation in architecture and its end-products have not
one realisation of that. One prevailing variation, CNC- been researched much. In this paper the term value
manufacturing, a method that removes material, has is described to mean valuation or usefulness through
revealed several new possibilities for architects and an evaluation. The evaluation is always subjected to
designers. The opposite method, which does not time, place and the person who acts as an evaluator.
waste material, is called additive manufacturing, and Value creation can be based on social or economic as-
this has started to spread only in recent years. At the pects, but in architectural design, values can be seen

also as a guarantee of quality. In architecture, value out examining thoroughly the possibilities of archi-
creation can be seen as a result of a series of selec- tectural design and favourable product applications
tions. The selection making can be based on valuing, or purpose-built concepts are huge. The hype is not
such as favouring a high-quality material or empha- enough to maintain the interest of customers. There
sizing environmental aspects during the design pro- have to be other methods, such as value-based re-
cess. Value creation, conscious or unconscious, can search and business that create a long-lasting evolu-
be done also through evaluations of appropriate mation of large-scale additive manufacturing.
terial features, such as surface structure, colour, form
or composition. The architect or designer can also in- Objectives, Methods, Data and Theoretical
clude intangible value-laden properties, such as feel- Framework
ings and individuality (Ashby 2014, 196). With these The objective of this paper was to find wider
choices, an architect or designer can guide the de- value-laden themes that can be found currently
sign process in a certain direction. When the com- in architectural-related additive manufacturing
bining of a material and a novel product method is projects. The selection process for the study cases
not yet established, different kinds of value-based as- was centred on the question: how additive manu-
pects can be seen as a guide to creating meaningful facturing can create added value for architectural
and appropriate solutions for the future. In addition, design and end-use applications. The framework for
when presenting the main value aspects clearly, as a this research was defined under themes where ar-
kind of guarantee, the novel practice can be given a chitects can use selection or decision-making as a
rationale. Quality or innovative thinking have been tool or method to creating values. Understanding
found in many current case studies and prototypes. has been based on comparing the data of projects
If additive manufacturing is to become more prof- which have used additive manufacturing to the cur-
itable, it can lead to more innovative visual design rent situation with more traditional manufacturing
and end-products. Although, additive manufactur- methods. The theoretical framework consists of a
ing is a very novel method in large scale architec- whole chain of design, from the basic elements of
tural structures, it might be a realistic way to create the design process to the finished outcome. Dif-
buildings or architectural-related products in the fu- ferent kinds of printing method or material did not
ture. However, without demand, a conceptual de- refine the sampling. The data have been analysed
sign and adequate applications, the new manufac- and structured using a qualitative method, more
turing method will not be successful. The focus of specifically, coding the data thematically using in-
this paper is on presenting different kinds of values terpretation and classifications that generalise the
that have arisen from the research material, to cre- phenomena under prominent themes. Knowledge
ate understanding of how the production method of different kinds of projects and cases constituted
can contribute to architecture in future. It is very the research material and it was found using the news
clear that the benefits of the architectural possibili- groups of additive manufacturing. Publications such
ties are not yet adequately understood or researched. as books on additive manufacturing did not have
An emerging method can be seen as a risk when enough information, because the subject has only
thinking of how to adapt the new method to the de- recently emerged. Non-scientific news groups were
sign process, or whether the new method will benefit harnessed to find up-to-date material and relevant
the construction process enough compared to tradi- research or the design projects of universities and
tional methods, or even how additive manufacturing commercial companies.
will change over the years. The investments costs of
starting a new additive manufacturing business with-

BACKGROUND Large Printers
Additive manufacturing or 3D printing as it is also The size of a printer determines the size of a printed
called, has been used commercially on a small scale object. In addition, printers can produce small mod-
since the 1980s, when the first commercial 3D printer ular elements that can then be used to create larger
was patented by Charles Hull for 3D Systems (Hoskins structures. If large and continuously printed objects
2013, 30). Since then the technique was used, for ex- are desired, gantry based or movable printers can be
ample, to speed up product development processes used, which can continuously manufacture walls or
by providing design teams with prototypes quickly other large structural elements.
for evaluation and further development. Today the
technique is used in a great variety of industrial fields VALUES AND APPLICABILITY IN ARCHI-
such as medical, automotive and aviation industries
TECTURE
and the scale of the printed object can be from the
In this paper, the question of how additive manu-
nanoscale to quite large structures. Several types
facturing can create added value in architecture has
of additive manufacturing techniques have been in-
been considered through two different themes. The
vented to serve different needs, industrial fields, and
first theme describes how added value can be cre-
printing materials. The technique is based on three-
ated by the selection process for using the basic el-
dimensional computer data that is then transformed
ements of architecture. The three sub-themes are
into a specific format for the printer and objects will
aesthetic aspects, the possibilities of the materials
usually be created using computer controlled noz-
and the design of structures. The second theme is
zles, layer by layer. Lately the printing technique has
value creation through end use applications. Value
been applied also to a wide variety of projects in the
creation is researched through themes such as new
construction industry. However currently there are
product properties, novel frameworks of design, new
not many building-scale 3D printed products on the
product fields and applications themes. Each sub-
market, just some prototypes and on-going research
theme includes case examples that clarify the appli-
projects. Projects differ from each other by material,
cability. The analysis of values considers how the se-
product type or printing technique.
lected case studies contribute to the new production
Techniques related to construction industry method. The same cases could be used as examples
Fused Filament Fabrication (FFF), stereolitography also in other themes or sub-themes. The themes also
and robotic arms are the most common addi- entail connections, such that value can be created by
tive manufacturing techniques when large scale combining structures and materials, when designing
architecture-related objects are printed. Printers are functional and novel end-use applications.
often custom made for the specific materials.
VALUE CREATION FOR ARCHITECTURAL
Materials DESIGN
Almost all materials, such as concrete, plastic, steel This section presents how the basic elements of de-
and wood that have been traditionally used in archi- sign, in addition to tools used in architectural design
tecture are used in large-scale construction industry- projects, can create added value when additive man-
related printing projects. Most of the large-scale ufacturing is chosen as a method (also refer to Figure
printing projects do not focus on developing envi- 1).
ronmental friendly printing material, though recy-
cled waste material has been used at least in some Aesthetic
cases as a concrete-based printing paste. In additive manufacturing, it is possible to enlarge
the aesthetic scope. The digital design and manu-

Figure 1
Two main themes
with the three
sub-themes and the
connections
between them. The
themes emerged
during the research
process.
facturing method is flexible when creating complex The case of algorithm-based design. The Digi-
forms and variations in the visual elements. Surface tal Grotesque project (Figure 2) by two architects,
structures, colours, the haptic properties or material Michael Hansmeyer and Benjamin Dillenburger, illus-
consistencies can be controlled and defined at the trates a new aesthetic approach. During the design
design phase. The surface patterns are not depend- process, the architects used an algorithm based tech-
able on the original structural aesthetic of the mate- nique for generating complicated forms and highly
rial and the design can be very precise. Biomimicry, accurate surface structures. [2]
algorithm-based design or parametric design can be
a source of a new kind of detailed design, where the Material
scalability of the pattern can be almost unlimited. Almost all traditional raw materials in the construc-
tion industry have been applied somehow in 3D
printed prototypes. The printing material can be also Figure 2
a product itself, designed or developed as a commer- 3D printed Digital
cial material.The benefits of an artificial printing ma- Grotesque project,
terial mixture is that the material features can be con- architects Michael
trolled beforehand and the design can be material Hansmeyer and
driven. A deep understanding of the material fea- Benjamin
tures and physical behaviour opens up new possi- Dillenburger [1].
bilities in product design and architecture. The ad-
justable or controllable qualities of the printing ma-
terials can add extra value to the end products. When
adding smart components directly to the printing

material, smart materials can be developed. It is also be manufactured using traditional methods. Due to
possible to print on electrically conductive material this, the structures can be an essential part of the de-
or to include printed parts. Nanotechnology can be sign idea and visuality at an early phase. It is possi-
applied to the printing material creating products, ble to design strength simultaneously with architec-
which become, for example, fire resistant, very stiff tural design if using an iterative process of the opti-
or easy to clean. Focusing on the tactile and aesthetic mal form creation. However, the design of structures
aspects of the material can generate added value in go also hand-in-hand with the material development
the end products. In other words, concentrating on and aesthetic aspects.
the design shapes, colours, scents, haptic warmness,
hardness, softness or structure of the material itself
Figure 3
can lead to something unpredictable and novel. Ma-
Cool Brick,
terial development gives possibilities to think of the
Emerging Objects
lifecycle chain of the material starting from the be-
[3].
ginning, from the manufacturing of the raw mate-
rial. Recycling or using recycled printing material
can give added value to end-use applications. Future
possibilities may lie also in renewable, more reusable,
bio-based construction materials. The technical and
functional aspects, such as moisture resistance, in-
sulation abilities, sound insulation, air tightness, fire
resistance and structural stability, are crucial proper-
ties when thinking about architecture-related build-
ing products. The emphasis regarding those features
are on dependable defined product fields and reg-
ulations, legislation, certification and local methods
of using building materials. The material properties
are crucial in terms of the building's lifecycle in the
longer term.
Figure 4
Function of the tile, Case of cooling bricks. The material and structure of
Cool Brick, a specially designed brick (Figure 3, 4) can bind mois-
Emerging Objects ture and due to this, when warm air flows through,
[3]. the brick gets cooler. The idea is based to the Mus-
cateses tradition of cooling their buildings in a very
warm climate by using evaporative cooling. The
sponge-like fine structure of the tile and the large sur-
face area would be difficult to achieve using other
current manufacturing methods, such as traditional
moulding techniques. [3]
Structure
The new manufacturing method in architecture can
be an enabler of novel visuality in design structures,
and can lead to the creation of structures that cannot

Case of varied density. In nature structures are of-
ten optimized and inherently aesthetic. A scaled or
Figure 5
complicated imitation of nature can lead to a very
3D printed concrete
small patterned design, with good strength or even
and variations of
resilient structures. At the MIT Media Lab, the Medi-
density, MIT [4].
ated Matter Group have studied 3D printed concrete
of varying density (Figure 5). [5]
VALUE CREATION FOR END-USE APPLICA-

TIONS
In architecture, additive manufacturing can be
utilised in several types of product development
projects. Nevertheless, due to the current phase of
development, expectations of when the products Figure 6
could be launched will have a different time span Involute wall,
and depends on the target applications. Currently, printed using sand,
most of the projects concentrate on printing just a prototype,
modular element, such as a tile, a building block or designers Virginia
a wall element, creating added value for the existing San Fratello, Ronald
product fields with this new method. In addition, Rael, Emerging
3D printing can be seen as a method of manufac- Objects [6].
turing unique pieces or industrial products that are
not very easy to produce using traditional methods
and where the size of the production line is limiting.
In addition to these, additive manufacturing can, if
successful, frame design in a new way in the future. Figure 7
Seismically resistant
New Product Properties structures, Quake
The new type of production technique can enhance Column, Emerging
the three dimensional abilities of the products in ar- Objects [7].
chitecture. New characters, such as printed func-
tional features, can lead to something unprece-
dented in the construction industry. Novel functions
can be added to the traditional or brand new build-
ing product, for example, products that have been
printed using four dimensional printing. The fourth
dimension is time, when for example the printed
piece will change shape after the material reacts to
an external stimulus. As a result, this can lead to pro-
grammable and movable prints.
Case of functionality in the designed object. An
architect office called Emerging Objects has devel-
oped some prototypes of functional products, such

as the Involute Wall (Figure 6). The curvy forms of the ways of working, and the needs of end-users. Since
surfaces absorb and redirect sounds. The three di- the advent of hi-speed data transfer and the rise of
mensional design creates shades on the wall and the information technology, an architectural design can
structure stays cool on a sunny day. Other wall struc- be global, but resources or manufacturing can still be
tures, the Quake Column (Figure 7), is also optimized local. In additive manufacturing the printing material
three-dimensionally. The tile is designed to be seis- can be produced locally, and even respect local build-
mically resistant, and reacts optimally to movements ing material traditions. The local way to production
during an earth quake. The idea is based on the inter- means that design can once again respect and arise
locking stones of the Inca culture, where tiles were from local traditions, cultures, resources and needs.
allowed to have slight movements, without falling In addition, the manufacturing method can create
apart. [6, 7] new methods for architectural-scale rapid prototyp-
ing or three-dimensional visualisation that strength-
Figure 8 ens and supports the design process.
Low-cost
Case of three-dimensional visualisation. Additive
replacements for
manufacturing has been used in walls to visual-
obtaining an
ize previously destroyed historical decorative parts
impression of the
and recreating already lost visual information. In
original idea of the
Castello di San Martino dall'Argine, Mantova, Italy,
decoration [8].
conservation has been carried out on a missing putti-
figure's head (Figure 8). After generating a 3D im-
age of the head, the missing part was printed three-
dimensionally to obtain an impression of the putti-
figure with the head on. [8]
Figure 9 Case of emergency housing. Instead of moving
Detail of the facade, building parts or element from the factory to the
Anne Pheiffer site, the factory can be seen also as the mobile unit.
Chapel, original There have been many projects that have consid-
design Frank Lloyd ered the question of how additive manufacturing can
Wright. The mould be applied in rebuilding projects in disaster areas.
of the replaced tile An expected added value can be created when low
has been 3D cost and rapidly constructed sheds or buildings are
printed. [10] needed. With the co-operation of other partners,
the University of Nantes and L'institut de Recherche
en Communications et Cybernétique de Nantes be-
gan a research project where emergency houses are
printed in 20-30 minutes apiece. The area of an insu-
New Framework of Design lated shed is 3 m², and the height is 3 meters. [9]
In recent years there have been some visions where
additive manufacturing can be applied to serve daily New product fields and applications
life, ease life after natural disasters, or create cost- As mentioned before, when using this novel pro-
effective living environments for poor areas. The duction method, value can be achieved through the
changing situations of environmental or political is- strengthening of current product lines or through to-
sues affect new standards for living environments, tally new product concepts. In addition, the tech-

nique can be used to produce customised products DISCUSSION
or spare parts, joint structures with old products, or At this moment in history, large-scale additive manu-
to complete parts for the existing products. The tech- facturing industry is concentrated mostly on improv-
nique can also be applied to replacing the manufac- ing printers. For this reason, technical challenges re-
turing techniques of products lines already on the strict design thinking, the creative process, and also
market. Reasons for doing this can include better visions of 3D printing possibilities. Research and
and unique design possibilities, easier manufactur- product development has basically fallen into two
ing processes, a need for customisation, or environ- parts. Typically, engineers try to improve the techni-
mental issues related to the printing material or pro- cal abilities, while groups of designer-architects with
duction. Additive manufacturing can be seen as a a more creative approach develop end products and
possible method when the moulds are made for large concepts. Most fruitful would be a co-operation be-
structures. The technique has been used extensively tween engineers, designer-architects, and commer-
on a smaller scale. Moreover, using design thinking, cial companies that would develop together new 3D
totally new products based on new needs can be in- printable product concepts, following a normal real-
vented. The novel production method has a great life design process. Some examples of co-operation
possibility to create unique joint structures, with tra- have however already emerged. Presumably, in fu-
ditionally produced new or old building parts. How- ture the development of printers will be more de-
ever, designing new 3D printed parts to complete pendent on future visions and the process will be
old products that have been manufactured earlier or more design driven. Design can also be material
that have a more traditional appearance can cause vi- driven, when the environmental aspects and mate-
sual or usability malfunctions. That can be prevented rial selection will give added value to design appli-
by design thinking at the early phase. As time goes cations. Currently the tendency seems to be that al-
by, the 3D-printed products will age and aging will most everyone wants to be the first to market. This
change how the product is valued, with the value fad- eagerness will decrease the quality of printed parts,
ing or even growing (Ashby 2009, 66). Design con- with concepts being under-developed and value as-
cepts should support the aging of printed structures pects not given sufficient thought. If it is not clear
or building parts, and replacing a printed part or re- how value is obtained and how the new technology
pairing prints needs to be considered in advance. is justified, catching up later might be difficult. Value
Case of replacing a traditional method. In the case creation can benefit architectural design by creating
of the restoration of the Annie Pfeiffer Chapel (Fig- quality and precision in the basic elements of de-
ure 9), additive manufacturing was used because the signed objects. The freedom of creating design pat-
old and original artisan skills had disappeared. The terns can be seen, for example, even as a new kind of
building, dated 1941 and designed by Frank Lloyd craft. Printing will create added value also in designs
Wright, faced a problem because the tiles were al- where form, functions or shapes are not possible to
most impossible to reproduce, while traditionally manufacture by using traditional methods. Design-
used methods were too expensive. In addition, not ers can present creative, even intangible feelings or
enough competent artisans were available. The orig- reminiscences by means of additive manufacturing.
inal design of the facade consists of 46 different kinds The optimal form of creation of structures can imply
of tiles, which were originally cast using wooden less material, but also form-based imaginative archi-
moulds. Using 3D-printed moulds, it was possible to tecture that is yet implementable. Currently there are
manufacture the facade blocks with ease and at a rea- no standards, norms or legislation to define the stiff-
sonable cost [10]. ness of structures, the use of 3D printed products or
to guide 3D-printed construction projects. Calculat-

ing the added value is not yet possible. This broad, REFERENCES
wild and open scene can be seen as an advantage, Ashby, M. F. 2009, Materials and the Environment, Eco-
while almost anything is still possible and is yet to be Informed Material Choice., Elsevier Inc.
strictly defined. On the other hand, without certified Ashby, M. and Johnson, K. 2014, Materials and Design,
The Art and Science of Material Selection in Product
products, markets are difficult to conquer. There is
Design, Elsevier Ltd.
a risk for companies in starting a new product line if Hoskins, S. 2013, 3D printing for artists, designers and
the future standardisation is unpredictable. Appro- makers, Bloomsbury
priate concepts for architects and the construction [1] http://www.digital-grotesque.com/installation_i
sector need to be clarified, and the possibilities need mages.html?screenSize=1&color=1#1
to be articulated more precisely. Value-based analy- [2] http://www.digital-grotesque.com/concept.html?s
creenSize=1&color=1
sis of architectural design and end-use applications
[3] http://www.emergingobjects.com/project/cool-bri
can lead to the stable use of this new manufacturing ck/
method. [4] http://matter.media.mit.edu/tools/details/3d-p
rinting-of-functionally-graded-materials
[5] http://matter.media.mit.edu/assets/pdf/Publicat
CONCLUSION ions_FGRP.pdf
Additive manufacturing has great possibilities to en- [6] http://www.emergingobjects.com/project/involute
hance the quality of architecture and end-products. -wall/
The potential is immense, yet the method is not com- [7] http://www.emergingobjects.com/project/quake-
monplace, and research in architecture is still scarce. column/
To sum, the potential lies in the properties of mate- [8] http://www.digitalmeetsculture.net/article/3d-p
rinting-applied-to-cultural-heritage/
riality and the possibilities to design complex forms
[9] http://www.iutnantes.univ-nantes.fr/14355915450
and surfaces and to control or adjust different aspect 31/0/fiche___actualite/
during the design process. In the early stages, there is [10] http://www.smithsonianmag.com/arts-culture/rebu
a significant possibility to develop products and find ilding-frank-lloyd-wright-classic-3-D-blocks-
solutions using a more value-based approach. A flex- 180953007/?no-ist
ible manufacturing method can be a source of inspi-
ration as an enabler, and has the potential to create
value through the design process. In future, a de-
sirable research focus in architecture-related additive
manufacturing could be to find new and more sus-
tainable practical solutions and iterations. Additive
manufacturing can also be seen as a fresh way to en-
hance architecture, the materials industry, and the
construction industry, as well as a means to utilise
information technology more efficiently without di-
minishing environmental issues. In the coming years,
the interests of companies, researchers and design-
ers can be expected to work to identify more clearly
what the new trends are and where the real values
lies.

Groove Light
Adding Physical Reality to Virtual Projections Using 3D-printed Lanterns
Shinya Okuda1 , Lin Zhenyi2

1,2
National University of Singapore
1
akiso@nus.edu.sg 2 a0085912@u.nus.edu
How might 21st-century computational technologies enhance lighting

functionality in architecture? The Groove Light provokes relationships between
light and shadow, adding a new dimension to future lighting. A series of
distinctive and complex 3D-printed lanterns, which cast identical patterns of
geometric shadows, creates the optical illusion that they are floating above a
continuous, geometric carpet of shadows. The authors tested this concept in three
steps: 1. 3D printability, 2. a lighting test and 3. interactivity. The paper also
reports the effectiveness of a selective and custom support strategy for printing
overhanging geometries with fused deposition modelling (FDM) 3D printers and
further highlights differences between the computational models and physical
prototypes in the quest for 'point' light sources.
Keywords: Digital Physicality, 3D Printing, Self-supporting Geometry,

Stereographic Projection, Projection Mapping
INTRODUCTION We refer to the 3D-printed lanterns of Segerman and

Jun'ichiro Tanizaki's novel In Praise of Shadows, origi- Schleimer [1], who inversely applied a stereographic
nally published in 1933 in Japanese, expresses favour projection principle to define their spherical lanterns.
for obliqueness over brightness as one of the charac- Stereographic projection is an ancient mapping tech-
teristics of Asian culture as opposed to the Western nique to project a sphere onto a plane; however, we
preference for clarity. Typical Asian lanterns, express- hypothesized that the lanterns themselves would not
ing such subtlety with their translucent paper skins, necessarily need to be spherical in order to control
are often seen as more of a sort of lighting sculp- the projection onto a plane from a point light source.
ture than functional lighting equipment. Inspired by As long as precisely computed light rays intersect
the concept, we started to question the rationality with any kind of lantern geometries, shadow patterns
of unconditional brightness in contemporary life. If could be controlled in a more complex and three-
lighting is not only a matter of level of illuminance- dimensional manner.
the goal at which most current lighting simulation
tools primarily aim-how might 21st-century compu- GROOVE LIGHT
tational technologies enhance lighting functionality The Groove Light concept presents provocative rela-
in Architecture? tionships between light and shadow by using precise

Figure 1
Left: Groove Light
requires precise
relative spatial
positioning among
light sources,
3D-printed lanterns
and geometric
shadows. Right:
Overall rendering of
the Groove Light
layout. Five
distinctively
complex lanterns
lighting simulation models in which computational came more striking. Five distinctive lanterns are po- cast precisely the
designs and prototyping processes are instrumen- sitioned at precise distances in order to match their same geometric
tal. We invert the typical projection mapping pro- peripheral shadow patterns to each other. The over- shadows, creating
cess, which usually projects computed visual images all rendering of the layout demonstrates that it cre- the optical illusion
toward non-planar objects, creating optical illusions. ates the optical illusions that a series of distinctive that they are
Instead, we penetrate sculptural lanterns with light lanterns are floating above a continuous, geometric floating above a
rays toward geometric shadow patterns, which result carpet of shadows (Figure 1, Right). continuous,
in complex porous volumes with a series of groove- geometric shadow
like light cannons. If typical projection mapping aims 3D PRINTABILITY carpet.
to add virtual reality to physical objects, Groove Light We demonstrated the Groove Light concept using a
aims to craft physical reality from virtual projections large-scale 3D printer, BigRep One, one of the largest
(Figure 1, Left). commercially available Fused Deposition Modelling
Following several iterations using 3D CAD soft- (FDM) 3D printers that has output of approximately 1
ware (Rhino 5), the authors tested the concept over m3 (Figure 2, Left). The large-scale lanterns not only
the following three steps: 1. 3D printability of frame a public square with geometric shadows but
overhanging geometries, 2. a lighting test using also simultaneously have their own sculptural pres-
high-power point light sources and 3. interactiv- ence.
ity between the geometric shadow patterns and ob- One of the key challenges of large-scale 3D print-
servers. Instead of using a spherical interface, as in ing is how to materialize such complex forms in a
the case of the stereographic projection, we experi- cost-effective manner. In Digital Fabrication, we are
mented with several distinctive lantern shapes. We no longer designing the form that will ultimately
began with various rotational geometries, such as El- be produced, but the production process itself (Gra-
lipsoid and Cylinder, as it is relatively easy to con- mazio and Kohler 2008). As such, we tested the fol-
trol explicitly their degrees of overhang through their lowing three steps to design the lanterns and their
sectional profiles. Subsequently, we also verified the 3D printability: 1. self-supporting geometries that do
concept with asymmetric lantern geometries. not require additional support (Figure 2, Right); 2. ge-
We set an identical geometrical shadow pat- ometrical shadow patterns that result in fewer over-
tern for all individual lanterns so that the differences hanging elements; and 3. selective custom supports
among the lantern shapes, which are enabled by that require less material than the printer-default au-
computational design and fabrication strategies, be-

Figure 2
Left: Large-scale 3D
printer BigRep One,
with maximum
printing volume of
approximately 1
m3. Right:
3D-printing process
of one of the
Groove Light
lanterns with
controlled
overhanging
geometries.
tomated vertical supports. We took the overhanging constraints not only as
limitation but also as an opportunity to integrate the
Self-supporting geometries fabrication process into our lantern geometries. Ela-
Selected Laser Sintering 3D printing, which can the- dio Dieste took a similar approach in the Church of
oretically print any kind of geometry, is still a rela- Atlantida (Pedreschi 2000), which demonstrates the
tively costly process, with output size confined by the structural integrity of brick-layered walls in its con-
printing powder bed. In contrast, variations of Fused struction process (Figure 3, Left). On the other hand,
Deposition Modelling (FDM) 3D printing are gener- Contour Crafting (Hwang 2004) separated most of the
ally used for large-scale 3D printing experiments in horizontal elements in architecture, such as floors,
architecture, as the cost of printing is lower, and since ceilings and roofs, into prefabricated components in
output size is basically only limited by the machine's order to get around the limitations of 3D-printing
frame size. FDM extrudes various types of filaments overhanging geometries on architectural scales.
layer by layer in a technique similar to brick layer- We conducted experiments to test the maximum
ing. However, the most crucial limitation of FDM is cantilevering angle in our 3D-printing system, as this
its limited ability to print overhanging geometries, maximum is material- and design-specific. We occa-
which may require vast amounts of supporting mate- sionally succeeded in 3D printing up to 60 degrees
rial, just like the scaffolding required to support over- of overhang, which requires sensitive calibrations of
hanging elements during building construction. the print-head and filament-extrusion speeds. How-
Figure 3
Left: Church of
Atlantida by Eladio
Dieste (Pedreschi
2000). FDM 3D
printing is similar to
the brick layering
process. Right:
Lantern’s sectional
profile controls the
maximum degree
of overhang.
(Design: Chua, S.C.
and Chan, L.R.)
Figure 4
Left: Triangular
grid, funnel-shaped
lantern itself
requires 5 kg
material and 70
hours printing time
with additional
custom supports.
(Note: The image is
shown without
supports.) Right:
Square grid lantern
itself requires 2 kg
ever, as the geometries of the Groove Light lanterns tional profile. material and 28
vary, we set our design guideline to the much safer By comparison, the square-grid lantern itself re- hours printing time
maximum of 45 degrees of overhang. Figure 3, Right quires only 1.4kg total weight and 17 hours print- with selective
illustrates how the design guideline defines one of ing time without support, primarily because it has custom support,
the Groove Light lanterns, named Seashell, which is fewer branches. This is just over half the weight of indicated as
generated by rotational geometry and then split hor- the triangular-grid lantern. Additionally, as half the diagonal fillers
izontally into two parts. Each piece-top and bottom- elements are within the overhang limit and thus self- within square grids.
is 3D printed separately. The maximum 45 degrees supported, only selective parts require additional The square grid
of overhang is controlled by its sectional profile. Ad- custom support, resulting in total of only 2 kg weight lantern requires
ditional geometrical complexity is added to the rota- and 28 hours printing time. Overall, the square-grid 60% less material
tional body by carving out grooves for light rays. lantern requires 60% less material and printing time and printing time
than does the triangular-grid one (Figure 4, Right). than the one of
Geometrical shadow patterns triangular grid.
Secondly, we study how geometric shadow patterns Selective custom supports
affect the 3D printability of the lanterns, comparing Most FDM printers automatically generate vertical
triangular and square grids. We used a funnel-shape supports to compensate for their overhanging ge-
lantern of 60 cm height and diameter as a base geometry limitations, often significantly increasing the
ometry and then evaluated the total 3D printing time required amount of printing material and time (Fig-
and weight required to create triangular and square ure 5, Left). As such, we developed a custom sup-
shadow grids. Generally speaking, triangular grids, as port strategy, which enables us to support overhang
in the case of the truss, have better structural rigid- geometries within the sectional profiles of lanterns
ity, requiring thinner overall sectional profiles. We with the smallest possible amount of support (Fig-
computed that the triangular- grid lantern requires ure 5, Right). We studied two properties of custom
2.7Kg weight and 31 hours printing time without sup- supports: thickness and angle. Thinner and steeper
port (Figure 4, Left). However, as the majority of el- angles of support require less material, whereas sup-
ements exceed the limit of 45 degrees of overhang, ports that are too thin or too steep supports may
total printing weight increases to 5 kg and 70 hours not sufficiently hold up overhanging elements. Af-
printing time with additional custom supports (Sim- ter several experimental iterations, we identified that
plify 3D). This easily counteracts the advantage of po- 1.1mm thick, 4.5mm pitched supports minimize the
tential weight reduction from its thinner overall sec- total amount of printing while delivering satisfactory

Table 1
FDM 3D printing
time and weight
comparison of four
support strategies
for overhang
geometries.
Figure 5
Left: Generation of
automatic vertical
supports for
overhanging parts
by the default 3D
printing software
(MakerBot Desktop
3.8). Right:
Illustration of
selective custom
support strategy for
FDM 3D printing.
Red-lined elements
require support as quality. These custom supports also follow the over- size of the light-emitting element. However, bigger
they exceed our hang design guidelines, inclined no more than 45 de- light sources produce more light rays, thus tending
overhang design grees. to project rather blurred shadow edges. Computer
guideline, while We use the previously described funnel-shaped, rendering (Rhino 5), by contrast, uses ideal 'point'
green-lined square-grid lantern to test the effectiveness of the fol- light sources that have no dimensions and which can
elements are lowing four support strategies: 1. without support, 2. therefore render crisp shadow patterns. As such, in
self-supported. automatic support, 3. custom support, and 4. selec- designing and selecting intense 'point' light sources,
Blue dotted lines tive custom support. Table 1 compares their printing we need to be mindful of the gap between computa-
indicate selective time and weight. While automated support requires tional models and physical prototypes.
custom support, 10 times more material than the actual lantern itself, Three types of light sources were empirically
which drastically the selective custom support strategy only increased tested. Firstly, a traditional incandescent light bulb
reduce the total total weight 43% to 2 kg (Simplify 3D). merely projected blurred, unrecognizable shadow
amount of patterns, perhaps because its light-emitting wire fil-
supporting material A QUEST FOR 'POINT' LIGHT SOURCE ament has a certain length, generating multiple light
required compared Identifying a suitable light source for the Groove Light rays from supposedly a single origin of light.
to the automatic installation required balancing between the bright- Secondly, we tested a High-intensity Discharge
vertical supports. ness of light and the crispness of shadows. Generally (HID) lamp, a type commonly used for vehicle head-
speaking, lighting power rises in proportion to the lights. The light-emitting part of the HID lamp is
much more compact than that of the incandescent

Figure 6
Left: Full-scale
lighting test with
HID lamp for square
grid shadow. It
creates crisp
shadow patterns,
however the light
source is
unappealingly
bright and
generates excessive
heat. Right: A series
of energy-efficient,
light bulb, which means it is closer to an ideal 'point' speaking, smaller LED chips cast crisper shadow pat-
custom-made LED
light source. The lighting test showed impressively terns, being closer to an ideal 'point' light source;
light prototypes of
intense and crisp shadow patterns; however, the however, their light is considerably dimmer. Though
various power
light source itself is unappealingly bright and gener- the selection of suitable LED chips was not straight-
(15–25W) and size
ates excessive heat. Perhaps the HID lamp is too small forward, after several trials, we concluded that 25W
(10–25mm
for the level of illuminance meant to be projected. It LED chips with 20 mm diameter could achieve suffi-
diameter). LED
also seemed too dangerous to expose such a lamp cient, if not perfect brightness and acceptable, if not
chips are soldered
without its proper casing (Figure 6, Left). flawless crispness of the shadows (Figure 6, Right).
directly onto
Lastly, we tested a Light-emitting Diode (LED)
aluminium heat
chip. Generally speaking, the LED is much more INTERACTIVITY sinks.
energy-efficient than the HID lamp, thus producing Technically speaking, Groove Light comprises three
less heat while still being sufficiently bright to project elements: 1. point light sources, 2. a geomet-
shadow patterns over several meters. However, un- ric shadow pattern and 3. 3D-printed light grooves
like the HID lamp, LED chips glow over certain surface that project the shadow pattern from the point light
areas, resulting in relatively blurred shadow patterns. sources. The project requires precise spatial relation-
In other words, the larger LED chips project brighter ships among those three elements in order to cast
light but less crisp shadow patterns. Conversely the intended shadow patterns. Conversely speaking,
Figure 7
Left: Dynamic
shadow patterns
reflect various
interactions with
the observers.
Right: Illustration of
Pull mechanism.
(Design: Bek, T.K.,
Lim, B.W. and Tan,
Y.T.)

Figure 8
Interactions with
the public provoke
relationships
between light and
shadow enhanced
through
computational
design & fabrication
technologies.
(Design Left: Bek,
T.K., Lim, B.W. and
Tan, Y.T. Right: Kui,
H.G., Lee, G.L. and
Wong, W.)
Figure 9
Groove Light at the
international
lighting festival in
Singapore, i Light
Marina Bay 2016,
which attracted
over 700,000
visitors.

any slight deviations in the relationships may drasti- site management; BigRep GmbH for technical sup-
cally change the shadow pattern. We took this fact port and sponsorship; Takeo Sugamata from SWITCH
not as a constraint, but as an opportunity to create for the lighting consultation; and Rapid Prototyping
interactive relationships with prospective observers module participants from the Department of Archi-
(Figure 7, Left). tecture, School of Design and Environment, National
We experimented with three types of interac- University of Singapore for the design and fabrication
tions: 1. pull actions, moving the point light source of the Groove Light project.
vertically away from the 3D-printed lantern (Figure 7,
Right); 2. rotation of the 3D-printed lantern around REFERENCES
a fixed-point light source; and 3. swinging the 3D- Gramazio, F and Kohler, M 2008, Digital Materiality in
printed lanterns to create ephemeral shadow pat- Architecture, Lars Müller Publishers, Baden, Switzer-
terns similar to the surface of water. As a whole, land
the Groove Light shadow patterns dynamically corre- Hwang, D and Khoshnevis, B 2004 'Concrete Wall Fabri-
spond to various degrees of interaction between the cation by Contour Crafting', 21st International Sym-
posium on Automation and Robotics in Construction,
3D-printed lanterns and the observers (Figure 8).
Jeju Island, Korea
Pedreschi, R 2000, Eladio Dieste: The Engineer, Thomas
CONCLUSION Telford, London
The Groove Light project presents a new type of [1] http://www.theguardian.com/science/alexs-advent
ures-in-numberland/2014/oct/30/pumpkin-geomet
sculptural lighting equipment fabricated through
ry-stunning-shadow-sculptures-that-illuminate
advanced computer-aided design and 3D-printing -an-ancient-mathematical-technique
technologies. Technically speaking, Groove Light
could consist of any combination of lantern shapes
and geometric shadow patterns. However, FDM
3D printers have major limitations with overhang-
ing geometry, tending to require disproportionate
amounts of additional support materials. Controlling
overhang elements, choosing square shadow grids
and applying selective custom supports can optimize
the printing time and weight of the lanterns, which
is critical for any large-scale 3D-printing project. A
selective custom support strategy and the lessons
learned from the quest for a 'point' light source could
also be applied to projects on an architectural scale.
The Groove Light project was invited to a show-
case at the renowned international lighting festival
in Singapore, i Light Marina Bay 2016 (Figure 9).
ACKNOWLEDGEMENTS
We would like to express our sincere gratitude to the
following dedicated collaborators: the Urban Rede-
velopment Authority in Singapore and curator Randy
Chan for the invitation to the i Light Marina Bay 2016;
Pico Art International Pte Ltd for the installation and

FABRICATION | Robotics: Design &
Assembling
Building Traditions with Digital Research
Reviewing the Brick Architecture of Raúl Hestnes Ferreira through Robotic
Fabrication
Rui Oliveira1 , Jose Pedro Sousa2

1,2
Faculty of Architecture, University of Porto + DFL/CEAU
1
rui.santosoliveira@gmail.com 2 jsousa@arq.up.pt
Brick construction has a strong tectonic tradition in architecture, being used both
as a structural and as an expressive material. Despite several technological
innovations at the composition and production level, its application still relies on
talented craftsmanship, which has some natural human limitations and has
becoming harder to find in the present days. To overcome this problem, robotic
assembly technologies have been introduced in the field, opening new design and
construction possibilities. In this context, this paper intends to examine their
application but from a different perspective, by examining how they can be used
to connect with the traditions in brick construction. To do so, it presents and
analyses the work of Portuguese architect Raúl Hestnes Ferreira, and develops a
computational design and robotic fabrication research on the topics of corner,
column and dome bricks. The production of a column design at the 1:1 scale
using an automated process serves to reflect on the relevance of new technologies
to innovate in accordance to tradition.
Keywords: Brick Construction, Hestnes Ferreira, Robotic Assembly,

Computational Design, Digital Fabrication
INTRODUCTION effects. More than simply constructing with bricks, ar-

Bricks are a man-made material, which has been used chitects can design with bricks, as Louis Kahn contin-
in architecture since the ancient times of the build- uously referred to on his lectures and texts (Kahn and
ing construction history. Over time, the evolution Twombly 2003).
of its composition and production possibilities, from
crafts to industrial ages, never ceased to stimulate de- The integration of robotic technologies
signers to explore not only their structural quality but Aside with this creative intention, architects had also
also, their aesthetic potential. On his recent book, to devise ways to communicate and instruct workers
William Hall (2015) collects a series of notable build- on how to lay down the bricks. Until recently, this has
ings where architects creatively employed such small always been a manual process requiring talented and
and regular components to achieve unique tectonic dedicated craftsmanship. But in 2005, Gramazio and
FABRICATION | Robotics: Design & Assembling - Volume 1 - eCAADe 34 | 123

Kohler (2007) started to implement robotic technolo- on robotic technologies (Sousa 2015), and introduces
gies to perform the assembly of brick structures. By the work of Raúl Hestnes Ferreira, who is a renowned
recognizing the human limitations in understanding Portuguese architect who mastered the use of bricks
and executing intricate brick arrangements that can in his buildings. By studying his work, it describes a
be easily conceived with computers, they showed research on using computational design and robotic
how robots could be used to expand the design in- technologies to review and extend his design inter-
terests into a larger domain of geometric complex- est on brickwork. By preparing and discussing the
ity. However, aside with this creativity-oriented ap- experience with the architect himself, it was possi-
proach, another technology innovation path has also ble to reflect and validate better the relevance of us-
been followed. The repetitive brick laying procedures ing such advanced technologies in the current days.
when building regular structures are also an evident Aside with this goal, this paper also presents a fully
scenario for potential automation. That is the exam- automated system for robotic brick assembly, which
ple of the Semi-Automated Mason (SAM) robot de- is also another contribution to the field.
veloped by the company Construction Robotics (Pet-
ters and Belden 2014), which claims to achieve a high THE BRICK ARCHITECTURE OF RAÚL HES-
level of productivity. What is interesting to observe is
TNES FERREIRA
the fact that robots are being explored as a solution
Raúl Hestnes Ferreira (b. 1931) is a preeminent fig-
for both standard (i.e. repetitive) and non-standard
ure of Portuguese contemporary architecture, espe-
(i.e. variable) design scenarios.
cially in the second half of the 20th century. His
very unique style was marked by a strong concern
Rethinking traditions with the building details (Neves 2002). This interest
When new technologies can open new design and
derivates from the influence of a life experience in the
construction opportunities, it becomes also impor-
United States in the 60's, where he post-graduated at
tant to examine the way they can contribute to re-
the University of Pennsylvania and worked for Louis
think, or update, architectural traditions. In this disci-
Khan, and the personal search for incorporating Euro-
pline, technology is not an end but a mean, and only
pean values. His built work exhibits elementary com-
by working in both research directions it is possible
position methods, based on symmetry, axiality, pro-
to understand the whole spectrum of its disciplinary
portion, basic shapes and historical values, as an ar-
impact. The work on the Palladian Grammar by Stiny
chitectural alternative to Modernism (Tavares 2003).
and Mitchell (1978), the Malagueira houses by Duarte
By carefully realizing the physical properties and lim-
(2001), or the recent one on the Alberti's treatise by
itations of the materials, Raul Hestnes Ferreira looks
Kruger (2015), are just a few examples of such interest
for taking the most out of their design potential. This
in using digital technologies to revise traditions. But
tectonic approach is noticeable on his brick build-
perhaps the most brilliant applied work at this level
ings where, aside with the influence from Kahn, he
has been the one conducted by Mark Burry in the
invokes and reinvents the traditional Portuguese ar-
extension works at the Sagrada Familia in Barcelona,
chitecture (Tavares 2003).
where both digital design and fabrication tools have
been orchestrated to understand and continue the
Two reference brick buildings
legacy of Antoni Gaudí in the present days.
Among the brick buildings designed by Raúl Hestnes
Ferreira, the Municipal Library in Moita (1986-1997)
The paper and the House of Culture in Beja (1975-1985) were se-
Following this last trend, this paper summarizes and
lected as case studies for this research work (Figure 1).
extends a Master Thesis (Oliveira 2015), which was
In both buildings, the structural and expressive po-
developed in the scope of a larger research project
124 | eCAADe 34 - FABRICATION | Robotics: Design & Assembling - Volume 1

tential of bricks were carefully intertwined. In the Mu- brickwork with some elements placed perpendicular
nicipal Library, Hestnes Ferreira employed the mate- to their surface. The structural imposition was thus
rial in several interior and exterior applications. In managed in favor of the tectonic expression that can
specific wall corners and columns the architect care- be visible in the interior.
fully designed the brickwork to deploy original ma-
terial effects. In the House of Culture, Hestnes Fer- Motivations for digital research
reira employed bricks in the construction of domes The work of Hestnes Ferreira provides a rich demon-
as a reference to the traditional brick vault structures stration of a tectonic approach to brick construction.
that can be found in the region. Due the necessity Thinking in regular patterns and using hand drawing,
of defining a thicker base in the domes to support he found the way to translate his design intentions
their own weight, the architect conceived specific and prescribe the corresponding rules for laying the
Figure 1
Brickwork details of
a column and a
corner in the
Municipal Library
building in Moita
(top). One of the
domes in the House
of Culture in Beja
and the detail of its
brickwork, which
was specifically
designed for
structural purposes
(bottom).

bricks. By then, it would be very difficult to explore THE DIGITAL RESEARCH
more intricate arrangements with the existing repre- The digital research process unfolded in two sequen-
sentational tools. Furthermore, on the construction tial steps: a computational design exploration fol-
side, he started to become concerned with finding lowed by a robotic fabrication experiment.
talented craftsmanship to assure the production of
more complex brickwork structures, like the domes Computational design research
in the House of Culture (Ferreira 2002). Thus, with the This work started by looking for the main brickwork
digital technologies available today, it seemed op- themes found in the reference buildings described
portune to investigate how: before: the corner, the column and the dome. By
capturing the design intention presented on the ex-
isting constructions, the authors developed a com-
• Computational design tools can support ar- putational design exploration using the software
chitects in the tectonic exploration of brick Rhinoceros and the Grasshopper plugin (Figure 2).
structures, by expanding the design space of Supported by conversations with the architect
geometric possibilities; (Oliveira 2015), the methodology followed the initial
• Robotic assembly processes can provide an codification of the geometric rules, the parametric
efficient mean to assure complex traditional generation of alternative solutions, and finally, their
or novel brick constructions in a moment of formal verification through the production of 3D
scarce skilled labor. printed models using FDM technology. This digital-
to-material continuity was essential to examine the
Figure 2
Summary of the
computational
design exploration
developed on brick
corners, columns
and domes.

tectonic quality of the many different generated so- Robotic fabrication research
lutions to study the corner, column and dome situa- The results of the computational design research
tions (Figure 3). were presented to Raúl Hestnes Ferreira who ac-
knowledged the relevance of using such methods to
Figure 3 operate with rule-based design interests. Further-
The parametric more the 3D printed models proved to be an effi-
exploration of cient way to communicate and evaluate non-regular
column designs brick structures. Facing such information, the archi-
using 3D printed tect was invited to select one of the digitally gener-
models to evaluate ated solutions to be robotically fabricated at 1:1 scale.
each solution. The Attracted by the rhythm of its torsion and subtle vari-
same procedures able openings, his selection fell on a twisted column
were followed to design (i.e., the column in the bottom of Figure 3). On
explore corner and his words, "it has a certain complexity in which what
dome designs. one sees from one side is not exactly the same of what is
seen from another side. This piece is very beautiful and
variable (...) I find it simple and complex. It's a good
work." (Oliveira 2015). For this reason, this structure
was named as the "Hestnes" column. A quick ini-
tial experiment trying to manually assemble ceramic
bricks, made clear the difficulty in materializing its
variable geometry conventional means (Figure 4).
Figure 4
Manual assembly
experiment of part
of the Hestnes
Column to realize
the degree of
complexity of its
geometry.

Thus, the development of a robotic assembly automated process. After laying one level of bricks,
research became relevant to test the physical pro- the robot was paused to allow the manual deposi-
duction of the Hestnes column. This research part tion of glue on the top of the bricks, before running
was run in the Digital Fabrication Laboratory (DFL) the pick and place operation of the next upper level.
at the Faculty of Architecture of the University of This was done in a first moment when the glue spray
Porto (FAUP), using a large size KUKA robot equipped system was not yet available. The second experiment
with a vacuum gripper mounted on it. Due to the took advantage of the fully automated system, which
lack of equipment in the laboratory to carry heavy consisted in the recursively doing the following steps,
weights, the ceramic bricks were replaced by the use until finish the desired construction (Figure 6):
of EPS ones, cut in the real size (i.e. 200x120x50mm).
The rest of the fabrication setup included a long tray • pick the brick in the tray by turning the vac-
to feed the bricks for the picking operations and a uum on;
wooden base in the ground for placing them. To al- • move it inside of the box and wait for 1 second
low for a complete automated assembly process, an to be sprayed with glue from below;
original system for spraying glue was devised and in- • move outside of the box and place the brick in
stalled between the feeding tray and the assembly the specific position over the base by turning
base. All fabrication instructions were defined us- the vacuum off.
ing the KUKA|prc plugin for Grasshopper, which con- With its 246 bricks and 2.10 meters height, the Hes-
verted the movement trajectories and speeds, and tnes column took 52 minutes to be assembled in the
the on/off activation of the tools (i.e. gripper and glue semi-automated process, and 57 minutes in the fully
spray) in the KRL language of the robot (Figure 5). automated process, running at 75% of the maximum
The robotic fabrication of the column unfolded speed (Figure 7). Despite this minimal time differ-
in two experiments. The first one explored a semi- ence, the second test avoided the pause for man-
Figure 5
The digital
parametric design
environment that
supported the
design and robotic
fabrication of the
Hestnes column. It
is possible to see
the preview of all
pick and place
trajectories with a
stopping moment
inside the glue
spraying machine.

Figure 6
The key moments
of each robotic
assembly
movement: picking
the brick from the
feeding tray, gluing
its bottom surface
in the spraying
machine, and
placing the brick on
its specific position
in the assembly (top
row). The column
near completion in
the setup at the
DFL, with the glue
spraying machine
on the left.

ual intervention, which made it a safer and human CONCLUSION
fatigue-free process. Although this was the only so- The research work presented in this paper explored
lution that was materialized, it was possible to realize the possibility to employ robotic fabrication pro-
that any other design variation would require a simi- cesses to revise the traditions of brick architecture.
lar time to be built by the robot, since it has the same To do so, the authors studied the built work of Por-
amount of bricks. tuguese architect Hestnes Ferreira, and discussed
with him the tectonic interests that guided his brick-
Figure 7
work designs.
Photo of the
By using computational design tools, it was pos-
robotically
sible to encode the rules of the existing solutions and
fabricated Hestnes
generate new alternatives with a higher degree of
Column.
geometric complexity. Without the use such digital
processes, architects would have limited representa-
tion means to think, evaluate and describe certain
tectonic solutions. In continuity with building tra-
ditions, the computational design research revealed
the potential of such technologies in expanding the
creative space of design exploration in the present
times. Facing the design possibilities allowed by
computational design, the lack of talented crafts-
manship opens the door for using robotic technolo-
gies for brick assembly purposes. In the past, Hestnes
Ferreira described and told the workers the simple
rules they had to follow to build the expressive effects
of his brick details. In a similar fashion, the architect
can now assure their design intention by dictating
the assembly rules to the robot with the help of dig-
ital programming. Like in the domes of Casa da Cul-
tura, brickwork design is not only essential for serving
aesthetic purposes but also for achieving structural
efficiency. Thus, brick construction nowadays can
benefit from the use of computational design pro-
cesses to define customised brickwork arrangements
for more complex structural challenges.
Moving to a more generic level, the research
work demonstrates how the articulation of digital de-
sign and fabrication technologies can support an ex-
ceptional design freedom without loss the control
of its materialization. Recalling the 50 minutes that
took to fabricate the Hestnes Column, one can imag-
ine the significant number of columns that could be
robotically built in a single day, independently of
their geometric complexity or variation. The use of

robotic assembly technology seems thus very appro- Kahn, L. and Twombly, R.C: 2003, Louis Kahn: essential
priate to suit prefabrication production logics. In the texts, W.W. Norton
construction site, and despite some advancements Kruger, M., Duarte, J.P. and Moniz, G.C. (eds) 2015, Joelho
n.5 - Digital Alberti: tradition and innovation, EDARQ,
like the SAM robot mentioned before, it still seems
Coimbra
difficult to be able to explore the flexibility of oper- Neves, J.M. 2002, Raúl Hestnes Ferreira: Projectos 1959-
ations that can take occur within the controlled and 2002, Edições ASA
safe environment of the factory. Oliveira, R. 2015, Possibilidades de desenho e construção
Drawing from the experience described in this digital em tijolo a partir da obra de Raúl Hestnes Fer-
paper, future research would look to adjust the labo- reira, Master's Thesis, Faculty of Architecture, Univer-
sity of Porto
ratory conditions to use real ceramic bricks. Although
Petters, S. and Belden, R. 2014, 'SAM, the robotic brick-
this objective does not imply any substantial change layer', SMART/Dynamics of Masonry, 1(4), pp. 10-14
in the actual process, different glue will have to be Sousa, J.P. 2015, Robotic Technologies for Non-Standard
incorporated in the system. At the same time, a sen- Design and Construction in Architecture, Faculty of Ar-
sor for measuring the structure during its assembly chitecture, University of Porto, Porto
should be also considered to allow real-time adjust- Stiny, G. and Mitchell, W.J. 1978, 'The Palladian Grammar',
Environment and Planning B, 5, pp. 5-18
ment of the robotic operations. This research avenue
Tavares, M.A. 2003, Além da Geometria, Edições Estampa,
is a very important one, especially when dealing with Lisbon
larger structures where the accumulation of small de-
viations that are not anticipated in the computer (e.g.
brick size variations, structural deflection, glue thick-
ness) can become critical in the building process.
ACKNOWLEDGMENTS
The work presented was co-financed by the Euro-
pean Regional Development Fund (ERDF) through
the COMPETE 2020 - Operational Programme Com-
petitiveness and Internationalization (POCI) and na-
tional funds by the FCT under the POCI-01-0145-
FEDER-007744 project, and the previous PTDC/ATP-
AQI/5124/2012 research project. The authors are also
greatful to industrial partner Cerâmica Vle da Gân-
dara and to architect Raúl Hestnes Ferreira.
REFERENCES
Duarte, J.P. 2001, Customizing mass housing: a discursive
grammar for Siza’s Malagueira houses, Ph.D. Thesis,
Massachusetts Institute of Technology, Department
of Architecture
Ferreira, R.H. 2002 'Conhecer o Tijolo para Construir a
Arquitectura', Seminário sobre Paredes de Alvenaria,
Porto, Portugal, pp. 111-132
Gramazio, F. and Kohler, M. 2007, Digital Materiality in Ar-
chitecture, Lars Muller Publishers, Baden
Hall, W. (eds) 2015, Bricks, Phaidon Press Limited, London

Biomimetic Robotic Construction Process
An approach for adapting mass irregular-shaped natural materials
Chi-Li Cheng1 , June-Hao Hou2

1,2
National Chiao Tung University, Hsinchu, Taiwan
1
micky@arch.nctu.edu.tw 2 jhou@arch.nctu.edu.tw
Beaver dams are formed by two main processes. One is that beavers select proper
woods for constructing. The other one is that streams aggregate those woods to
be assembled. Using this approach to construction structure is suitable for
natural environment. In this paper, we attempt to develop a construction process
which is suitable for all-terrain construction robot in the future. This construction
process is inspired by beavers' construction behavior in nature. Beavers select
proper sticks to make the structure stable. We predict that particular properties of
sticks contribute gravity-driven assembly of wood structure. Thus, we implement
the system with machine learning to find proper properties of sticks to improve
selection mechanism of construction process. During this construction process,
3D scanner on robotic arm scans and recognizes sticks on terrain, and then robot
will select proper sticks and place them. After placement, the system will scan
and record the results for learning mechanism.
Keywords: Biomimetic Design, Machine Learning, Natural Material, Point

Cloud Analysis, Robotic Fabrication
INTRODUCTION industrial revolution, having a negative impact on

Contemporary construction process is good enough, the environment as a whole [2]. One of the biggest
Today's construction projects are characterizing by challenges of natural materials to be used in digital
short design and build period, increased demands crafting process is its unevenness and controllability
of quality and low cost. These problems can be (Chen and Hou 2016). Therefore, architects usually
approached by a flexible automation using robots give up using natural materials in their design.
based on computer assisted planning, engineering
and construction management (Bock 2008). How- Local natural materials
ever, there are some shortcomings. Construction Rivers bring and erode trees into mass driftwood es-
waste is becoming a serious environmental problem pecially in Taiwan's typhoon season. But, it is difficult
around the world; This is due to the growing in vol- to utilize by conventional industrial process; If we can
ume and complexities of modern economies (Erekpi- make good use of these natural materials instead of
tan et al. 2015). Using natural materials for architec- throwing them away, we can save lots of resources
tural designs was abandoned during and since the and provide architectural field with new possibilities.

Inspiration from nature for adapting natural softness materials. Those men-
On the Earth we live in, the ecosystem works without tioned cases also explore how to use robotic arm
waste. Living creatures develop ways to build shelter for adapting complex properties of natural material.
by following the character of natural material. For ex- Those research involved material and behavior levels
ample, termite mounds, bird nests and beaver dams of biomimicry in architecture.
(see Figure 1b). Beavers especially make good use Woodchip Bran project of The Architectural
of natural materials for large-scale construction. The Association School of Architecture(AA) utilized
biggest scale of beavers' construction is a beavers' irregular-shaped materials to construct the main
dam located in Wood-Buffalo-NP, north-east Canada structure of a barn[5]. The team used optimized al-
(see Figure 1a). The dam's width reaches 850 me- gorithm to figure out a suitable placement. Then,
ters [3]. This scale is comparable with human being's employed robotic arm milling to form the connec-
construction. Beavers' construction can be general- tions solving the design problem of using branch.
ized into three types such as beaver dams, lodges, Additionally, Robotic Softness project of ITECH chal-
and food caches. The amazing scale and variety of lenged softness materials to construct woven space
construction is rare in nature. Hence, beavers are re- improving the possibility of combining 3D scanner
garded as an architect in the animal world. and robotic arm for adapting natural softness mate-
Beavers' construction can even adapt its struc- rials [6]. Rock Print Project is the first architectural
ture to hydro-logical environment and surrounding installation to be built from low-grade granular ma-
forest species effects (Janiszewski 2006). The main terial and constructed by robotic machines. Con-
structure is composed of woods. Mud, weeds, and ceived as an intriguing vertical object, the instal-
tiny sticks can be used as infill. In fact, the comple- lation presents a new category of random packed,
tion of beavers' construction is attributed merely half potentially fully reusable, poly-dispersed jammed
to beavers' construction skills. Self-assembly mech- structures that can be automatically fabricated in
anism of materials adapting river force makes half non-standard shapes [7].
contribution to the construction (Blersch and Kan- Those mentioned cases also explore how to use
gas 2014). Both beavers' construction skills accumu- robotic arm for adapting complex properties of nat-
lated with experiences and self-assembly mechanism ural material. Those research involved material and
of materials play essential role in the construction. If behavior levels of biomimicry in architecture. How-
those principles of beavers' construction are grasped, ever, Robots are still difficult to utilize local resource
it is possible to make construction robots utilizing lo- and construct without materials processing.
cal resource for full-scale buildings.
RESEARCH OBJECTIVES
Robotics Fascination In this research, we investigate how the robotic
In other cases of using natural materials, the Wood- construction utilizing unprocessed natural materials.
chip Bran project of AA's team challenges using The dimensions of beaver's constructions reach hun-
branch to construct barn in main structure, they used dreds of meters long, as a result, using unprocessed
optimized algorithm in this process of project for natural materials as construction materials is a poten-
finding suitable arrangement then employed robotic tial approach for artificial buildings. However, human
arm milling to form the connections solving the de- beings depend on low technology and conventional
sign problem of using branch. Additionally, Robotic design methodology. This research proposes a solu-
Softness project of ITECH challenged softness mate- tion to solve the design problems.
rials to construct woven space improving the pos- The system of biomimetic robotic construction
sibility of combining 3D scanner and robotic arm process is designed to simulate beavers' construc-

Figure 1
A:Largest beaver
Dam in the
world[4].B:Large
beaver
dam[1].C:Beaver[1].
tion, including beavers' construction skills and self- SYSTEM DESIGN AND PROTOTYPING
assembly mechanism of materials. Therefore, the sys- In this research, we investigate two key techniques of
tem contains two major parts. One part is to arrange biomimetic robotic construction process. One tech-
materials placements automatically. The other part nique is to make robotic arm know materials char-
is finding appropriate materials by machine learn- acteristics, including curvature, diameter, and length
ing mechanism. This mechanism can judge whether through 3D scanner. The robotic arm can also lo-
current materials are suitable. Unlike conventional cate the materials and grip them precisely. The other
industrial process, which involves processing raw technique is to sense the structure stability after plac-
materials, this biomimetic robotic construction pro- ing materials and activate machine learning mech-
cess can save mass resource by utilizing unprocessed anism. This mechanism will give priority to better
irregular-shaped natural materials like sticks (see Fig- material characteristics when selecting construction
ure 2). materials (see Figure 3).
Figure 2 Prototyping
Concept The prototype is established on common tools that
development of usually appear on academic and industrial fields.
biomimetic robotic Therefore, it is convenient to be modified by other
construction people. For instance, using 6 degrees of freedom
process. robotic arm (KUKA KR6R700) usually appearing in in-
dustrial production and common parametric design
Figure 3 software (Grasshopper) used in design field as devel-
The robotic arm can opment platform. In the tendency of architects using
also locate the digital design and fabrication, the workflow combin-
materials and grip ing parametric software and robotic fabrication be-
them precisely. came more and more popular, we expect that devel-
oping biomimetic robotic construction process in it
as open source can accelerate this construction pro-
cess practiced in real environment.
As a result, in later versions, the robot fist scans
all materials. Next, every time it places each stick,
it scans all materials again. This revision helps the
robot take instant reaction to environment changes.
Besides, in the beginning, we use computer vision

to recognize the shape of sticks and use gravity sen- Figure 4
sor on gripper to sense the center of gravity of ma- System of
terials. By doing so, we can use physical simulation biomimetic robotic
program to find appropriate placements. However, construction
those methods are failed, because physical simula- process.
tion may optimize construction for fewer, large, or
regular-shaped materials. Nevertheless, in this re-
search, we focus on mass, small, and irregular-shaped
sticks for building structure as beaver lodges. Ac- Figure 5
cording to our experiment, physical simulation is un- The framework of
suitable for mass, small, and irregular-shaped sticks. system.
Gravity is ineffective in small-scale sticks structure but
curvature and length are keys to construct success-
fully. Therefore, we use 3D scanner instead of gravity
sensor to scan sticks because it is useful to analyze
those properties from point cloud data.
The completed version of system generalizes se- Figure 6
ries components from various incidents on Grasshop- Workspace is made
per (GH) for adaption. For example, when construc- of sawdust on a
tion robot perceives structure sway, robot will adopt wood tray.
infill program gripping sawdust from terrain as infill
to strengthen structure (see Figure 4). In this way,
construction robot can adapt various sticks and ter-
rain to improvise a stick structure as beaver lodge.
Components of system
Figure 7
The followings are components in the system of
Robotic arm and
biomimetic robotic construction process (see Figure
gripper assembly.
5).
Natural materials
The natural materials in this research are about 20-
centimeter long and irregular sticks (without any
branches). These sticks are used as scaled-down drift-
wood, which are common natural materials in Tai- Figure 8
wan. Typhoons and rivers erode driftwood into sticks Gripper assembly.
without any branches. If we can make good use of
these natural materials as construction materials in-
stead of throwing them away, we can save lots of re-
sources and provide design field with new possibili-
ties.

Figure 9
The robot sense
environment by
Kinect.
Workspace ior, and the component of machine learning. Point

A workspace simulates complex natural environment clouds processing first deals with information from
to test the adaption of construction robot. It is made Kinect (see Figure 9).
of sawdust on a wood tray (see Figure 6), the sawdust Firstly, the soft program receives points outside
terrain is easy to be shaped, as a result, the workspace the tray, and these points will be filtered. However,
can be reused for different rugged terrains. A soft of there may be some unreasonable points in the filtrate
sticks will put into the workspace randomly captured because Kinect IR points may not be detected, which
by gripper as construction materials during experi- causes the errors of point clouds location. The com-
ence. ponent of machine learning can transform received
points into mesh by Delaunay Mesh (an algorithm).
Robotic arm and gripper assembly By calculating each mesh area, we can know the ac-
The prototype of biomimetic construction robot con- curacy of point distribution. Besides, errors may exist
sists of KUKA kr6r700, KinectV2, and a material- when the gripper moves back to the observation po-
adaptive gripper (see Figure 7). sition. The component of machine learning can also
This gripper is made of PLA through 3D printer solve this problem through measuring the original
(see Figure 8). The robot makes good use of PLA's position of tray and moving the observed points to
high flexibility to adapt natural materials' various that original position. Then, the component exports
sizes. To add friction, abrasive papers are stuck to the data and its job is done. (see Figure 9) left is a real
the joints of the gripper and materials. The servo picture, and the right is the scanned data from points
empowers the engaging lever and transform rotation cloud.
torque into holding power. Robotic arm's IO controls
Figure 10 the gripper, and send digital information to Arduino.
The Environmental Arduino analyzes digital signals and commands the
sensing component servo on the gripper. The maximum weight for the
captures gripper to lift is 500 g and from 1 mm to 30mm in
geometrical length.
properties.
Environmental sensing The component captures geometrical proper-
The soft program involves point cloud processing, ties. These geometrical properties include materi-
geometrical poverties, robotic construction behav- als' position, length, curvature, and radius. They are

Figure 11
A:Terrain analysis.
B:Stability analysis.
Figure 12
Sticks are stacked to
form a structure
step by step during
construction.
obtained from the analysis of point clouds distribu- area is, the more unstable the structure is (see Figure
tion (see Figure 10). Nearby points can be viewed as 11). All geometrical properties will be recorded and
a group. By analyzing whether the groups present analyzed during construction process.
a linear layout, the component can judge if an ob-
served object is a stick. The point clouds data can be Figure 13
further analyzed by Circle Fit (an algorithmic mecha- This process is
nism) to get the object's curvature. Its length comes similar to FDM.
from measuring point clouds distribution centerline,
whereas the radius is calculated from the distance be-
tween centerline and point clouds.
The component also finds a flat area as a proper Figure 14
construction site. By analyzing point clouds, it can A: This gripper can
test the stability of the structure when placing each grip sawdust and
material. For example, through comparing new sticks.B:Sawdust
scanned data (material's actual place) with previous mode.C:stick mode.
scanned data (material's expected place), the compo-
nent can calculate a swept area. The bigger the swept

Procedural construction process until the circle is too small. Then, construction is fin-
The procedural construction process component can ished. It is expected that when the materials are with
dominate the execution order of other components. high quality, the robot will construct a towering and
For example, when we first start up the robotic con- tall cylinder, whereas it will be a short and low arch if
struction behavior component, the robot arm will the materials are in poor quality. With more experi-
move to the observation position and make Kinect ences, more data can be generalized, and the robot
scan point clouds. After scanning, it will wait un- can make better choice of materials. The structure
til the data is analyzed. Then, the construction be- will be more solid and the construction can be taller.
gins. The component controls robot according to
point cloud analysis. After determining the construc- Construction types and processes
tion site through point cloud analysis, the robot will In natural environment, there are many incidents
define a circle to measure a construction range. This happening at any time during construction (For ex-
circle is cut into several segments to show directions ample; structure sway, depression, lack of materi-
and positions for piling the materials. Then, it will als), as a result, we implemented the system with
control gripper to grips the materials to a proper incident response mechanism for adapting complex
position based on the data from the previous com- construction environment.Considering in natural en-
ponent. As the terrain changes and materials keep vironment, beavers use smaller materials as infill to
being piled, this circle curve will be submerged by strengthen structures, the robotic construction be-
newly scanned point clouds. After each layer is sub- havior component provides two selecting modes
merged, the new layer will on top of the old curve and such as stick mode and sawdust mode. When the
become a new placement of materials. In this way, construction structure is stable, the stick mode is on,
sticks are stacked to form a structure step by step (see whereas it switches to sawdust mode and grips saw-
Figure 12). The placements will be piled up along the dust to strengthen the main structure when the con-
Z-axis. This process is similar to 3D printer's Fused De- struction structure is unstable (see Figure 14).
position Modeling (FDM) (see Figure 13).
In this way, the new placement of materials Data recording
will form as the new circle curve exists until the Data recording component will record the data af-
biomimetic construction robot detects that the struc- ter it scans the materials (see Figure 15). Machine
ture is swaying. The robot will offset to narrow the learning component will analyze the correlation be-
circle curve's size to keep the structure from swaying tween each material's geometrical features and its
Figure 15
Data recording
component will
record individual
material properties
and stability.

construction process. Different materials may have Figure 16
different advantages and limitations. As more experi- The learning
ences and data are collected, machine learning com- mechanism is
ponent will help robot select the best material fea- developed based
ture from the database as a model. Whenever new on Spearman Rank
materials are found in the natural environment, the Correlation .
component will choose the material which best fits
the criteria of that model as the material for main
structure.
Figure 17
Machine Learning The system can
System will be executed with machine learning select proper sticks
mechanism and keeping improving stability of as- through evaluation
sembly by choosing appropriate sticks. We predicted of machine learning
that particular properties of sticks (for instance, cur- mechanism.
vature, length and radius) will influence the stability Figure 18
of assembly. Thus the system is designed to find the Proper curvature
best value for assembly by machine learning. The can grab other
method of measuring assembly stability is generat- materials tightly
ing a swept area from predicted place to actual place. and avoid them
The smaller the swept area is, the higher of the as- rolling away when
sembly stability is. Thus, we can analyze correlation CONCLUSION swaying.
between assembly stability and individual material The results in this study indicate that the system
properties. Considering that the correlation between gives priority to materials with 65-centimeter curva-
material properties and assembly stability may be ture radius when selecting materials. The curvature
nonlinear. As a result, we develop the learning mech- plays the most important role in the selecting sec-
anism based on Spearman Rank Correlation in this tion, while length is ranked as second crucial factor
system (see Figure 16). and diameter seems not an essential factor. Proper
At the beginning, the system records several curvature can grab other materials tightly and avoid
data of material properties and assembly stability. them rolling away when swaying(see Figure 18). Be-
Next, the system will define the value of assembly sides, material gaps and infill can be organized well
ability as different scores. Then, it will calculate tem- with proper curvature. Machine learning mechanism
porary appropriate properties for assembly. If the can preclude unsuitable materials and strengthen
current property is closer to temporary appropriate the structure of the dome. As construction expe-
property, the assembly is likely to be more stable. In riences increase, machine learning mechanism has
the algorithm, the temporary appropriate property is made the construction height higher than the first
dynamic may due to data accumulation. Therefore, time(see Figure 19).
machine learning mechanism can find influences of In this research, we focus on how to develop
those material properties (curvature, length and ra- the process for adapting complex environment in-
dius) on assembly stability. The system can select stead of constructing a predefined structure; each
proper sticks through evaluation of machine learning final form in an experiment are unpredicted. Final
mechanism(see Figure 17). form is influenced by context of materials and ter-
rain. In the context, if materials are too short or bent

Figure 19
Performance after
training(left is initial
outcome and right
is the performance
after training).
Figure 20
Any variety can
cause structures to
be different.
Figure 21 and terrain is too rugged, it will cause the structure

Imagination of to be failed, the final form might be low or small,
Using natural as a result, the performance of Biomimetic Robotic
material in full-scale Construction Process resemble "form follows natural
architecture. materials". Any variety can cause result to be differ-
ent(see Figure 20).
On the whole, this research attempts to seek a
solution to two issues in architecture. First one is
environmental impact from architecture and second
is architectural impact from automatic, we combine Robot adapting natural materials to reduce resource
these two to develop a Biomimetic Robotic Construc- waste. Adapting natural materials is widespread in
tion Process as a solution; We review architecture in- natural environment, especially beaver construction
dustry bring material and energy waste from large is a paradigm. Through studying beaver construc-
scale production and then propose a Construction tion behavior, the Construction Process combining

new technologies is possible to be achieved. This Polonorum: Silvarum Colendarum Ratio et Industria
research (Biomimetic Robotic Construction Process) Lignaria, Poland, pp. 63-70
is different from most of digital fabrications because Zari, Maibritt Pedersen 2007 'Biomimetic Approaches
to Architectural Design for Increased Sustainability',
this Construction Process need not predefined form
Regional Sustainable Building Conference 2007, Auck-
for input, the final form is due to materials and terrain land, New Zealand
in environment. [1] https://en.wikipedia.org/wiki/Beaver_dam
We expect biomimetic robotic construction pro- [2] http://www.arct.cam.ac.uk/research/researchgrou
cess can be used for real construction environment ps/natural-materials-and-structures
the after improving reliability, it can be use for full- [3] http://news.discovery.com/animals/zoo-animals/b
eaver-dam-canada-space.htm
scale material (larger stick or driftwood) (see Fig-
[4] https://www.google.com.tw/maps/place/Largest+be
ure 21). And therefore, workers are not necessary aver+Dam+in+the+world/@58.271435,-112.254433,
to be exposed to danger. Instead, Construction 17z/data=!3m1!4b1!4m5!3m4!1s0x53b76aac4c14e1b
Robot can be employed on rugged disaster area for 1:0xf5bd60b121769202!8m2!3d58.271435!4d-112.2
makeshift bridges, embankments and shelters. On 52239
the other hand, in non-disaster area Construction [5] http://designandmake.aaschool.ac.uk/woodchip-ba
rn/
Robot can be employed for river and landscape im-
[6] https://www.linkedin.com/pulse/robotic-softness
provements. After progressing biomimetic construc- -behavioural-fabrication-process-woven-ehsan-
tion robot through newer technology, the updated baharlou
robot may be widely used on permanent buildings. [7] http://www.selfassemblylab.net/RockPrinting.php
By doing so, this natural material-based biomimetic
construction robot will be able to reduce waste of re-
source from construction in the future, this approach
will allow designer to use unprocessed natural mate-
rials more easily and stimulate the design concept of
"Form Follows Natural Materials"
REFERENCES
Blersch, David M and Kangas, Patrick C 2014, 'Signatures
of self-assembly in size distributions of wood mem-
bers in dam structures of Castor canadensis. Global
Ecology and Conservation', ELSEVIER, 2, pp. 204-213
Bock, Thomas (eds) 2008, Robotics and Automation in
Construction, InTech, Rijeka, Coratia
Bunge, Eric and Hoang, Mimi (eds) 2008, Narchitects Eric
Bunge & Mimi Hoang 2000-2008, Damdi
Chen, I-Chin and Hou, June-Hao 2016 'DESIGN WITH
BAMBOO BEND', caadria2016
Erekpitan, Ola-Adisa, Yohanna, C. Sati and Ojonugwa,
Israel I. 2015, 'An Architectural Approach to Solid
Waste Management on Selected Building Construc-
tion Sites in Bauchi Metropholis', International Jour-
nal of Emerging Engineering Research and Technol-
ogy, 3( ), pp. 67-77
Janiszewski, Pawel, Gugolek, Andrzej and Lobanowska,
Anna 2006 'Use of shoreline vegetation by the Eu-
ropean beaver (Castor fiber L.)', Acta Scientiarum

Discrete Assembly and Digital Materials in Architecture
Gilles Retsin1
1
Gilles Retsin Architecture
1
gilles.retsin@gmail.com
The paper will discuss two projects which explore the territory of discrete or
digital material organisations in an architectural context. Taking inspiration from
the field of Digital Materials, this paper presents an approach to architectural
design which is fundamentally "digital" - not just in the process but also in its
physical organisation. The use of discrete and digital materials in architecture is
argued for from both an architectonic point of view, as well as from efficiencies
related to automation of construction. Experiments with robotic assembly are
caught between on the one hand the desire to increase speed, and on the other
hand increased complexity. This paper argues that robotic assembly on the scale
of architecture is only feasible and scalable in the context of digital materials and
discrete computation, which has a limited set of connectivity problems. The two
projects are a first attempt to translate the concept of Digital Materials to the
domain of architecture. The result is an architecture which is digital in its
physical organisation. It demonstrates how differentiated, complex and
heterogeneous spaces can be achieved with just serialised, discrete elements.
Keywords: Discrete Computation, Robotic Assembly, mereology, Digital

Materials, Additive Assembly
FROM CONTINUOUS TO DISCRETE the University of East-London.

This paper will discuss two projects which explore the Theorist and architect Neil Leach argues that "
territory of discrete or digital material organisations while there is clearly a practice of designing that in-
in an architectural context. The projects are practice- volves the use of digital tools, there is no product as
based, and have been developed at Gilles Retsin Ar- such that might be described as digital" (Leach 2015).
chitecture, a London based architecture and design Digital design and fabrication tools merely allow a
practice. They are part of the long term research specific type of design to be realised, but they can as
agenda of the office, and are partially developed in well be used for objects which do not appear "digital"
collaboration with consulting engineers Price & My- or "parametric". This critique ties into a contemporary
ers. The work also interfaces with academic teach- discussion on the post-digital. The use of digital tools
ing into robotic assembly at RC4, a research cluster and techniques has matured to such an extent that
at UCL the Bartlett School of Architecture, as well as their use is more critically questioned. So what does
with a ongoing research agenda into radical discrete- it mean for buildings or material organisations to be
ness which has been developed in M.Arch Unit 6 at digital ? Can material be organised in the same way

as data? Or is there really no such thing as a digital ar- by arguing for a discreteness both in the design
chitecture? Taking inspiration of Neil Gerschenfelds' and in the physical organisation of an object. There
concept of Digital Materials, this paper will present are not only philosophical reasons to explore this
two projects which are fundamentally "digital" - not proposition, but also a series of more pragmatic con-
just in their design process but also in their physical cerns. Continuous systems have fundamental prob-
organisation. In line with Mario Carpo's description lems with transitions of material, require a lot of time
of the Second Digital Turn (Carpo 2014), the projects to compute and fabricate, are less adaptive and not
presented understand computational processes as reversible. Moreover, in architecture, it is hard to cal-
fundamentally discrete. The research proposes to culate the cost of continuous systems, as they are not
align digital with physical materiel organisation. Ev- suited to the increased discrete character of work-
ery bit of digital data is the same bit of data in the flows such as BIM. Discrete construction systems are
physical world. generally dry and prefabricated, which means that
Whereas Carpo only talks about discreteness in they can be constructed faster and with more preci-
the design process, Gerschenfeld's research goes one sion (Knaack 2012). With an increased interest in au-
step further to also argue for discreteness in fabri- tomation of the construction industry, discrete build-
cation (Ward 2010). Essentially, most current fabri- ing blocks can significantly speed up robotic assem-
cation technologies are analogue processes, despite bly process (Gerschenfeld et al. 2015). This paper ar-
the fact that they are computer controlled. These gues for discrete systems as an efficient method to
computer controlled actions mimic human modes design, fabricate, automate and build on the scale of
of production, they are essentially mechanised arti- architecture.
sanal procedures. An industrial robot for example,
can be understood as a computer controlled version DISCRETENESS AND AUTOMATION
of a human arm. Techniques such as CNC-milling are There has been significant research into automation
actually based on centuries old artisanal methods, of the building industry. Moreover, there is a his-
just as additive manufacturing is a technique com- torical relation between architecture and machines
monly found in pottery. These analogue techniques for building. Vitruvius treatise De Architectura , con-
share the property of continuously adding or remov- tains one book solely devoted to machines. The
ing material - they are continuous fabrication tech- mechanisation and industrialisation of architecture
niques. These tools are computer controlled, but not was of course also a dream of the modernists, and
"digital". The analogue character of computer con- has recently again gained momentum as a result of
trolled fabrication tools logically gave rise to a gen- increased access to robotics. The most prolific at-
eration of architects identifying as digital craftsmen. tempt for automation is probably large scale additive
These digital craftsmen or makers are essentially not manufacturing or 3D printing. Key precedents here
"digital" - they are mere sculptors using analogue are Khoshnevis Contour Crafting process and Enrico
techniques in a computer environment instead of in Dini's D-Shape printer. The main interest in automa-
the physical world. Associated with this practice is tion of the building industry lies in the promise of in-
also the culture of the artisan: a culture of intuitive, creased speed, simplicity and reduced human labour
small-scale, aesthetic practice . In contrast to this is on the building site. However, the previously men-
the Albertian paradigm of the architect as a the anti- tioned attempts have had only a limited impact on
craftsman, in communication with the building site the building industry until now. One of the main con-
only through notation and instructions (Carpo 2011). straints with large scale additive manufacturing is un-
The projects presented in this paper extend doubtedly the speed. Robotic assembly processes
Carpo's understanding of the Second Digital Turn, should be in principle quicker, as they make use of

larger elements. Advances in process planning and local positions which in itself provides geometric
BIM software allow prefabricated modular assembly constraints for assembly. No plans or tools are re-
to become an ultra-fast construction method. For quired as the parts geometrically define the assembly
example, the Broad Groups project for a fifty-seven (Cheung 2012). Digital Materials construct objects
story skyscraper was assembled in just nineteen days which are discrete in their physical materiel organi-
in Changsha, China. Moreover, assembly-based pro- sation. These structures are reversible: they can be
cesses have the advantages that the components or re-assembled into other types of structures. The con-
particles for assembly can be composed of multiple cept of digital materials also becomes an interesting
materials. Multi-materiality is a fundamental prob- opportunity for computational methods, as the orga-
lem of additive manufacturing processes. Large- nization of physical parts is the same as the organi-
scale additive manufacturing is almost exclusively zation of the digital data. The part computed digi-
based on a single material. A process which can print tally is also the part assembled physically. The male-
at the same time glass and concrete, is hard to imag- female composition of digital materials establishes a
ine, as both materials have entirely different mate- binary 0-1 relation between elements. In that sense,
rial properties. To produce a complete building, one a lego-brick can be understood as a digital material.
would still have to use traditional means of construc- (Cheung 2012.) Gerschenfeld sees their applications
tion for insulation, windows, finishes, cabling and so in domains as varied as aerospace, building and land-
on. Experiments with robotic assembly, and more scape. Gerschenfeld and his team have developed a
specifically, additive robotic assembly, have been un- number of prototypes of digital materials, most no-
dertaken by Gramazio Kohler at the ETH. As Ger- tably the GIK or General Instruction Kit and the Digital
schenfeld points out, these experiments are caught Flexural Materials (Cheung 2012; Ward 2010).
between on the one hand increasing speed, and in The projects presented in this paper are one of
the other hand increasing complexity ( Gerschenfeld the first explorations of the concept of "digital ma-
et al. 2015). Automation of discrete or digital materi- terials" on an architectural scale. Another precedent
als can result in both increased speed and increased is Jose Sanchez' Polyomino research at USC, which
complexity. However, architects seem to be reluctant is based on a game-like assembly of discrete parts
to adopt the homogeneous and repetitive lattice-like (Sanchez 2014). Skylar Tibbits' self-assembling ma-
structures associated with Digital Materials. terials are also understood as digital materials. Archi-
tectural Digital Materials share the basic properties of
DIGITAL MATERIALS AND DISCRETE FAB- the digital materials developed at MIT: they are dis-
crete, have a limited connection possibility with a bi-
RICATION
nary logic, and relative local positioning.
Although projects such as Gramazio Kohler's roboti-
There are some key-differences with the Dig-
cally assembled brick walls also make use of bricks as
ital Materials proposed by Gerschenfeld. First of
discrete elements (Kohler M et al. 2014), technically
all, architectural digital materials aim for fast assem-
this has to be considered as a continuous project. Ev-
bly, and therefore should be multi-material or multi-
ery brick has a highly varied position, which is inde-
functional. Lattice-like structures like the GIK and
pendent of the geometry of the part. In an analogue
the Digital Flexural Materials would need a significant
or continuous system, a piece of matter has infinite
amount of cladding to create inhabitable spaces.
connection possibilities, whereas a discrete or digital
Secondly, the proposed architectural digital materi-
system only has a limited number (Ward 2010).
als have a much larger scale - meters instead of mil-
Digital materials are a concept pioneered for me-
limeters. The larger scale allows for a much quicker
chanical engineering by Neil Gerschenfeld at MIT.
assembly, and responds better to the amount of res-
A Digital material is a building block with relative

olution or detail required in day to day living envi- whole relationships - a concept referred to as Meron-
ronments or spaces. The use of much smaller units omy or Partonomy (Simmons 2000). Object Oriented
is currently subject of parallel academic research in Programming has a mereological character through
Bartlett RC4 , under guidance of the author. its structure of classes and inheritance of classes. Al-
Lastly, as a result of the larger scale of the digital though mereology is an actual scientific study, devel-
materials, these also heavily impact the design as a oped by the Polish mathematician Lesniewski, it was
whole. Every vector set out by a single element has a initially the work of Levi Bryant which was important
significant impact on the space and structure of the for the development of the projects below. In the
proposed building - in contrast to the smaller scale Democracy of Objects, Bryant describes the concept
elements of non-architectural digital materials. If the of "Strange Mereology", as a relation where every part
digital material is small enough, then heterogeneity is in itself a whole which is not reducible to its parts
can be achieved through their assembly - just as a and where every part is not reducible to the whole
pixel. In this context, Joris Laarmans' Digital Matter (Bryant 2011). This kind of democratic part to whole
Table is a good reference. However, as architectural relationships are typical for Object-Oriented philos-
digital materials are much larger, they need to estab- ophy. It can be argued that Architectural part-to-
lish a precise design agency or bias within the com- whole relations are traditionally a non-strange Mere-
ponent itself. The architect essentially has to embed ology, which means that the parts are never really au-
a series of design decisions in the part, keeping the fi- tonomous from the whole, but are always reducible
nal building in mind. Digital Materials in mechanical to the whole. They have no existence outside of the
engineering are generally homogeneous and repeti- whole. This non-strange mereological character is
tive, and would as such not satisfy the desire for var- typical for top-down design methods, but essentially
iegated and heterogeneous spaces - which has been also for continuous design in general. To establish
one of the driving forces behind the digital. continuity, every part has to be reduced to a deriva-
The design method described in this paper is tion of the whole. A good example of non-strange
based on the assembly of cheap, standardized, dis- mereological structures is parametric surface paneli-
crete elements into indeterminate, heterogeneous sation. In panelisation, every piece is unique, and de-
and differentiated spaces with a high degree of econ- rived from a specific location on a surface. The large
omy. The focus is on a minimum degree of cus- differentiation of pieces requires micromanagement
tomization for a maximum of differentiation, detail, of thousands of components, resulting in excessive
adaptability and economy. The serialisation and labour and energy in assembly. Gramazio Kohler's
metrology embedded in the pieces allows for effi- Programmed Wall, previously mentioned, also falls
cient robotic automation. in this category, as the orientation of every brick is
a result of a continuous parametric function. Every
MEREOLOGY brick is rotated in a specific position to establish the
The emphasis on the part or bit re-emphasizes the overall image of a gradient. This mereological atti-
discussion of part-to-whole relations . This notion, tude is controversial in relation to the premise of dig-
which has traditionally been important for architec- itally intelligent architecture, which argues for com-
ture, disappeared from the discussions surrounding plexity. Often taking inspiration from natural systems
the digital as a result of an interest in holistic, contin- such as swarms, digital architects like to argue for
uous tropes such as topology and surface. Mereology bottom-up systems where the whole emerges from
( from the Greek mereo or "part") is the study of the re- part-part interactions. This type of system necessar-
lations between parts and the wholes they construct. ily requires a democratic interaction of parts and pre-
Mereology specifically deals with hierarchies in part- sumes that every part is actually equal , resulting in

Figure 1 a homogeneous population. From the interaction
Precast Plaster of this homogeneous population, a heterogeneous
Digital Material, whole emerges. The use of the term mereology in
Gilles Retsin the context of the work presented in this paper, is
Architecture. mainly notional and pro-forma, merely stressing the
renewed importance of part-to-whole relationships
and the part itself. At least in the work presented be-
low, it does not directly refer to Lesniewski's philoso-
phy as a scientific framework.
PROJECTS
Two projects, Blokhut (2015) and Diamond Strata
(2016) developed at Gilles Retsin Architecture, start
Figure 2 to explore the potential of digital materials in archi-
Digital model tecture. Both are based on a serialised building block.
indicating serialised The first project, Blokhut , still needs some degree of
versus customised customisation, where Diamond Strata is completely
areas in the Blokhut serialised.
prototype.
Blokhut
The Blokhut (2015) is a study for a villa in the Bel-
gian community of Wetteren. The term "Blokhut" is
a Dutch for Log Cabin - a hut built of whole or split
logs. As a response to budget constraints, and at the
same time a request for a differentiated and highly
articulated space, the project aimed to extensively
use serialised building elements. The Blokhut estab-
lishes a differentiated and adaptive architectural sys-
tem which consists for 90% of serially repeated, dis-
crete, prefabricated concrete elements, and for 10%
of unique, customized elements (Figure 2). To test
Figure 3 the design method, a large-scale prototype of 2 x
Blokhut, prototype 1.5 x 0.3m was developed. It was assembled using
for a villa using more than four thousand pre-cast plaster compo-
discrete assembly. nents, intersecting and joining around a number of
customized, 3D printed zones (Figure 4).

The precast component is designed as an arrow- Figure 4
shaped brick, with a male and female connection Part of the physical
(Figure 1). This arrow-like connection is able to inter- prototype for the
lock two bricks together in a fixed position. This dis- Blokhut, discrete
crete building element can be understood as a digital plaster cast
material: it has a relative local positioning, and the components
design possibilities are defined by the geometry of aggregated on 3D
the element itself. However, the edge of the element printed
has no clear metrology, nor a defined binary con- connections.
nection. This is inconsequential, but allows for some
more continuity in the system which helps absorbing
some of the issues with tolerance. The discrete pieces
in this case are also not reversible, and were joined to-
gether with glue and glass fiber. The tectonic quality Figure 5
and structural behaviour of the Blokhut is rather lim- Combinatorial
ited, and is further improved upon in the next project, logics of the
Diamond Strata. L-Shaped Digital
The Blokhut prototype proves that serial repeti- Materials.
tion of very simple, cheap, prefabricated digital ma-
terials is a feasible and accessible method to achieve
detailed and adaptable forms. The argument shifts
from a system where everything is mass customized,
with a labor intensive assembly process, to a lim-
ited number of intensive, rule-changing customized
zones or glitches and a large number of serially re-
peated, cheap material. The Blokhut can be con-
structed without the need for micro-managing thou-
sands of unique, numbered pieces. Instead, the 3D-
printed components and bricks set out the instruc-
tions for assembly. The assembly is "plan-less" and
"tool-less", as the geometry of the pieces defines the
aggregation.The finished state of the prototype is un-
determined. It can be extended or contracted at any
time. The final geometry is messy, redundant and un-
simplified (Figure 3).

Figure 6 Diamond Strata
Combinatorial Diamond Strata (2016) is a further development of
Aggregation of the Blokhut project, but does not require any cus-
L-Shaped Digital tomised pieces . The project brief was a multi-family
Materials. house with three apartments in a rural area close to
Brussels, Belgium (Figure 7).The project is an aggre-
gation of a single, beam-like timber element. In this
case, a hierarchical digital material was used - a dis-
crete building element with multiple different scales
which can cross-connect. The system consists of a
linear and L-shaped element with a lego-like male-
female connection( Figure 5). The elements can con-
nect through sliding joints on the lateral faces of the
piece. The hierarchical system is similar to OcTree op-
timisation, a procedure used in 3D graphics where
space is partitioned with different scales of voxels de-
pendent on the resolution required. Assembly time
Figure 7 can be reduced as the scale of the parts is adaptable
Diamonds (2016) to the resolution desired.
Gilles Retsin Moving on from the rather simple and con-
Architecture. strained digital materials used in the Blokhut, the
elements in Diamonds Strata can act at the same
time as brick, surface, column and beam, which im-
proves structural performance. Increased capabili-
ties for parts to interlock and support neighbouring
parts are developed, introducing patterns of struc-
ture in the system. Two types of pieces are used: an
L-Shaped element and a straight piece. The L-shaped
element allows the system to grow in all directions
(Figure 6). None of the elements have to be cus-
tomised. The beam-like character of the piece allows
for a more efficient and hierarchical structure than in
the Blokhut. Central columns and beam-like patterns
can be developed first, using the large scale of ele-
ments. Afterwards, smaller scale elements are used
to bridge between the structures aggregated in the
first iteration. The digital materials in this case are
made of Laminated Veneer Lumber (LVL), with steel
plates for the joints. The fibre direction of the LVL
plates is aligned along the long span of the digital
material. The elements in this case effectively oper-
ate as lego-like pieces, which have to perform effi-
ciently both as column, beam or brick. The inside of

Figure 8
Diamonds House
(2016) Gilles Retsin
Architecture.
the elements is hollow and reserved for services and elements' weight has to be limited to 150 kg to be
cabling. feasible for assembly with large industrial robots. In
The resulting building is highly differentiated the case of the proposed projects, assembly could be
and complex, however none of the elements are cus- done with a gantry robot. This gantry robot could as-
tomised (Figure 8). The complexity is based solely on semble chunks or parts of the building off-site, which
part-part relations. The parts are not reducible to just can then be transported to the site on a normal truck.
parts of the whole, and as such also have an existence The inherent disadvantage of this method is the lack
and relevance outside of the proposal for this multi- of scalability: only one gantry robot can be used at
family residence. a time. The previously mentioned hierarchical digital
materials used in the Diamonds house take into ac-
6. TOWARDS DISCRETE ASSEMBLY count a certain degree of scalability, so a gantry struc-
Robotic assembly is only feasible and scalable in the ture could be feasible in this context. Distributed
context of digital materials and discrete computa- robots would become more desirable if the scale of
tion, which has a limited set of connectivity prob- the elements would be smaller, but bring more con-
lems. The high degree of serial repetition in the straints for the design. Next iterations of work should
projects presented make a robotic assembly pro- look into material efficiencies and structural optimi-
cess more feasible. Parts are organised in a grid or sation. Further research with robotics could test as-
voxel-like pattern, the connection between elements sembly sequences and optimise the proposed ge-
is repetitive, and the connection problems them- ometries for robotic handling.
selves are always discrete, neighbour-neighbour or This paper presented a series of arguments for
part-part problems. In contrast to projects based the use of discrete and digital materials in archi-
on continuous variation, these projects can be fab- tecture. This arguments come both from a archi-
ricated through very simple, serialised actions. The tectonic reasoning, as well as from an increased in-

terest in efficiencies associated with automation of
construction. The proposed method was tested on
two case studies for practice-related projects. The
projects presented prove that serial repetition of for-
mally very simple digital materials is able to achieve
complex and articulated designs. The projects pre-
sented articulate a case for a new kind of digital ar-
chitecture - which is physically digital, not just in
the process. This architecture is fundamentally non-
analogue, and non-craft based. It offers a counter-
point to Neil Leachs' argument that there is no such
thing as digital design. It also expands the scope of
the digital beyond its bias against standardisation.
REFERENCES
Bryant, L 2011, The Democracy of Objects, Open Human-
ities Press
Carpo, M 2011, The Alphabet and the Algorithm, MIT
press, Boston
Carpo, M 2014, 'Breaking the Curve', ArtForum, 52(6), pp.
168-173
Cheung, K 2012, Digital Cellular Solids: Reconfigurable
Composite Materials, Ph.D. Thesis, MIT
Gerschenfeld, N, Carney, M, Bennett, B, Calisch, S and
Wilson, S 2015, 'Macrofabrication with Digital Mate-
rials', Architectural Design, 85(5), pp. 122-127
Knaack, U 2012, Prefabricated Systems: Principles of Con-
struction, Birkhauser
Kohler, M, Gramazio, F and Willmann, J 2014, The Robotic
Touch, Park Books, Zurich
Leach, N 2014, 'There Is No Such Thing as Digital Design',
in Gerber, D and Ibanez, M (eds) 2014, Paradigms in
Computing, Evolo, New York
Sanchez, J 2014, 'Post Capitalist Design: Design in the
Age of Access', in Gerber, D and Ibanez, M (eds) 2014,
Paradigms in Computing, Evolo, New York
Simmons, P 2000, Parts: A Study in Ontology, Oxford Uni-
versity Press
Ward, J 2010, Additive Manufacturing of Digital Materials,
Ph.D. Thesis, MIT

The Robotic Production of the GRC Panels in the CorkCrete
Arch Project
A stratified strategy for the fabrication of customized molds
José Pedro Sousa1 , Pedro Filipe Martins2

1,2
Faculty of Architecture, University of Porto + DFL/CEAU/FAUP
1,2
{jsousa|pcarvalho}@arq.up.pt
The CorkCrete Arch was an experimental prototype built in the scope of a

research project concerning the use of robotic fabrication technologies for
non-standard solutions in architecture. It combined 2 materials, cork and GRC
into a self-supporting lightweight building system, designed to explore the
integration of different robotic fabrication technologies in one constructive
solution. This paper is focused in providing a detailed description and analysis of
the robotic fabrication process used in the production of the GRC components.
The presented solution integrated robotic milling and hot-wire cutting
technologies with a stratified mold design strategy that allowed for overcoming
the limitations of each and enabled a time and cost efficient production process.
Keywords: Robotic Hot-Wire Cutting, Digital Fabrication, Glass Fiber

Reinforced Concrete, Computational Design, Corkcrete
INTRODUCTION Objectives
The CorkCrete Arch was developed as a design-based The proposed building system was tested through
research activity concerned with using robotic fab- the design and construction of an arch structure. Fea-
rication technologies in the production of a novel turing double-curved and textured cork panels com-
building system. By exploring the combination of bined with ruled-surface GRC elements, its geometry
two different materials - cork and Glass-fiber Rein- challenged the use of robotic fabrication technolo-
forced Concrete (GRC) - the goal was to merge the gies for its production. Since it was not a single ma-
sustainable and insulation properties of the first with terial installation, like in many actual robotic exper-
the structural efficiency of the second. The result is a iments, the process had to coordinate the different
lightweight and performative material system suited physical tolerances resulting from employing differ-
for customized prefabrication and easy on-site instal- ent materials and different fabrication processes, tak-
lation. ing place in different locations (i.e. laboratory, fac-
tory and on-site). Working in collaboration with in-
dustrial partners, from the cork and GRC fields, this
experience envisioned and tested a fabrication strat-

egy with both materials that revealed to be very pro- which the architects developed modular elements
ductive and efficient for both areas. While an over- for a window frame with curved profiles. However,
all design and tectonic description of this project can since they can assume any shape depending on the
be found in a previous text by the authors (Sousa et mold, the use of GRC has become an increasingly
al. 2016), the present paper is focused in providing a popular solution in recent years to address the com-
detailed description and analysis of the robotic fab- plex and variable challenges in contemporary archi-
rication process used in the production of the GRC tecture. Built in 2009, the Roca London Gallery de-
components. In order to address the materialization signed by Zaha Hadid Architects featured an inte-
problems, put forward by the geometry of the panels, rior continuous curved space, built with 272 differ-
this process explored a novel stratified mold concept, ent self-supporting GRC sandwich panels. Despite
contributing to the existing state-of-the art in robotic the formal versatility of the material, the produc-
hot-wire cutting of formwork. For its understanding tion of different molds defines the main econom-
and assessment, this paper describes the traditional ical constraint of the system. For instance, in the
GRC production process and recent technological in- Heydar Aliyev Center in Baku, also designed by the
novations before illustrating the CorkCrete arch GRC same office, the large-scale application of GRC panels
components and the production strategy employed. required a geometric rationalization process to de-
fine and organize the panels into planar, single and
THE GRC IN ARCHITECTURE double-curved ones. Despite the exceptional eco-
GRC is a composite material developed in the late nomical means involved in this project this simplifi-
1960s that combines the resistance and binding cation process had to be considered for its construc-
properties of a cement matrix, with the tension resis- tion (Bekiroglu 2010).
tance properties of glass fibers to achieve thin, shell-
like construction elements. The traditional fabrica- Technological Innovation
tion processes of GRC panels can be subdivided in Due to its singular properties and its potential
two: premix or sprayed. In the first case, the mate- to match contemporary design interests, many re-
rials which compose the GRC are mixed in advance searchers have sought opportunities for innovation
and cast into a mold, press molded, extruded or slip- in GRC production over the last 2 decades. In or-
formed. In the second, and most common scenario, der to overcome geometric, manual production and
the cement paste and the glass fibre are simultane- economical limitations, this continuous investigation
ously sprayed using a two headed nozzle into a mold has been mainly focused in the automation of two
(Brookes and Meijs 2008). This procedure is a com- moments: the production of molds and the spraying
plex iterative manual process of depositing and com- procedure.
pacting the mix of cement paste and fibres, and its Regarding the first, the use of Computer Numer-
success depends on the skill and judgment of the ically Controlled (CNC) milling processes has been
workers. A finishing procedure is used, depending on adopted to fabricate complex shape molds in wood
the necessity of different surface qualities or effects. or foam parts directly from digital models. Although
these subtractive processes are widely available, they
Architectural applications are time-consuming, expensive and produce large
Given that GRC components are thinner and lighter amounts of material waste in the form of dust, which
than precast concrete ones, their main use in archi- is critical for reuse purposes. To minimize these is-
tecture is in cladding applications. One of the first sues, other digital fabrication strategies have been
relevant works with this material can be found in the developed, such as Robotic Hot-Wire Cutting (RHWC)
Credit Lyonnais building in London built in 1977, in to quickly fabricate molds in EPS for ruled surface-

based designs with minimal material waste (Feringa The geometric complexity of these design elements
and Søndergaard 2012). was heightened by the necessity of fabricating the
Regarding the automation of the spraying mo- corresponding negative molds for GRC spraying. The
ment, the work in the 90's of Carlos Balaguer and José main fabrication challenge was thus the production
Manuel Pastor Garcia with the Spanish company Dra- of a large-size concave ruled surface mold with vari-
gados y Construcciones (Balaguer et al. 1993) intro- able lateral walls. This kind of geometry requires a
duced the use of robotic spraying processes. Despite customized mold that can not be achieved with a
the advantages that were verified (Peñin et al. 1998), single-run RHWC process, over a single raw material.
its adoption by the industry has not happened yet. In
Figure 1 this summary it is interesting to see how the robotic
The Corkcrete Arch arm seems to play a key role in driving the develop-
3D model with ment of innovative solutions in both moments.
exploded GRC and
cork components.
A STRATIFIED STRATEGY FOR ROBOTIC
FABRICATION OF GRC COMPONENTS
The CorkCrete Arch was specifically designed for pre-
fabrication. With 280cm wide and 260cm height, it
is a composite system built from the assembly of 3
structural GRC elements and 18 cork panels. The ge-
ometry of the CorkCrete arch was driven by two fac-
tors: the catenary curve and the robotic fabrication
technologies. On one hand, the catenary curve was
used to introduce a clear structural principle in the
arch. On the other, robotic fabrication technologies
were used as a design driver to shape the material
surfaces, which for the GRC panels signified an explo-
ration of ruled geometries and their materialization
through robotic hot-wire cutting.
The specific geometry of the 3 GRC panels posed
several challenges. For each one, its interior face was
designed as the intersection of two ruled surfaces
producing an emergent and progressive crease in the
central axis. The exterior face was delimited by two
planar surfaces to connect with the flat bottom sur-
face of the cork panels. The lateral faces of the GRC
panels performs as the structural reinforcement of
the thin panels, and was also designed as a continu-
ous ruled surface, which increases in width from the
base to the apex of the arch (Figure 1).

The Stratified Concept binations of 100x50x50cm high-density EPS blocks as
To overcome the challenges mentioned, a strategy stock material, with the expected seams considered
for a stratified process to produce the mold was in the design of the arch elements. The fabrication
developed by taking advantage of the benefits of routines were composed of four cutting operations
robotic hot-wire cutting and addressed its limitations performed sequentially: (a) hot-wire cut to define the
to produce the necessary mold geometry. RHWC is interior surface of the arch and removal of the result-
generally characterized by its limitation to the pro- ing top half; (b) milling of the ruled contour of the
duction of ruled surfaces and its ability to do so with panel with a 20mm diameter bit, in a single pass; (c)
considerably low fabrication times when compared hot-wire cut, defining the planar surfaces for the ex-
to traditional subtractive technologies (i.e. milling) terior face of the GRC elements; (d) hot-wire cut of the
(Feringa and Søndergaard 2012). Considering the EPS core resulting from the previous operations to fit
fabrication of molds in EPS, one of its main advan- the interior of the panels (Figure 3). Independently
tages is the preservation of excess material when cut- of surface complexity, each routine was completed
ting a surface. This feature enables the production of in approximately 10 minutes for all 4 cuts, a relatively
successive cuts, subdividing a stock block into several fast process when compared to the necessary hours
different layers. Applying this notion to formwork de- of CNC milling to achieve similar finishing quality.
sign results in the concept of the "stratified mold". A
subdivision of a mold geometry into several layers, Figure 2
cut from a single EPS block in a short time span and The stratified
assembled into a final mold that couldn't otherwise concept applied to
be produced with RHWC technology. the Corkcrete Arch
Following this concept, the molds for the Corkcrete molds: bottom
Arch GRC panels were subdivided into 3 layers to surface layer;
be sequentially cut: the bottom surface; the con- contour layer;
tour layer, creating the lateral structural folds and the sprayed GRC panel;
overall boundary geometry of the panels; the top interior core; outer
surface, defining the exterior flat surface of the pan- layer.
els (Figure 2). To overcome size and geometric con-
straints (the size of the available hot-wire and the
concave, intersecting ruled surfaces), each mold was
divided in two halves through their central axis for
fabrication, resulting in six halves for the three GRC
elements.
Robotic Fabrication of the Molds

The RHWC system at the DFL laboratory consists
of a stationary Kuka KR 120 R2700 industrial robot,
mounted with a 100 cm wide hot-wire bow, featur-
ing a 0,25mm thick wire. Simulation and toolpath
generation was done in the Grasshopper plugin for
Rhinoceros, with the Kuka|prc software, developed
by the Association for Robots in Architecture. Us-
ing this setup, the molds were produced from com-

Figure 3
Mold production
process featuring
the four sections for
the center GRC
element: bottom
curved layer,
contour milling, top
flat layer and
interior core
cutting. Completed
mold pieces for the
bottom arch
elements,
illustrating the
mold subdivision.

After the fabrication process, the first and second Engil, S.A.. Minimal work was done on the mold
mold layers were assembled and glued together, surfaces and a simple coating of demolding oil was
along with their corresponding halves. Because the applied in order to minimize the overall production
GRC spraying process requires an open mold, the time and costs. Before the spraying procedures took
third layer was not necessary, being only used for place, a set of metal anchor elements was introduced
stacking and transportation purposes. Nevertheless, in the molds to be embedded in the GRC matrix for
if necessary, the process could have been further op- transportation and as fasteners between the 3 arch
timized to nest more molds in the remaining EPS ma- elements. The spraying process used a white GRC mix
terial. At this stage, the robotic fabrication phase to make a clear distinction between the two materi-
proved to be a straightforward process, clearly adapt- als and to accentuate the geometric features of the
able for larger production scales, maintaining eco- arch. It consisted of an initial layer of GRC with an av-
nomic viability. erage thickness of 15mm, followed by manual com-
paction. After this step, the EPS cores were placed in-
The GRC Production side the mold and a finishing layer was applied. Em-
The production of the GRC panels took place at a pre- bedding the excess EPS cores in the panels added to
cast facility of the projects' industrial partner, Mota- the structural stability of the system and created a
Figure 4
GRC production
process: spraying
and EPS core
positioning; center
arch component
after demolding.

suitable support surface for the cork elements. After ceptable smooth finish. This is due to the poros-
curing for 24 hours, the panels were easily demolded ity of the EPS material which enables the transfer-
and shipped inside the used molds. ence of its cellular texture the finished surfaces. Also,
The robotically produced molds were effortlessly faint marks of the hot-wire could be found on the fi-
used in the industrial setting of the precast facility by nal surface which were traced back to vibrations of
the workers. They proved to be suitable for regular the wire during the cutting process. Although unin-
spraying equipment and operating procedures with- tentional and generally undesired, these textural ele-
out adaptations or special knowledge required, vali- ments added an interesting aesthetic layer as finger-
dating its usage in this environment. (Figure 4) prints of the fabrication process. After the production
of the GRC components and the corresponding cork
Results and Installation panels, the Corkcrete Arch was manually assembled
As expected, the surface quality of the GRC panels several times in different settings, without mechani-
was not completely polished as is the case with tra- cal aid (Figure 5). This demonstrated the lightweight
ditional flat steel molds. Nevertheless it had an ac- properties of the system and suggested the feasibil-
Figure 5
Manual assembly
process and
finished arch
structure. A
continuous
geometrical
transition between
ruled surfaces and
the well defined
central crease
reflecting the
precision of the
robotic fabrication
process.

ity of expanding it for larger applications. The assem- REFERENCES
bly stage also confirmed the accuracy of the robotic Balaguer, C, Rodriguez, FJ, Pastor, JM and Peñin, LF 1993
fabrication process. Small geometric deviations from 'Robotized System of GRC Panels for Construction
the modeled geometry were found to be well inside Industry', Proceedings of the 10th International Sym-
posium on Automation and Robotics in Construction
construction tolerances and were absorbed in the
(ISARC 93),, Houston
construction joints of the arch system. Bekiroglu, S 2010 'Assembling Freeform Buildings in Pre-
cast Concrete: Heydar Aliyev Cultural Center by
CONCLUSION Zaha Hadid Architects', Reader Symposium, TU Delft
Brookes, AJ and Meijs, M 2008, Cladding of Buildings,
The research presented in this paper addressed
Abingdon: Taylor & Francis
the problem of producing custom molds for non- Feringa, J and Søndergaard, A 2012 'Design and Fabrica-
standard GRC components in a composite building tion of Topologically Optimized Structures', Proceed-
system. To achieve this goal, the authors devised a ings of the 30th International Conference on Education
stratified strategy that combined the efficiency of hot and Research in Computer Aided Architectural Design
wire cutting and milling in one fabrication process. in Europe, Prague
Peñin, LF, Balaguer, C, Pastor, JM, Rodriguez, FJ, Barrien-
The produced results revealed the flexibility and
tos, A and Aracil, R 1998, 'Robotized Spraying of Pre-
versatility of the use of robots in relation to tradi- fabricated Panels', Robotics & Automation Magazine,
tional CNC systems, by being able to use different IEEE, 5(3)
tools and optimize their inherent abilities. The strati- Sousa, JP, Martins, P and Varela, P 2016 'The CorkCrete
fied mold strategy proved to be a time, material and Arch Project: The Digital Design and Robotic Fabri-
cost-efficient solution for the production of complex cation of a Novel Building System Made out of Cork
and Glass-Fiber Reinforced Concrete', Proceedings of
molds when compared to both traditional manual
the 21st CAADRIA – Computer Aided Architectural De-
methods and contemporary milled formwork. The sign Research in Asia Conference, University of Mel-
validation of this solution in an industrial setting pro- bourne
moted the interest of the precast industry in contin-
uing the research into this topic. In this regard, fur-
ther developments will seek other applications of the
stratified mold strategy and address 2 key issues: im-
proving the surface quality of the GRC without incur-
ring in a substantial increase in production time and
optimizing stock material usage.
ACKNOWLEDGMENTS
The authors would like to thank our industrial
partner, Mota-Engil in the production of the GRC
components for the Corkcrete Arch. The work pre-
sented was co-financed by the European Regional
Development Fund (ERDF) through the COMPETE
2020 - Operational Programme for Competitiveness
and Internationalization (POCI) and national funds
by the FCT under the POCI-01-0145-FEDER-007744
project, the previous PTDC/ATP- AQI/5124/2012 re-
search project, and the SFRH/BD/79227/2011 PhD
scholarship.

FABRICATION | Robotics: Generative
Design
Calculated Matter
Algorithmic Form-Finding and Robotic Mold-Making
Elif Erdine1 , Alexandros Kallegias2

1,2
Architectural Association (AA) School of Architecture
1,2
{elif.erdine|alexandros.kallegias}@aaschool.ac.uk
The paper addresses a specific method for the production of custom-made,

differentiated moulds for the realization of a complex, doubly-curved wall
element during an international three-week architectural programme,
Architectural Association (AA) Summer DLAB. The research objectives focus on
linking geometry, structure, and robotic fabrication within the material agency of
concrete. Computational workflow comprises the integration of structural
analysis tools and real-time form-finding methods in order to inform global
geometry and structural performance simultaneously. The ability to exchange
information between various simulation, modelling, analysis, and fabrication
software in a seamless fashion is one of the key areas where the creation of
complex form meets with the simplicity of exchanging information throughout
various platforms. The paper links the notions of complexity and simplicity
throughout the design and fabrication processes. The aim to create a complex
geometrical configuration within the simplicity of a single material system,
concrete, presents itself as an opportunity for further discussion and development.
Keywords: robotic fabrication, custom form-work, generative design, structural

analysis, concrete
INTRODUCTION multi-axis freedom of an industrial robot arm, its

The digital era in architecture has witnessed the speed, precision, and low tolerances (Menges 2013).
production of a vast array of geometrical assem- This development in turn has fuelled the revival of
blies through computational form-finding methods complexity found in volumetric assemblies, moving
in previous decades. With the developments in digi- away from previous standardized / sheet-material
tal fabrication, the production and assembly of com- component fabrication (McGee et al. 2013). The
plex forms has been compromised by the constraints capacity to design and produce volumetric build-
of selected fabrication techniques. In recent years, ing components through robotic carving strategies
robotic fabrication processes implemented in archi- presents itself as a novel approach where custom-
tecture have begun to incorporate digital and phys- made moulds can be produced with speed and pre-
ical paradigms in an unparalleled way due to the cision, allowing for the creation of complex forms
FABRICATION | Robotics: Generative Design - Volume 1 - eCAADe 34 | 163

which would not have been possible with standard- materiality. Recent developments in robotic fabrica-
ized building materials and processes (Brell-Cokcan tion techniques offer designers with the capacity to
and Braumann 2010). fabricate complex geometrical configurations thanks
This paper presents a specific method for the to their multi-axis freedom. As such, the major ob-
production of custom-made, differentiated moulds jective of this case study has been directed towards
for the realization of a complex, doubly-curved wall the coupling of doubly-curved complex geometrical
element during an international three-week archi- assemblies informed by structural analysis and their
tectural programme, Architectural Association (AA) realization through robotic milling processes.
Summer DLAB, which took place during 27 July - 14 Initial form-finding experiments have focused on
August 2015 in AA's London home and Hooke Park the manipulation of a vertical planar geometry to-
premises. The method engages an industrial robotic wards the creation of varying degrees of curvature
arm mounted with a milling head, the mould material through a range of computational techniques. Key
is medium-density EPS foam, and the final assembly parameters in this phase have comprised the dis-
is produced with fast-setting concrete. The research tribution of openings and areas of local differenti-
objectives focus on linking geometry, structure, and ated curvature values that stem from a global mor-
robotic fabrication within the material agency of con- phology. After preliminary tests conducted in differ-
crete. In this respect, it aims to link the notions of ent algorithmic platforms, the open-source program-
complexity and simplicity throughout the design and ming environment Processing (Shiffman 2012) has
fabrication processes. While complexity is generated been chosen for further exploration for its resilience
throughout computational form-finding techniques, and accessibility to create custom-made tools. In
simplicity lies in how architectural information re- this setup, an agent-based model approach has been
lating to geometry, analysis, and fabrication can be adopted. These computational algorithms simulate
seamlessly transferred between various simulation, the local interactions of agents in order to evaluate
modelling, analysis, and fabrication platforms. More- complex behavioural patterns, which can then be
over, the aim to create a complex geometrical con- further developed to manipulate geometrical vari-
figuration within the simplicity of a single material ation according to a set of predefined criteria. In
system, concrete, presents itself as an opportunity for the early tests of the custom tool, an agent-based
further discussion and development. system has manipulated a vertical planar geometry
constructed as a mesh, imported from the three-
COMPUTATIONAL WORKFLOW dimensional modelling software McNeel Rhinoceros.
A research methodology has been structured around These digital experiments have been valuable in set-
a set of experiments in line with the research objec- ting up constraints for the agent system in relation
tives. The design brief is to propose, fabricate, and to the Z-axis coordinate of the agent, the distance
assemble a one-to-one scale architectural wall ele- between the agent and the planar mesh, and the
ment from concrete in a forest located in the south amount of displacement the mesh face can achieve
of United Kingdom within a limited time frame, three in relation to its normal vector.
weeks. The employment of complex formwork for The goal of the second set of computational ex-
concrete structures has the potential to yield mor- periments has been the integration of structural anal-
phologically interesting and materially efficient as- ysis tools from the outset of design explorations in or-
semblies. In the initial stages of design development, der to inform global geometry and its structural per-
real-time generative form-finding methods have set formance. In this regard, the planar mesh in the pre-
the correlations between the computational process vious experiments has been replaced with a global
of design with the physical world of fabrication and geometry which is defined by a straight section at its
164 | eCAADe 34 - FABRICATION | Robotics: Generative Design - Volume 1

top and an "S-curve" section at its bottom, modelled ture generation. The agent system influences local
in McNeel Rhinoceros, in order to assist in structural mesh curvature throughout the overall form along
performance. This geometry has then been evalu- the force flow lines following flocking and proximity
ated under its self-weight via FEA analyses with the rules. The magnitudes of vectors manipulating the
Grasshopper add-on Karamba as a shell structure [1]. form increase as the agents flock towards the ground
For the analyses, material properties of high-strength plane resulting in greater mesh deformation in the
concrete (C90/105) have been selected. After various bottom areas. As a result, the final mesh geome-
tests with shell depth, a value of 5 cm. has been se- try is characterized by more local curvature and ma-
lected in regard to keeping displacement values at an terial accumulation in the bottom areas compared
optimal level. The output mesh model of Karamba to its relatively straight and thin cross-section at the
has then been connected to the 'ForceFlow' com- top. The differentiation in thickness has been a focus
ponent, a native component of Karamba, which en- for the corresponding physical experiments in order
ables the visualization of force flow lines in a shell in to examine the variation of cross-sectional depth in
the global direction provided. The generation of the conjunction with the proposed fabrication process.
force flow lines has been design-oriented, with the With the resulting tapering effect, the final outcome
aim of achieving wall openings which do not inter- of the simulation has more structural load-bearing
fere with the transfer of loads throughout the global capacity (Figure 2).
geometry. As such, the FEA stage has been con-
cluded with the creation of a 5 cm. deep shell split by ROBOTIC MILLING AND ASSEMBLY PRO-
openings following the direction of force flow within
CESS
the wall (Figure 1).
As it has been described above, the second goal
The next stage in the computational design pro-
of the simulation process has been to test robotic
cess entails the generation of doubly-curved geome-
milling processes with the purpose of experiment-
tries following the initial shell model as an input. The
ing with complex curvatures for form-work building.
purpose of this investigation is two-fold, pertaining
In this respect, the choice of form-work material and
to structural performance and exploration of robotic
robotic milling time have served as major inputs for
milling techniques. The tool in Processing is devel-
the fine-tuning of applied forces in the Processing
oped further as a combination of agent-based simu-
simulation. Computational experiments have been
lation and mesh relaxation techniques with the pur-
simultaneously coupled with scaled fabrication ex-
pose of creating a doubly-curved geometrical aggre-
periments before the final design output. The robotic
gation that increases in density towards the bottom
fabrication processes in the design allow for moving
section of the wall. After the generation of open-
away from a direct design-to-production approach,
ings, the mesh with its corresponding force flow lines
whereby the final outcome is predefined and fabri-
has been imported in Processing for complex curva-
Figure 1
FEA conducted on
preliminary design
in Karamba add-on
for Grasshopper,
followed by the
extraction of force
flow lines and the
generation of
openings on the
wall.

cation solely offers a "means to an end". The tool- for robotic simulation followed by the milling pro-
ing path serves as a direct visual connection between cess. The output mesh generated in Processing is
the global geometry and local surface manipulation; exported as a text file into McNeel Rhinoceros, be-
therefore, the robotic end effector plays a crucial role coming the negative geometry for the preparation of
as a design tool in the generation of localized surface form-work. In Rhinoceros, each form-work is created
textures on the global configuration. Several path by taking into consideration material dimensions and
milling tools have been tested on scaled physical ex- tolerances. The dimension of each EPS block is 200
periments regarding linear vs. elliptic paths, radius of cm in length, 125 cm in height, and 50 cm in depth,
tool and fabrication time. Due to the time limitation resulting with the employment of a total of 8 EPS
pertaining to the fabrication and assembly of the final blocks. As the fabrication and assembly processes
prototype, a linear tool path has been chosen to cre- need to have high precision for desired outcomes,
ate local surface textures with a robotic milling tool the placement of steel bars connecting the form-
defined by a ball nose of 25 mm. diameter and 250 work elements on two sides has been calculated in
mm. length. Rhinoceros as well. The location of the steel bars fol-
The material for the form-work has been selected lows the distribution of the openings of the final wall
as medium density (30g/l) fire retardant grade ex- structure (Figure 3).
panded polystyrene (EPS) blocks, as EPS offers a suit- The end-effector for the robot serves as a design
able compromise between milling time and strength means that aids in the generation of surface textures
for form-work construction. Several iterations inves- in the EPS boards. After the completion of the milling
tigating the increase of surface area in relation to process that lasted 30 hours in total, the areas of con-
milling time have been generated in Processing. The tact between the scaffolding and concrete have been
final output model demonstrated in Figure 2 is the re- treated with a mixture of silicone and mold releasing
sult of this iterative process, providing a conclusion agent in order to assist with the de-molding process
that the total amount of milling time would approxi- (Figure 4). Accordingly, the EPS foam boards have
mately be 30 hours. been connected and secured with steel bars and ply-
The next stage of design development has in- wood panels to enhance their stability. It is impor-
volved the creation of the necessary form-work files tant to note that the structure is made of a special
Figure 2
Instances of
Processing
simulation
demonstrating the
applied forces to
the input mesh.

Figure 3
The creation of
form-work files for
robotic milling.
concrete mix with fiberglass additives which has en- concrete has been selected as the most suitable ma-
abled it to be cast, dried and held strongly in place terial for casting. The final concrete mixture is fast-
in a period of several hours without being limited by setting with fiberglass additives thus excluding con-
the constraints of applying conventional reinforcing ventional reinforcing systems such as rebar through-
systems such as rebar. The only location where rebar out the wall element, except along the bottom parts
has been used in the final fabrication has been along where it is linked to its foundation. The final assem-
the foundation of the wall, bearing a depth of 30 cm. bly stages have involved the casting of fast-setting
concrete in the EPS form-work, followed by the cur-
Figure 4 ing time of approximately 12 hours. With dimensions
A section of EPS of 2.2 meters height, 4 meters width, and a varying
form-work after it depth of 30 - 250 mm, the final wall assembly has
has been treated been developed, fabricated, and assembled in 7 days
with silicone and (Figure 5).
release agent.
CONCLUSIONS
While this research has been applied for the realiza-
tion of a wall in the context of the international pro-
gramme, it has future potentials applied on more
complex architectural elements. One of the limita-
Figure 5
tions during the fabrication stage has pertained to
Final wall structure,
the time constraints to produce the moulds; there-
overall perspective.
fore, the tooling path had to be comprised in order
to maintain realistic production times. The modular-
ity of the form-work was a challenge for the concrete
pouring of the entire structure as the final placement
of the form-work pieces caused a thinner space in-
between due to the milling pathways used. However,
due to the precision of the robotic milling the surface
continuity was achieved via specific markings on the
blocks. Due to the increased temperature of the fast-
setting concrete, the EPS blocks were fused with the
As the final structure would be placed outdoors, it thin partition rendering of the latex coating, making
has been crucial that it resists the cold and humid it highly challenging for the demoulding. Together
weather conditions of its environment; therefore, with the double-curved profile of the wall as well as

the undercuts, the removal of the EPS blocks required We would like to thank our tutoring team and stu-
delicate but powerful approach. An ideal formwork dents for their great efforts. Summer DLAB 2015:
setup would comprise EPS blocks with thicker coat- Tutors - Elif Erdine, Alexandros Kallegias, Daghan
ing, larger openings on the wall surface, and wider Cam, Angel Lara Moreira, Zeynep Aksoz. Students:
profiles at the lower parts of the partition in order to Lorenzo Pellegrini, Jackson Lindsay, Xiaoxiang Li,
ensure the consistency in the concrete pouring. The Matthew Eng, Angelo Figliola, Nicolo Cernigliaro,
computational simulations on structural analysis that Jos van Roosmalen, Areti Sanoudou-Dramalioti, Ju-
showed the distribution of forces proved to be true lian Heinen, Mathias Paul Reisigl, Stavroula Kakaletsi,
in terms of the even distribution of displacements Pallavi Gupta, Vikas Sharma, Emine Ece Emanetoglu,
along the height. As the final prototype resulted in Leo Claudius Bieling, Florian Kaiser, Jin Li. Robotic
being critically thinner than initially computed, the Setup: Pradeep Devadass.
fact that it can hold its own weight acts as a proof of
the successful implementation of physics behaviour, Figure 6
the material properties of concrete (Figure 6). Final wall structure,
The integration of robotic techniques is advan- detail displaying
tageous due to the capacity to explore volumet- surface textures.
ric spatial formations as well as to incorporate real-
time feedback for future explorations. The design
and fabrication processes have demonstrated the
strong independence between the digital and phys-
ical paradigms. The computational simulations have
taken real-world constraints into account with the in-
tegration of structural analysis. Digital simulation on
structural analysis proved to be accurate compared
to the final outcome. Throughout the design, fabri-
cation, and assembly processes, the interactive asso-
ciations between different simulation software have
been a key driver in recognizing the ways of inte- REFERENCES
grating architectural criteria with the structural per- Brell-Cokcan, S. and Braumann, J. 2010 'A New Paramet-
ric Design Tool for Robot Milling', LIFE in:formation,
formance of the pavilion. Furthermore, the compar-
On Responsive Information and Variations in Architec-
ison between the digital simulation of the architecture: Proceedings of the 30th Annual Conference of the
tural output and the final output, the pavilion itself, Association for Computer Aided Design in Architecture
provides useful information to be considered and (ACADIA), pp. 357-363
embedded in the future digital simulations. Overall, McGee, A., Feringa, J. and Søndergaard, A. 2013 'Pro-
the research aims to illustrate the architectural possi- cesses for an Architecture of Volume: Robotic Hot
Wire Cutting', Rob|Arch 2012: Robotic Fabrication in
bilities of using concrete in a non-conventional way
Architecture, Art and Design, pp. 62-71
with limited resources and period of time by creating Menges, A. 2013 'Morphospaces of Robotic Fabrication',
strong associations between computational design Rob|Arch 2012: Robotic Fabrication in Architecture, Art
methodologies and digital fabrication processes. and Design, pp. 28-47
Shiffman, D. 2012, The Nature of Code: Simulating Natural
Systems with Processing, The Nature of Code
ACKNOWLEDGEMENTS [1] http://www.karamba3d.com/.
The work presented is part of the research under-
taken at AA Summer DLAB Visiting School 2015.

Optimization Process Towards Robotic Manufacturing in
Actual Scale
The Implementation of Genetic Algorithms in the Robotic Construction of
Modular Formwork Systems
Odysseas Kontovourkis1 , Panagiota Konatzii2

1,2
Department of Architecture, University of Cyprus
1,2
{kontovourkis.odysseas|konatzii.panagiota}@ucy.ac.cy
The application of optimization processes in architectural design has gained

significant attention among architects and recently has become a driving force
towards more robust, reliable as well as flexible design investigations. Such
application, require handling of multiple parameters, aiming at finding the range
of possible solutions in morphological or topological problems of optimization,
mostly during the design decision-making process and under the influence of
functional, environmental, structural, or other design criteria. This ongoing
research investigation puts forward the hypothesis that optimization processes
might be equally applied during the construction decision-making process where
architectural systems are examined in terms of their ability to be statically
efficient and easily manufactured through the use of robotic machines. This is
important to exist within a bidirectional platform of communication where the
design decision-making will inform decision taken during pre-construction stage
and vise versa. In order to test our hypothesis, two case studies are developed
that implements genetic algorithms to examine the geometric and static behavior
as well as the construction ability of proposed flexible three-dimensional modular
formworks and overall systems for concrete casting, aiming to be robotically
manufactured in actual scale.
Keywords: Optimization process, genetic algorithms, robotic manufacturing,

modular formwork system.
INTRODUCTION topological issues based on a number of criteria.

The implementation of genetic algorithms in archi- Specifically, the process attempts to increase the
tectural design as a multi-objective optimization that number of possible solutions using a feedback loop
occurs during the design decision-making process logic that cyclically iterates between modification
aims to find suitable solutions in morphological or and evaluation of results through the use of genetic

mechanisms of crossover and mutation of a popula- solve such issues and in parallel might offer solutions
tion of solutions based on a number of generations. in cases where time and precision of executed tasks
In this process, the evaluation criteria are represented or heavy-duty works are important to be taken into
as objectives and are used to evaluate the popula- account enabling accuracy and hence error elimina-
tion of solutions. In order to produce optimal solution during the process.
tions, at least one or more evaluation criteria are im- Nowadays, robots can be involved and taken
portant to be taken into consideration. These can into account throughout the whole manufacturing
be values within a minimum or maximum range de- workflow, from design to construction. Within this
rived throughout the digital simulation process. In frame, the standardization of dimensions and con-
the case of multi-objective optimization, the evalua- struction details of building elements during the de-
tion criteria and the number of possible solutions are sign stage, which aim to automate the process of
increased dramatically and as a result, mechanisms manufacturing and the assembly of parts are impor-
such as Pareto front are used to search for solutions tant to be investigated. This can be found in cases
in a multidimensional space. where the systems under construction are character-
Among others, these might involve geometrical ized as adaptable or flexible, aiming to achieve opti-
and volumetric criteria that are influenced by sound mized design results based on various criteria. In ac-
(Foged et al. 2012), thermal performance (Poblete cordance with robotic manufacturing, the optimiza-
and Ignacio 2011), etc. In other examples, with the tion processes and especially the use of genetic algo-
assistance of genetic algorithms, a large number of rithms can be applied in research works where stan-
data related to the natural lighting performance in dardization of system's structural members and in-
relation to the sufficient number of structural mem- vestigation of static behavior are examined, leading
bers and the minimum cost during manufacturing towards physical prototyping of results that can be
of a static system are examined (Richardson et al. implemented via robotic mechanisms. In a number
2013). Similarly, the implementation of genetic al- of cases (Søndergaard et al. 2013), the investigation
gorithms in construction decision-making process is is concentrated towards the optimization and fabri-
found, mostly isolated from the decisions taken dur- cation of space-frame structures through the digital
ing the conceptual stage where the morphological design process. Then, the robotic arm is used during
development occurs. In contrasts, examples can only the production and assembly of structural elements
be found in cases where genetic algorithms are used (Søndergaard et al. 2013; Kurilla and Svoboda 2013),
to calculate the formation and adjustment of form- while in other cases a post-optimization procedure of
works in already existing designs through the ar- modules rearrangement is applied to check the fea-
rangement of slabs and columns. In this case, aim is sibility of structure (Kurilla and Svoboda 2013).
to calculate the size and position of formworks in or- The formwork development for concrete cast-
der to minimize cost during construction (Lee et al. ing is a time consuming procedure, which requires
2014) or to optimize shape and layout (Pronk et al. the assembly but also the disassembly of multiple
2013). components after the concrete is cured. Contem-
On the other hand, the construction of any build- porary and advanced methods are developed in or-
ing part in actual scale has been largely controlled der to minimize the time and automate the manu-
by the ability of workers on site who are responsi- facturing steps of concrete structures. More specif-
ble to implement design ideas mainly through deci- ically, design examples that require the assistance
sions taken during the construction stage and after of robotic arms to implement unconventional mor-
the design is completed. The introduction of robots phologies of concrete construction systems seek to
in construction industry appears to eliminate or re- save time and cost in the manufacturing process. In

the example of 'Smart Dynamic' system by Gramazio effector tool. The positive and negative parts of form-
and Kohler (Kristensen et al. 2013), the investigation work are then assembled in alternative ways formu-
is concentrated towards the formation of columns us- lating the mold system for concrete casting, leading
ing a flexible formwork that is used as an end-effector to the construction of different morphologies.
mounted on the edge of an industrial robotic arm. The embedding of optimization processes and
In a continuous process, the formwork is filled with especially genetic algorithms on the one hand
self-compacting concrete and in short time period and robotic manufacturing techniques on the other
is moved to the next layer to achieve the desired hand, presuppose a bidirectional form of communi-
height of column. In a different approach and specif- cation between digital design and manufacturing re-
ically in the project 'Mesh molding' (Hack et al. 2014) spectively, aiming to find desirable solutions that can
the robotic arm is responsible to develop a plastic be functionally, statically and morphologically robust
mesh using 3D printing principles, which in turn is and can be directly implemented in actual scale.
used as the formwork and reinforcement for the sub- This research hypothesis is examined further in
sequent concrete filling process. In different mor- several case studies including the ongoing work pre-
phological cases that include the filling procedure, sented in this paper. In this case, the design and
the static behavior as well as the time needed for robotic construction of an optimal concrete wall is in-
the development of mesh construction system are vestigated through the simultaneous development
evaluated. Other techniques like in the example by of two cases of flexible modular formwork systems.
(Stavric and Kaftan 2012), the robotic arm is respon- The application of an optimization mechanism, in
sible to produce a number of formwork units made this case genetic algorithms, aims to allow an in
of polystyrene using a hot-wire cutter as the end- depth investigation towards morphological and con-
Figure 1
Conceptual
framework of the
proposed
methodology.

struction development of flexible modular formwork interrelated and interdependent logic.
systems for concrete casting. Based on a number of Throughout the process of investigation, two
criteria related to the static performance of the con- modular formwork systems are developed that fol-
crete walls, the design of modular systems, the feasi- low respective robotic construction scenarios. The
bility of manufacturing as well as the assembly logic robotic construction criteria play an important role
are examined. in the design of modular formwork, dominating the
manufacturing process as well as the physical results
PROPOSED METHODOLOGY AND RE- obtained during robotic construction in actual scale.
In addition, the robotic construction presupposes
SEARCH DEVELOPMENT
an in depth investigation and construction decision-
The proposed methodology is defined in the form
making in terms of the effective assembly and disas-
of a conceptual framework, describing various activ-
sembly of the modular formwork systems, the struc-
ities that take place in an interrelated manner (Fig-
tural detail of the modular elements, the use of ap-
ure 1). This consists of two basic levels of investi-
propriate end-effector tools, the scale of modular
gation, the digital design and physical construction.
formwork in regard to the dimensions of robotic tools
Within this framework, the design development and
and finally the working environment and logistics re-
assembly logic of the modular formworks is followed
garding the robotic construction in actual scale.
by the overall design and robotically driven physical
Apart from all the above, the digital design pro-
construction of the systems.
cess is also determined by the implementation of ge-
In a preliminary stage, the digital development
netic algorithms, wherein the evaluation of the con-
and configuration of selected modular formworks as
struction systems' static performance is examined
well as the material fill morphology are investigated
and optimized based on the minimum displacement
and optimized against static performance criteria us-
and volume of material fill. Throughout the optimiza-
ing genetic algorithms. The analysis results feed the
tion process, a number of results are derived, that
process of modular formwork design refinement and
consist the pool of possible candidate solutions for
improve the solutions in digital as well as physical
physical construction in actual scale using a number
level, which are determined via robotic construction
of robotic intervention scenarios. As it has been al-
principles. Then, the suggested design of modular
ready mentioned, within the framework of this paper
formworks and assembly as well as disassembly logic
the digital design level of investigation is followed by
are tested in relation to the physical construction per-
the selection of two case studies that are physically
formance of the overall concrete wall systems. In or-
built. This is done in order to test our hypothesis in
der to achieve this, physical prototypes in scale 1:2
an introductory and experimental level as well as to
are produced, examining in parallel potential man-
examine the static behavior of the systems. Also, to
ufacturing drawbacks and obstacles that might oc-
validate the robotic construction criteria applied in
cur during the process. In parallel, through digital
these cases so that effective assembly and disassem-
robotic simulation, the robotic construction proce-
bly procedures of the modular formwork systems can
dure is defined according to the given scenarios. Last
be achieved.
step in this workflow is the execution of robotic con-
struction process in real time, a final stage of develop-
ment that is influenced by previous conceptual and MORPHOLOGICAL DEVELOPMENT AND
refined design stages, which consists of a number of ASSEMBLY PROCEDURE
necessary cycles of evaluation in the form of feed- In the first case study (case A), the alternation of the
back loops. Actions taken within the digital design system in digital design level is defined by parametric
and physical construction levels are acting within an design exploration and criteria directly related with

the assembly and disassembly of the formwork units of data of the height and width of the construction
(Figure 2). In a preliminary phase, the appropriate system as well as on the thickness variability of the
morphology is selected through a design optimiza- module. By combining the three parameters in dif-
tion process and then, modules are assembled to- ferent ways, a range of possible morphological solu-
gether using an articulated robotic arm, which is re- tions in regard to the development of the system of
sponsible to perform the overall manufacturing pro- formwork can be achieved.
cess. In the second case (case B), the design as well In case A, an identical morphology of module
the robotically driven assembly procedure is inves- based on a folded geometry is used that is repeated
tigated using respective methodological principles and placed horizontally in layers from both sides cre-
(Figure 3). ating an interior space enabled to be filled using
building material, in this case concrete (Figure 2). The
Figure 2 morphology of the case B is the result of a single mod-
Morphology of ule that consists of four side surfaces of which two of
modular formwork them are folded. As a final result, a perforated mas-
system and results sive wall is developed, where the logic of layout as-
of material fill in sembly is performed based on individual units (Fig-
case A. ure 3). In both cases, different combinations of the
above parameters provide different results as regards
the dimensions and volume of the overall construc-
tion system as well as the size and shape of openings.
Figure 3
In physical construction level, the suggested
Morphology of
modular formworks are evaluated against specified
modular formwork
assembly, disassembly and filling procedures, lead-
system and results
ing to the production of the final construction out-
of material fill in
comes. For each case, a respective scenario of man-
case B.
ufacturing process that contains all the above is de-
fined. Regarding the construction scenario of case A,
Figure 4 two embodiments of the system assembly are sug-
Preliminary design gested whereas in the first case the entire formwork
investigation of the system is constructed by the robotic arm through a
end-effector tool for pick-and-place process using a gripper as the end-
each case scenario. effector tool (Figure 4), and then this is filled with
concrete involving human assistance. In the second
case, the system is manufactured and then filled with
concrete using a process of layering, which contains
a simultaneous assembly and filling process in each
row. In this case, the robotic arm is responsible for
the assembly of system, which is filled in a later stage
through human intervention. The process is gradu-
In this preliminary phase of investigation, both ally repeated according to predefined levels of the
morphologies are parametrically defined using the system, leading to the production of a complete wall.
parametric-associative design software Grasshopper For case B, the process is defined through the ap-
[1] (plug-in for Rhino [2]). The interchangeability of plication of a custom end-effector tool (Figure 4) in
geometry in both cases is based on the proliferation

Figure 5
Selection of best
solutions (BS) and
design variables for
case A.
Figure 6
Graph of the
average values of
displacement
through
generations for
case A and B.
the form of a kinetic modular formwork, which, once quently influences manufacturing implementation.
is filled with concrete, produces successive concrete The design optimization process is achieved using
unit components, aiming to develop a single con- the Octopus software [3] (plug-in for Grasshopper)
crete wall system at the end of the repeated proce- based on a number of objectives that includes min-
dure. The end-effector tool adjusts its morphology imum displacement under vertical loads and mini-
through a number of kinetic behavior mechanisms mum material volume of the construction system.
that contain rotation in longitudinal axis of formwork The static analysis and the numerical results are ob-
and horizontal movement. For filling and construc- tained using the Millipede software [4] (plug-in for
tion of a single component, the assistance of con- Grasshopper). This provides a continuously iterated
structors is required. feedback loop process, where results are statically
evaluated against the morphology of the construc-
DESIGN OPTIMIZATION PROCESS tion systems and hence the modular formwork de-
The optimization process is implemented in digital sign.
design level via static performance criteria in order In this research investigation, the optimization
search for best fitting solutions, leading to results is defined as multi-objective since the static analysis
that will be manufactured in actual scale. Also, this and overall evaluation contains not only morpholog-
might initially assess the possible design solutions ical but also static performance objectives, formulat-
as regards the dimensions of modules and hence ing the overall range of criteria that are interrelated
the overall dimensions of the system, which subse- and interdependent with each other. The Octopus
plug-in, an evolutionary search engine based on ge-

netic principles, is able to produce and export multi- metric design procedure of experimental settings is
tude solutions through the iterative process of gen- used as the starting point for the optimization pro-
erations. Each displayed solution contains different cess, aiming to search for best solutions that can de-
combination of variables of design values in order to scribe the morphology of construction systems ac-
achieve minimum objectives. In each generation two cording to robotically driven and static performance
solutions are reproduced to generate a new offspring criteria. In case A, based on the results obtained dur-
and hence new combinations of variables. ing a first run of optimization process, it can be ob-
served that over the generations, a number of combi-
Figure 7
nations that contain constant values in terms of thick-
Graph of the
ness and number of units in height axis and variability
distribution of
in terms of the number of units in width axis are ob-
solutions according
tained. The increase of dimensions of modules in the
to the objectives of
longitudinal axis results reduction of the static dis-
height and
placement performance of the overall construction
displacement in the
system. Figure 5 demonstrates a number of best so-
second run of
lutions (BS) together with the values of design vari-
optimization for
ables. Also, Figure 6 demonstrates the objective val-
case A.
ues of displacement and volume, which are gradually
reduced during the evolution of generations. This
shows the tendency of best solutions to reach mini-
mum values, which is also demonstrated through the
Pareto front, distinguishing in parallel the solutions
as dominated and non-dominated. The initial run of
optimization is followed by a second run where an ex-
tra objective is added and rearrangement of variables
Figure 8
is set, showing an alternative and improved method
Selection of best
of investigation leading to new results (Figure 7). In
solutions (BS) and
the next stage, the results derived from the design
design variables for
optimization process are feed into the robotic con-
case B.
struction process to produce the system in physical
actual scale.
In case B, the results shows downward value
trend of the objectives (Figure 8). Also, it can be
observed that values in regard to the thickness and
In each case, same parameters and objectives are number of modules in width axis remain fixed, but
applied in order to compare and correlate the results values that specify the number of modules in longi-
obtained during the optimization process. Initially, tudinal axis are differentiated. Based on the best so-
the design variables are based on empirical measure- lutions derived from this optimization run, it is con-
ments, involving robotically driven criteria that are cluded that by reducing the number of construction
described in previous sections. The thickness val- units in elevation, the static displacement is also min-
ues range from 0.2-0.5m and propagation parame- imized.
ters of units in height and width axes are defined by
integer values range from 2-8. This preliminary para-

Figure 9
Robotically driven
assembly
procedure for case
A.
PHYSICAL CONSTRUCTION DRIVEN BY physical environment where the assembly and disas-
ROBOTIC PRINCIPLES sembly of formwork modules occur, important fac-
The digital design and optimization parts of the pro- tors that influence robotic construction in various
cess influence decisions taken during the investiga- parts of investigation are also taken into considera-
tion of physical construction, especially in regard to tion. The aim is to automate the process and solve
the development of the modular formworks. The drawbacks as well as limitations of modular form-
results and best solutions derived from design op- work design so that this can be effectively executed
timization process are taken into account for both during the robotic construction implementation.
static performance reasons but mainly for the effec- For experimental purposes, the scale of modu-
tive execution of robotic construction process. In lar formwork assembly is determined by the work-
Figure 10
Robotically driven
assembly
procedure for case
B.

Figure 11
Physical protypes
demonstrate the
modular formwork
assembly logic,
material fill and
posible concrete
structure of case A.
Figure 12
Investigation of
modular formwork
positioning and
material fill in case
B.
ing space of the robotic laboratory and the dimen- place procedure using an industrial end-effector is
sions of the end-effector tools. Also, due to the lim- used, while in case B (Figure 10) a custom made end-
ited movement of robotic arm within the working effector tool is applied, driven by the specific scenario
cell, the structure is placed in an appropriate position that requires special positioning of tools in combina-
in order to enable the robotic performance and espe- tion with material fill.
cially the assembly and disassembly of modular form- In this stage of investigation, where preliminary
works. In case A, the dimension of modules is also design and construction of both concepts are tested,
influenced by the dimension of gripper so that an ef- physical prototypes in scale 1:2 are produced. In case
fective pick-and-place procedure can be achieved. In A (Figure 11), identical modules are manufactured
case B, where a custom end-effector is applied, de- and assembled in order to evaluate the static perfor-
cisions are taken in regard the design of formwork mance, the errors occurred during construction pro-
that include dimension limits for kinetic behavior and cess as well as the morphological result and aesthet-
overall movement of robotic arm. ical aspect of the material fill procedure. In case B
In both case, the robotic construction behavior is (Figure 12), the proposed custom made tool is man-
simulated using the HAL software (Schwartz 2012) [5] ufactured and tested in regard to the assembly, ma-
(plug-in for Grasshopper). The simulation is applied terial fill and disassembly processes. In both cases, as
to assess the effectiveness of the concept, confirm- filling material a perlite mix with cement and water
ing in parallel the design configuration of the pro- is used, which results light mass and rapid curing in
posed modules and systems in regard to the required short time period.
movement. Also, possible collision and constrains of
movement of the robot are investigated. As it has
been mentioned, in case A (Figure 9), a pick-and-

CONCLUSIONS York, Austria, pp. 184-195
The optimization process introduced in this ongoing Lee, D, Lim, H, Kim, T, Cho, H and Kang, K.I.A 2014
work, aims to define the boundaries of feasible so- 'Formwork Layout Model based on Genetic Algo-
rithm', 31st International Symposium on Automation
lutions and to allow assessment of robotic interven-
and Robotics in Construction and Mining, pp. 474-479
tion during different stages of the modular formwork Poblete, B and Ignacio, C 2011 'Evolutionary Strategy to
systems development. In addition, the results ob- Design Optimized Architecture, Respecting Fragile
tained might address an innovative design and con- Places', 29th eCAADe Conference Proceedings, Univer-
struction process of folded/porous concrete walls by sity of Ljubljana, Faculty of Architecture, Slovenia,
minimizing the use of material and hence its cost as pp. 293-301
Pronk, A.D.C, Dominicus, M.M.T, von Buelow, P, van Dijck,
well as the time needed for modular formworks as-
S.H.M and van de Koppel, W.J 2013 'Rigidized inflat-
sembly and disassembly. Through the standardiza- able structures a production method for optimized
tion and practicality of the process involved, feasible structures', Proceedings of the International Associa-
results might be obtained. Further work will consider tion for Shell and Spatial Structures (IASS) Symposium:
the physical development of the suggested solutions ”BEYOND THE LIMITS OF MAN”, Wroclaw University of
in actual scale based on the methodology described Technology, Poland
Richardson, J.N, Nordenson, G, Laberenne, R, Coelho, R. F
in this paper. This presupposes the parallel and in
and Adriaenssens, S 2013, 'Flexible optimum design
depth investigation as well as solution-finding in re- of a bracing system for façade design using multiob-
gard to the problems and drawbacks occurred during jective Genetic Algorithms', Automation in Construc-
the digital design and optimization phases, inform- tion, 32, p. 80–87
ing the physical construction process within a feed- Schwartz,, T 2012, 'HAL', in Brell-Cokcan, S and Brau-
back loop logic. mann, J (eds) 2012, Robotic Fabrication in Architec-
ture, Art and Design, Springer Wien New York, Aus-
tria, pp. 92-101
REFERENCES Stavric, M and Kaftan, M 2012 'Robotic fabrication of
Foged, I.W, Pasold, A, Jensen, M.B and Poulsen, E.S 2012 modular formwork for non-standard concrete struc-
'Acoustic Environments: Applying evolutionary al- tures', In Physical Digitality: Proceedings of the 30th
gorithms for sound based morphogenesis', In Digi- eCAADe Conference, pp. 431-437
tal Physicality: Proceedings of the 30th eCAADe Con- Søndergaard, A, Amir, O and Knauss, M 2013 'Topology
ference, pp. 347-353 Optimization And Digital Assembly Of Advanced
Hack, N, Lauer, W, Gramazio, F and Kohler, M 2015 'Mesh Space-Frame Structures', Acadia 2013: Adaptive Ar-
Mould: Robotically fabricated metal meshes as con- chitecture - Proceedings of the 33rd Annual Conference
crete formwork and reinforcement', Ferro-11: Pro- of the Association for Computer Aided Design in Archi-
ceedings of the 11th International Symposium on Fer- tecture (ACADIA), Cambridge, pp. 367-378
rocement and 3rd ICTRC International Conference on [1] http://www.grasshopper3d.com/
Textile Reinforced Concrete, pp. 347-359 [2] https://www.rhino3d.com/
Kristensen, E.L, Gramazio, F, Kohler, M and Langenberg, [3] http://www.grasshopper3d.com/group/octopus/
S 2013 'Complex Concrete Constructions: Merging [4] http://www.grasshopper3d.com/group/millipede/
existing casting techniques with digital fabrication', [5] http://www.hal-robotics.com/
Open Systems: Proceedings of the 18th International
Conference on Computer-Aided Architectural Design
Research (CAADRIA 2013), Asia, p. 613–622
Kurilla, L and Svoboda, L 2013, 'Geometry Optimization:
Realization of fluid-form structure composed of
spherical components, fabricated by means of com-
puter software and robotic arms', in Brell-Cokcan,
S and Braumann, J (eds) 2013, Robotic Fabrication
in Architecture, Art and Design, Springer Wien New

AMA - Additive Merged Appliance
Alina Rejepava1 , Mario Haring2 , Anna Gruber3
1,2,3
Graz University of Technology
1,2,3
{rejepava|mario.haring|ann.gruber}@student.tugraz.at
The production of concrete structure components is often a backbreaking process

due to the inconvenient and time-consuming process of producing the formwork.
Depending on the geometry, this can be a very time-consuming activity. A trial
will be made to solve these problems within the manufacturing process using
adapted binder jetting technology. First of all, the research will deal with the
behavior of concrete as well as the bond between the cement paste and the
aggregate. Various additives and grain sizes will be determined to find an
appropriate result. On the other hand, different spray and drop systems will be
analyzed to compare the new method and the common binder jetting setup. After
fabricating some geometrical shapes manually the study is going to be continued
using computer-aided manufacturing.
Keywords: 3D Print , Binder Jetting, Concrete, Robotics, Additive Fabrication
INTRODUCTION architecture.
3D printing or additive manufacturing is a process of The founder of "D-Shape" (Dini 2010), Enrico Dini,
making three-dimensional solid objects from a digi- implemented a methodology to create structures
tal file. Usually they are created by laying down suc- and sculpturesin a big scale. He developed a strategy
cessive layers of material until the object is entirely to produce a monolithic, seamless geometry.
assembled. Each of these layers can be seen as a Two scientific works from the ETH Zurich show
thinly sliced horizontal cross-section of the final ge- that the scope for designing digitally developed
ometry. This technology is presently being used for walls is almost limitless. "Digital Grotesque" com-
a number of different materials. At the moment, a bines technology and nature in a very novel way
lot of research is being carried out in the area of 3D (Mayer 2013). The commonly recurring problem is
concrete printing. The main questions that need to the fragility of the material that the structures are
be investigated, deal with the process, the materials made of. They can only bear their own weight and
and the scale. There are already several documented no additional loads.
projects about the progression of 3D printing tech- The aim of "AMA - Additive Merged Appliance",
nology. (Bogue 2013) a modified version of the binder jetting process, is
The main focus of "Concrete Printing" (Lim et al. to bind the cement paste with the aggregates and in
2011) is the construction of buildings. One major this way to produce load-bearing architectural com-
topic is how to include reinforcement into the print- ponents.Due to the application of a binder and sup-
ing process. Once this is possible, it opens the door port material it is an economic and ecological manu-
to 3D printed, self-supporting, sustainable, free form facturing process (Figure 1).

Figure 1
Drawing of
common binder
jetting and the
adapted setup.
EXPECTATIONS were rapid-hardening cement, water and

The trials are going to look at the bond between sand. Sand was used in the cement mixture.
the cement, the water and the aggregate. Tools for The water was then sprinkled onto the ce-
the robot need to be developed for the computer- ment mixture to create a 3D shape.
aided manufacturing. One tool needs to be devel- 3. Thick layers of rapid-hardening cement were
oped for each material. The computerized control spread before spraying water on them. This
system should enable adjustable layer height as well achieved a smoother surface and enabled the
as targeted application of cement paste. liquid to fill the small voids between the sand
AMA fabrication should be able to adjust the ag- particles.
gregate grain size meaning that the problems of scal- 4. Plain cement and sand, in a 1:1 ratio, were
ing will then be irrelevant. Also, the process should mixed and spread before moistening it.
allow for the accurate placement of cement paste.
This means that the process will be more precise. The first experiment showed that it is impossible to
achieve an effective combination of cement and wa-
ter when the water is being sprayed onto the layers
METHODS AND RESULTS
of cement.
A series of experiments were performed to get first
The results of the different methods showed that
impressions of how binder jetting works.
nothing had been produced that would be usable.
Water Spray Attempts The samples were fragile and horizontally stratified.
The first trials were based on the basic concrete in- The second and third experiments were dissected
gredients such as water, cement and sand. and it could be seen that internally the layers were
not totally connected.
1. Rapid-hardening cement was spread over a Cement Paste
small area to testing the method. Water was The previous test showed, water is not an effective
sprayed onto the thin layer of cement using binder on its own. A different binder had to be used.
a nebulizer. A second layer of cement was Different combinations of cements with water were
spread over the first sprayed with water. This tested to achieve an appropriate cement paste (Fig-
was repeated several times. After the fifth ure 2).
layer, a roller was used to compact the layers.
In comparison to binder jetting process, the Figure 2
quick cement substitutes the powder, and the Cement paste tests
fluid binder is substituted with water. using different
2. The main components of the second step aggregates. Sand,
fine aggregate and
coarse gravel
180 | eCAADe 34 - FABRICATION | Robotics: Generative Design - Volume 1 aggregate.
Figure 3 1. In the first tests quartz sand and cement paste
Test object. coarse were layered alternately.
angular aggregate 2. In the second test, fine gravel aggregate was
covered with mixed with sand and covered with cement
cement paste. paste.
3. Thirdly, coarse gravel was used in combina-
tion with the cement paste.
The results were much improved now that cement

paste was being used as the binder. It was difficult
to bind the sand using the cement paste as the pores
between the grains of sand were too small and the
paste could not permeate the first layer of sand. The Workability of the Cement Paste
second test with fine gravel and sand showed a sim- The experiments with the nozzle did not provide sat-
ilar result. The most successful test was the coarse isfying results. This meant the process needed to be
gravel. The cement paste could pass through the changed. The next investigation involved dropping
spaces between the grains of the gravel. Substituting the paste through a funnel that was placed in the cor-
rapid-hardening cement with normal cement made rect position.
no difference to the final appearance. It only speed Therefor investigations were carried out into the
up the drying time and changed the color of the ob- workability of the cement paste. Three different
ject. compositions were tested by dropping them on the
gravel by hand. The different mixtures effected the
Cement Paste Spray appearance of the surfaces. The flow and drop be-
Tests were then carried out to investigate the nec- havior were taken into account and the most appro-
essary features of the nozzle. It showed how accu- priate water cement ratio was selected based on the
rate the nozzle was and what pressure was needed appearance of the test objects.
and the size possible spraying radii of the deposited
cement. A wooden box was produced for the ex- Binder Jetting by Hand
periments. This made sure the grit was laterally re- The next step involved testing a possible binder jet-
strained during the drying phase. The tests were as ting setup for usability. A device consisting of two
follows: wooded boxed, one on top of the other, was used
(Figure 4).
1. Using an air brush and a compressor to spray
Figure 4 the cement paste.
Setup: Top = 2. The second system was a spray painting sys-
printing area, tem from a company called Wagner. This had
Bottom = space for a small nozzle and a large spraying radius.
the loose
These steps delivered important results. The airbrush
aggregate.
compressor was deemed unusable as the pressure
was far to high. The spray painting system was the
more appropriate. One disadvantage is the time fac-
tor (Figure 3). The size of the spray radius also needed
to be reduced as the application area of the cement
The two boxes were separated by two horizontal di-
paste was large and uncontrollable.

aphragms. One is perforated so that the aggregate 3. The mold was removed.
can pass through and the second is not. 4. The void from removing the mold was filled
The new printing process, performed by hand, with sand.
consisted of 3 steps: 5. The mold was put back in the box onto of the
sand layer and cement paste poured into the
1. First, the aggregate was spread on the base mold.
of the top box in a layer which had a defined 6. The mold was removed.
thickness.
2. Second, a mold was placed in the box and
Figure 5
filled with cement paste.
Cross-section of the
3. Finally the mold was removed and the process
test objects: crystal
started again.
quartz sand,
This process was repeated 10 times. The mold was a The test with sand as supporter and angular gravel expanded clay and
little different for each layer. The opening increased as aggregate was successful. The inside of the pyra- coarse grit.
in each layer. This meant that a self supporting object mid was homogeneous mixture of the cement paste
would be produced upside-down. and gravel and the surface was clearly defined by the
Once the cement paste had dried, the solid di- sand particles (Figure 6).
aphragm without perforations was removed. This al-
Figure 6
lowed all of the loose aggregate to fall down through
Pyramid printed
the perforated diaphragm. The upper box was a cas-
upside down -
ing for the printing process and the lower box was
angular grit as
now used to contain the loose aggregate.
linked material and
Different aggregates were tested to see how ef-
sand as supporting
fectively they worked. The first aggregate was a
material
coarse angular gravel with a grit size of 5-8 mm. The
next was crystal quartz sand with a grain size of 3mm
was tested. Another material used was expanded
clay.
The crystal quartz sand provided for a top qual-
ity surface but due to a small particle size which con-
Producing a self supporting object by hand
To test the possibilities of binder jetting a shape
dense the layers, the cement paste could not flow
which had large overhangs was printed (Figure 7).
through it. Both the crystal quartz sand and ex-
For the first samples crystal quartz sand was used as
panded clay samples presented a separated appear-
both supporting and binding bind material. As be-
ance in layers. The test with the chunky grit showed
fore, a layer of aggregate was spears and then the ce-
the best results: The layers were perfectly connected
ment paste poured.
and not visible. The sample appeared to be homoge-
nous (Figure 5). Figure 7
To optimize the surfaces, two different materials "Overhang" -
were then used. The coarse gravel was bonded and defined by a core
the sand was loose. and horizontal
These steps were repeated for each of the molds. elements.
1. The mold was placed in the retaining box.
2. The mold was filled with gravel. The cement paste did not seep through from the

most recently printed layer into the lower layers. The Hollow Cube
aggregate which was not bonded acted as a support A shape was digitally modeled that could show the
material for the next layer. capabilities of the adapted binder jetting process. It
was a simple shape with interesting surfaces. The
Robotics shape focused on testing the limits of the process us-
ing overhanging parts.
Figure 8 The hollow cube was the divided into 8 equal
AMA setup: robots parts. as the writing area was also limited in size. The
with linear print digital model was then sliced using Repetier Host a
head (left) and cap computer slicing software. To be able to test over-
for the cement hangs the digital model was turned upside-down.
paste (right). It needed to be sliced in order to obtain the target
points necessary to define the robots movements.
(Figure 9).
Figure 9
Drawing of the
hollow cube. Using a printing area that is surrounded by casing
a multitude of different shapes are possible. Two
industrial robots, ABB IRB 140s, with different print
heads were applied (Figure 8). One tool was used
for filling up the layer with coarse aggregate. It is a
linear funnel, which leveled the surface of the aggre-
gate. The second print head was a cone shaped fun- The next consideration of the computer-aided fabri-
nel with a small round opening at the bottom. The cation was the automatization of the cement paste
cement paste was filled into the cone where it was supply. This was carried out using a sturdy printer
stored until the locking bar opened and the cement head and a booster pump (Figure 10).
paste could flow out of the tube onto the aggregate.
Figure 10
Tool path with an
alternative printer 1. The exact amount of aggregate for one layer
head. was filled into the casing and Robot 1 leveled
the aggregate.
2. Robot 1 returned to its waiting position.
3. Robot 2 then moved to the correct position
and opened the locking bar and placed the
cement paste where it was needed. DISCUSSION
Although one suitable cement paste mixture has
been found for the relationship between the aggre-
The path of Robot 2 changed for every layer. The gate size some parts of the manufacturing process
thickness of each layer was defined by the size of the still need to be fine tuned.
aggregate. The workability was defined depending The modified binder jetting process is depen-
on the size of the aggregate grain so as to provide dent on the aggregate size, the aggregate shape, the
the perfect bond between the layers. height of the layers and the workability of the cement

paste and all of these parameters have an effect on out using conventional systems including form work.
the internal bond between the lines of printed ce- Because of the loose supporting materials it is possi-
ment paste. Due to high variance in the chosen scale ble to produce nested and twisted elements.
of the expanded clay and rounded shape of the crys- This technology can be used on and in any scale.
tal quartz sand as well as the expanded clay, the inner The possibilities of using the technology in a prefab-
structure of the object was non-satisfying. To achieve ricated concrete plant could be a viable application.
a better quality using these materials it is possible to This method is not expected to replace traditional
add aggregates with a larger grain size or more an- concrete production, it aims to offer more possibili-
gular aggregate grain. Using the angular aggregate ties is regards to the design and realization of mate-
achieved the best results for the scale of the printing rial efficient objects as well as speed up the produc-
area that was tested. tion process.
The possibility of using different kinds of aggre-
gate and severalgrain sizes makes it possible to pro- REFERENCES
duce smooth surfaces and therefor remove the ne- Bogue, R 2013, '3D printing: the dawn of a new era in
cessity for post processing. manufacturing?', Assembly Automation, 33, pp. 307-
Further investigations should include the rela- 311
tionship between the grain size and the layer height, Dini, E 2010 'Full-Size Stone 3d Printing', Smart Geometry
the mixture of water-cement with each grain to Symposium, Barcelona
Lim, S, Buswell, R, Le, T, Wackrow, R, Austin, S, Gibb, A
achieve a seamless bond between the layers.
and Thorpe, T 2011 'Development of a viable con-
The technique of removing the loose aggre- crete printing process', Loughborough University In-
gate by establishing space below the printing area, stitutional Repository, UK, pp. 665-670
demonstrates a promising approach to reduce the Mayer, H 2013, 'Aus Staub wird Gold', Archithese, 3, pp.
total fabrication time. However, the success of this 86-87
procedure depends on the form of the object. Hori-
zontal surfaces or enclosed regions prevent the ag-
gregate support material from falling away from
the object. When there is a horizontal surface or
an enclosed region, additional automated processes
would have to be developed. It could involve a lift-
ing tool for the finished shapes. As previously known,
the tool path is a continual path which does not have
any breaks in it. To create spaces it is necessary for
the locking bar of the nozzle to be mechanically op-
erated. Also the process of forwarding the cement
paste could be improved.
OUTLOOK
In comparison to other concrete 3D printing pro-
cesses, "Additive Merged Appliance" has demon-
strated great potential in fabricating complex geom-
etry for use as an architectural solution.
For example, modular structures consisting of
overhangs and unique parts can be produced with-

Generating an Anamorphic Image on a Curved Surface
Utilizing Robotic Fabrication Process
Marko Jovanovic1 , Vesna Stojakovic2 , Bojan Tepavcevic3 ,
Dejan Mitov4 , Ivana Bajsanski5
1,2,3,4,5
Faculty of Technical Sciences, University of Novi Sad
1,2,3,4,5
{markojovanovic|vesna100|tepavcevicb|dejan_mitov|ivana_b}@uns.ac.rs
The integration of industrial robots in the creative art industry has increased in
recent years. Implementing both brick stacking robotic fabrication, following a
curved wall, and generating an image viewed from a single point, by rotating the
bricks around their centres, has yet to be studied. The goal of this research is to
develop a functional, parametric working model and a workflow that ensure easy
manipulation and control of the desired outcome via parameters. This paper
shows a workflow for the automatic generation of anamorphic structures on a
curved wall by utilizing modular brick-like elements. As a result, a code for the
robot controller and the position of the structure during fabrication are provided.
Keywords: anamorphosis, brick lying, robotic fabrication, generative design
INTRODUCTION sign works generated by utilizing a robotic arm have

In the last decade, the implementation of industrial become part of performance art [1]. As a tool for
robots as fabrication tools has presented itself as be- fabricating artwork, industrial robots have been used
ing productive, effective and manageable (Keating for generating paintings with an element of ran-
2013; Braumann 2013). Therefore, it is important to domness (Aguilar & Lipson 2008; Alkhodairy & Pa-
consider fabrication type in the early design stages. tel 2014). Regarding studies examining the utiliza-
This is due to the multipurpose nature of the indus- tion of anamorphosis, with shapes on freeform sur-
trial robot, which is the primary fabrication tool used faces, recent studies have been published on project-
for a specific purpose. Robotic arms function by moving vector-based shapes (Di Paola et al 2015). In ad-
ing its parts via joints, and the robots are defined as dition, research regarding the generation of anamor-
having specific degrees of freedom, usually 3 to 6. By phic sculptures have discussed the question of el-
utilizing a gripping or suction tool, a robot can uti- evating ideas from a 2D image to a 3D sculpture
lize a specific modular element, such as a brick-like (Manal 2015). Furthermore, in 2006, Gramazio and
module (Yu Nam et al. 2009; Bärtschi et al. 2010; Kohler utilized a raster image as a guideline for brick
Bonwetsch et al. 2007), and follow a computational rotation following a planar surface in the Gantenbein
model to assemble a structure resembling the origi- Vineyard Façade in Switzerland.
nal (Rakovic et al. 2014; Vomhof et al. 2014). However, implementing both brick-stacking
The integration of industrial robots in the cre- robotic fabrication, following a curved wall, and gen-
ative art industry has increased in recent years. De- erating an image viewed from a single point, by ro-

Figure 1
a) The brick
tessellation
algorithm applied
to a extruded
surface b) The
extracted brick
center points.
tating the bricks around their centres, has yet to be tion and height of the desired image plane (the plane
studied. The goal of this research is to develop a where the image is formed) and the grayscale image
functional, parametric working model and a work- used as a guide. In this research, a custom-scripted
flow that ensure easy manipulation and control of component for the Grasshopper programming lan-
the desired outcome via parameters. As a result, a guage is used to generate the missing code for the
code for the robot controller and the position of the robot's RAPID programming language. Once the pa-
structure during fabrication are provided. rameters are set, a single text file is generated and
prepared to be implemented into a code for the robot
METHODOLOGY controller. The workflow is explained in the following
This research's starting point is the early implementa- section.
tion of robotic fabrication and its insight into the de- Figure 2
sign process. Here, it is the robotic arm that is utilized The setup of the eye
to place brick-like elements in the form of a curved position, the image
surface. Additionally the bricks are rotated around plane and the
their central axis in order to form an anamorphic initially tessellated
raster image that can be input and this image is user surface before brick
defined. Hence the end result has to comply with the rotation and image
necessities and requirements this fabrication type guideline is
dictates, which is a code for the robot controller - a introduced.
set of instructions that a robot follows and executes.
The main parameters used for defining the anamor-
phic sculpture are: the input surface, to which the
brick-like elements are stacked, the size of the bricks,
the minimal distance between the bricks, their max-
imum and minimal rotations, the height and posi-
tion of the viewer's eye point, the position, orienta-

ANAMORPHIC IMAGE ON A CURVED BRICK ements are taken and edge centre points are chosen
STRUCTURE for the position of the brick-like elements to come
In order to fabricate the desired anamorphic image as (Figure 1b). They are defined as C1 , C2 ... Cn , where
a curved brick-like structure, the workflow is divided n is the total number of bricks in the tessellated sur-
into three phases. The first phase explains the initial face. Afterwards, the eye-point position, defined as
brick setup following a curved surface and the eye Ep, has to be set, meaning the place where the viewer
position for the generation of the anamorphic image. needs to be in order to see the anamorphic image as
Based on the eye position, the second phase rotates a planar image. Next, the image plane, defined as α,
the bricks around their central axis in order to pro- needs to be set, meaning the position, the orienta-
cure adequate anamorphic image generation, where tion in reference to the eye point and the size. The
the bricks are viewed as pixels. Lastly, the structure setup is visible in Figure 2.
is then used to extract the necessary data and tran- Finally, a grayscale image is chosen and input as
scribe it into a code for the industrial arm to follow. a guideline for the brick rotation, which is explained
All the phases are explained in greater detail in the in the following subsection.
following subsections.
Brick Rotation for the Anamorphic Setup
The Initial Brick Setup The brick rotation is carried out in a twofold pro-
The first phase is the surface reference and parame- cess. The first rotation is done in accordance with
ters setup for the desired outcome. This phase starts the eye point. The brick-like element centre points,
by generating and referencing a single surface in the C1 ,C2 ...Cn , are evaluated according to the surface
Grasshopper programming environment. Here, an they follow and normal vectors, defined as nn , are
extruded curved shaped is chosen. Once the param- extracted as a result (Figure 3a). By connecting these
eters for the brick-like elements are set, including the points to the eye-point position, Ep , another set of
length, width and height, the surface is trimmed, if vectors is generated, en (Figure 3b). The angles be-
necessary, to address the brick-fitting tessellation ac- tween these two sets of vectors are then compared
cordingly (Figure 1a). to the maximum brick rotation angles, set at around
The lower horizontal edges of the tessellated el- 45 degrees. This corresponds to the borderline rota-
Figure 3
a) The first set of
vector, normal
surface vectors; b)
The second set of
vectors, connecting
the brick positions
to the eye point.

tion angle, while keeping in mind the stability of the Generating the Code for the Robot Con-
entire structure, as one brick lays on top of the two troller
below. The values are then remapped as to atone for The code for the robot controller consists of two
the corrections. This step is necessary since this is a parts: the 'target' of the bricks in the final anamor-
central and not an orthogonal projection and map- phic structure and the stacking process. The first part
ping process. of the code, the 'target', refers to the position and ori-
The second rotation is carried out in accordance entation of the bricks in the anamorphic structure.
with the image chosen for the anamorphosis. The al- They are extracted from the Grasshopper environ-
gorithm references the gray-scale image as a guide- ment by using a custom script that references the
line for the brick rotation in reference to the eye bricks and generates a code that the robot can in-
point. The algorithm starts by utilizing the en set terpret. The stacking process is written in Robot Stu-
of vectors to generate the lines connecting C1 , C2 , dio, a software for the offline robotic programming
...Cn and Ep , respectively. These lines intersect the and robotic arm control. The stacking consists of two
image plane and produce a set of points, defined as functions or processes: picking up the brick from a
I1 , I2 , ...In , as a result. The points have coordinates desired position and then placing it in the previously
in the Cartesian coordinate system. These coordi- defined targets defined previously. The final phase
nates cannot be utilized as a reference for the two- implies taking the generated code for the robot con-
dimensional gray-scale image in this form. Therefore, troller, derived from the input parameters previously
the image plane is reparametrized, meaning its do- described, and using it in RobotStudio software for
main is set as a function of percentage in reference controlling the industrial robot. Robot Studio then
to the two main surface directions, U and V, ranging utilizes the centre points and orientation of the bricks
from 0 to 1 and not according to its actual size. The in- from the 'target' code. The robot picks up the brick
tersecting points are evaluated according to the im- from a single position in its vicinity and places it in the
age plane, and a set of UV coordinate points is gen- respective points in an iterative manner. In tribute to
erated. By changing the size of the image plane, the Alvar Aalto, who used bricks and masonry as material,
resulting points are not impacted domain-wise. The along with curved lines, for the purposes of generat-
main reason for generating a UV set of coordinates ing an anamorphic sculpture, his portrait is chosen.
is to adequately align them to the two-dimensional
planar gray-scale image, where the domain is also RESULTS AND DISCUSSION
set from 0 to 1. The aligned points on the gray-scale First, a desired input surface needs to be generated
image reference the colour values which are then and an image for the anamorphosis chosen. In order
remapped to the desired minimum and maximum ro- to generate an anamorphic structure that is visible
tation values for the brick-like elements. These values from a single viewpoint, its position has to first corre-
depend on the desired end result, the eye-point posi- spond to the size and position of the surface guide-
tion and the initial surface chosen as a guideline and line. Given the narrow field of clear vision in the hu-
they are empirically determined. Finally, those values man visual cone, around 30 degrees, the eye point
are used on the previously rotated bricks to generate has to be positioned so that the visual cone com-
the desired image. After the bricks are rotated into pletely encloses the input surface. Otherwise, the en-
place and the anamorphic image visibility is checked tirety of the structure is not visible and, hence, the
in the 3D environment, it is necessary to prepare the anamorphosis of the image is lost. In addition, the
code for the robot controller. image plane has to be enclosed by the cone as well,
or the image will spread beyond the borders of the
input surface.

Figure 4
The grayscale
image as a
guideline (left) and
the final brick
structure according
to the desired
image (right).
Figure 5 the bricks cannot be rotated to adequately suit the

The anamorphic image. In addition, the bricks in the front can ob-
structure with an scure the bricks in the back and produce inaccurate
increased number results. The rest of the wall can occlude bricks de-
of bricks and colour pending on the wall's curvature. Therefore, the vis-
input. ibility of the bricks varies, and the image is simulated
differently. However, if the change of the curvature is
subtle, these variations will not influence the image's
readability.
In order to mitigate this, we used bricks with
edges of only two different lengths (a square brick),
which avoids the rotation issues and, thus, the stabil-
ity issues as well, without impacting the appearance
of the anamorphic image. The number of bricks is
also an important factor. A smaller number of bricks
gives anamorphic structures lower resolution and,
hence, smaller resemblance to the desired image and
vice versa.
Since the bricks form a non-smooth structure,
the overhanging parts of the bricks cast a shadow
Once these parameters are set, attention must turn on the rest of the structure, which can mitigate the
to the choice of brick-like elements. Utilizing a final image's appearance. In order to avoid these
brick with edges of three different lengths, in this issues, two nonadjacent, vertical sides of the bricks
case 70mm length, 30mm width and 19mm height, are coloured black, while the rest is left white. With
proves to be inadequate, as can be seen in Figure 4. an increased number of bricks and the difference in
The bricks cannot rotate fully to atone for the colour, the end result (Figure 5) seems much more
eye-point position correction angle due to the loss realistic than the one in Figure 4b. Even though the
of stability and the support under each brick: hence, structure is anamorphic, it offers a different experi-

Figure 6
The different
depictions of the
anamorphic
structure outside
the eye point
position.
ence while moving from one side to the other, not has a different appearance in reference to the view-
honouring the viewpoint (Figure 6). ing. The result can be applied in fabricating various
The limitation of using this workflow are exhib- wall panels, where the expected result is a different
ited in the distortion and appearance change due to appearance depending on the viewing angle. The
the use of elongated brick shapes, the lack of sup- control can be achieved by changing the parame-
porting area and the instability given the large an- ters, (brick size, viewpoint and image plane), along-
gle rotations. Also, a lower number of bricks will side a desired raster image, generating a computa-
produce inadequate results in terms of similarity, be- tional working model.
tween the anamorphic structure and the desired im- This model can be used for further refinement,
age. Finally, the surface design and the position of utilizing the aforementioned parameters, with an in-
the eye point can cause the viewing directions to be- stant result in viewport. Furthermore, the model is
come tangent to the surface generating a smeared used as a base for generating a code for the robotic
appearance of the image, as seen in the Figure 6c. fabrication process. By manipulating a set of param-
Regarding the robotic fabrication application aspect eters in the above-mentioned workflow, the design
of this research, the size of the robot and the work- process is much more fluent and straight forward,
ing area it can service, limit the size of the structure; with a clear path from the design phase to the fabri-
hence, they must be taken into account at the start cation phase. Limitations of this workflow are the re-
of the workflow. strictions in brick shape, robot working area and the
influence of the wall curvature on the visibility of the
CONCLUSION bricks.
In this paper, a workflow for automatic generation of Future work will focus on fabricating the desig-
anamorphic structures is shown: a curved wall gener- nated structure and comparing the results in real-life
ated by modular brick-like elements. As opposed to conditions with the ones generated here. Also, the
the previous approaches when generating anamor- workflow can use an upgrade in the area of utiliz-
phic structures, the fabrication process is automated ing elongated brick shapes, with the similar results as
by utilizing an industrial robot. Furthermore, the sur- seen here.
face, the image is generated on, is curved. The main
characteristic of this structure, is that the end result

ACKNOWLEDGEMENT [1] http://motionographer.com/2013/10/01/bot-dolly-
This research was supported by the Serbian Ministry box-interview-and-behind-the-scenes/
of Education and Science (projects no. TR36042 and
III44008) and by Provincial secretariat for science and
technological development.
REFERENCES
Aguilar, C and Lipson, H 2008 'A robotic system for inter-
preting images into painted art-work', International
conference on generative art
Alkhodairy, A and Patel, S 2014 'Robot Artist - Automated
Picture Portrait', ASEE 2014 Zone I Conference, April 3-
5, 2014, University of Bridgeport,, Bridgpeort, CT, USA.
Bartschi, B, Knauss, M, Bonewetsch, T, Gramazio, F and
Kohler, M 2010, The wiggled Brick Bond, Wien/New
York, Springer
Bonewetsch, T, Gramazio, F and Kohler, M 2007 'Dig-
itally Fabricating Non-Standardised Brick Walls',
ManuBuild, 1st International Conference
Braumann, J and Brell-Cokcan, S 2013 'Parametric Robot
Control: Integrated CAD/CAM for Architectural De-
sign', Proceedings of the 31st Annual Conference of
the Association for Computer Aided Design in Architec-
ture, ACADIA
Helal, M 2015, 'A simplified procedure for anamorphic
sculpture', International Design Journal, 5, pp. 317-
326
Keating, S and Oxman, N 2013, 'Compound fabrica-
tion: A multi-functional robotic platform for digi-
tal design and fabrication', Robotics and Computer-
Integrated Manufacturing, 29, pp. 439-448
Yu Nam, S, Gab Ryu, B, Jin Lim, S, Jun Kim, C, Kyu
Kang, M and Soo Han, C 2009, 'Feasibility verification
of brick-laying robot using manipulation trajectory
and the laying pattern optimization', Automation in
Construction, 18, pp. 644-655
Di Paola, D, Pedone, P, Inzerillo, L and Santagati, C 2015,
'Anamorphic Projection: Analogical/Digital Algo-
rithms', Nexus Network Journal, 17, pp. 253-285
Rakovic, M, Jovanovic, M, Borovac, B, Tepavcevic, B,
Nikolic, M and Papovic, M 2014 'Design and Fabri-
cation with Industrial Robot as Brick-laying tool and
with Custom Script Utilization', , IEEE International
Conference on Robotics in Alpe-Adria-Danube Region
Vomhof, M, Vasey, L, Brauer, S, Eggenschwiler, K, Strauss,
J, Gramazio, F and Kohler, M 2014 'Robotic Fabrica-
tion of Acoustic Brick Walls', Proceedings of the 34th
Annual Conference of the Association for Computer
Aided Design in Architecture (ACADIA)

CAAD EDUCATION | Design
Concepts & Strategies
Digital Architectural Composition - 30 years of experience
and experimentation
Aleksander Asanowicz1
1
Bialystok University of Technology
1
asanowicz@gmail.com
In this paper the evolution of architectural composition teaching will be

presented and discussed. The main point of these deliberations is to consider how
digital methods and technologies have influenced the teaching methods and the
understanding of compositional rules/principles by students. Another important
factor which will be analysed, is the difficulties faced by teachers in the
evaluation of students' work. The paper is divided into four parts, each of which
concerns different approaches to architectural composition teaching. In the first
part the historical background of architectural composition teaching is presented.
In the second part traditional teaching methods is described, the third concerns
teaching of digital architectural composition, and the fourth - digital architectural
composition in virtual space. At the end of each section, the advantages and
disadvantages of the discussed methods are presented. These disadvantages were
a stimulus for changes in the teaching process, which was enabled thanks to the
development of digital technologies (software and hardware) during the last 20
years. The last part of the paper will be devoted to the presentation of potential
directions of development of architectural composition teaching methods.Please
write your abstract here by clicking this paragraph.
Keywords: Architectural Composition, digital modelling, Virtual Reality
HISTORICAL BACKGROUND Tekhnicheskiye Masterskiye (Higher Art and Tech-

The main problem at the first stage of architec- nical Studios)). The main goal of both schools was
tural education is activation of the students' cre- an integration of all kinds of art and industry, and a
ative abilities. At this stage the lack of empiric ex- break from old design tradition. Searching for a new
perience should be compensated by the knowl- kind of art meant a destruction of the old, formed
edge of composition principles of spatial forms. This new types of problems, and therefore required a new
knowledge creates the basis for all future architec- theory of creation. Kandinsky has written: "Creating
tural professionals' activities. This teaching method abstract space forms will be very important there-
was introduced for the first time about 90 years fore in result the tradition will be destroy. We will see
ago in two great architectural schools - Bauhaus the new tendencies in architecture. [...] Searching
and VKHUTEMAS (Vysshiye Khudozhestvenno- should start from elementary spatial forms as pyra-
CAAD EDUCATION | Design Concepts & Strategies - Volume 1 - eCAADe 34 | 195

mids, cube, and sphere." (Matz 1933, 130) abstraction is quite justified." (Krinsky et al. 1968, 5)
The new idea of composition teaching belonged Eighty years later P.D. Plowright in his book "Re-
to N. F. Ladovsky, who claimed that teaching should vealing Architectural Design. Methods, Frameworks
concentrate on abstract composition themes, with- and Tools" has written: "The manipulation of form is,
out indicating a concrete function. Models were to then, a primary aspect of architecture, and one of the
be used at all stages of the search for composition main ways in which discipline is engaged. It is not
and in the final representation of the task. Thus, the the goal, however, but a means to an end." (Plowright
foundation of a new method of teaching architec- 2014, 18.)
tural design, which had not been known to the aca- Our course of architectural composition was in-
demic schools previously, was laid out. The princi- troduced to the curriculum in 1987. It elaborated
ples of this method of teaching were formulated as on the basis of Bauhaus and VKHUTEMAS. In this
follows: study of composition means and rules of cre- traditional approach to the architectural composi-
ating spatial three-dimensional forms for solving ar- tions teaching cardboard models were used. At
chitectural compositional design tasks; development the beginning our course of architectural composi-
of spatial thinking and compositional skills; master- tion consisted of three abstract composition exer-
ing the skill of masking models as one of means used cises (façade, solid form, linear composition (walk-
during the process of architectural designing. The through the space)). According to the VKHUTEMAS
entire theoretical and methodological material was rules, a composition does not have to be similar to
presented in 1934 by V.F Krinsky, I.V. Lamtsov and anything that exists in the real world, but it should
M.A. Turkus in the book "Elements of architectural be approached as an individual singular action that
and spatial compositions". completes itself. For each exercise we have elabo-
The teaching structure at the Bauhaus was devel- rated a short description including general remarks
oped by Walter Gropius in 1922. The central part of it on the activity and its goal.
was the Building. At the start of the studies, students
attended what was called a "preliminary course", in Façade
which they were experimenting with colour, shape The goal is to create a relief composition - a façade
and materials with no specific goals. This approach of an object of unspecified purpose. This façade
was developed by Gropius, who also chose the teach- will only be viewed from the front. The composition
ers which was the most influential for the develop- should be arranged on a 30x30 cm field. The viewer
ment of the Bauhaus's viewpoints. Johannes Itten is on horizontal 30x30 cm surface. Perception point
was the first person to run the preliminary course in should be marked with a human shape, as its position
which he had broken the academic tradition by ask- determines the size of the façade. The façade may
ing the students to create spatial forms based on their be built of different types of cuboids. Their size and
own subjective perceptions instead of copying from number depends on the conception. The elements
models. [1] may be arranged within the plane in any way. The
only limit concerns the depth of the composition, i.e.
TRADITIONAL ARCHITECTURAL COMPO- the distance from the furthest element to the compo-
sition plane, which cannot exceed 5 cm. The model
SITION
should be monochromatic.
"Elements of spatial forms under appropriate combi-
nations, relationships, proportions and rhythms are
Solid form with specific formal value
a means of expression of the content of the architec-
The aim of the design is an exploration of a 3-
tural object, including its artistic qualities. In the anal-
dimensional space and 3-dimensional volumes in
ysis of architectural forms, the use of the method of
space, which is one of the most basic and funda-
196 | eCAADe 34 - CAAD EDUCATION | Design Concepts & Strategies - Volume 1

Figure 1
Examples of
traditional
mock-ups: A –
Façade, B – Solid
form, C – Walk
through.

mental architectural activities. Students should de- board mock-ups. At this stage we observed that
sign a solid composition with a specific emotional their manual skills were totally insufficient. Mock-ups
value (dynamic, light, massive, monumental). The were inaccurate and many students claimed that re-
composition size should not exceed 30x30x30 cm. sults looked different than they expected. Due to a
The composition will be viewed from different per- lack of skills students did not want to develop de-
ception points located around the object. A point sign variants as they were too time consuming. In the
from which, according to the designer, the composi- case of the next exercise (Solid form) the beginning
tion appears most effective should be specified. Any of the design process was the same, but the mock-
cuboid solid can be used to build the form. Their size up quality has improved significantly. However, be-
and number depend on the conception. To make the cause the compositions were more complex (they in-
evaluation of the object scale, a human 1,7 cm fig- cluded emotional values), the students had trouble
ure should be placed on horizontal plane. The model to imagine the real emotional impact of the composi-
should be monochromatic. tion. This aspect also evoked a biggest discussion be-
tween teachers during the assessment of the works.
Linear composition - Passing through an Often, it was difficult to assess whether the designs
open space presented by students using scale mock-ups would
The goal is to create a Passage through something actually be, for example, monumental or massive. In
important, something that has no beginning or the last project - Go through - the evaluation has be-
end. A passage 'Through' and not a way 'From- come even more difficult, as it was required to imag-
To'. Students should anticipate the interlinks be- ine the appearance of the space, which cannot will be
tween present, past and future impressions; produce viewed.
a space in which things happen and different moods
are created. When applying architectural forms, a DIGITAL ARCHITECTURAL COMPOSITION
linear composition should be designed within a 0,3- In 1987 R. Schijf (1987) has written that the de-
0,6 km space. The composition will have different sign studio took up the central place in architectural
emotional impact. Any geometric forms and various teaching and it should be determined to introduce
materials (cardboard, glass, mirrors or metal) can be CAAD into the design studio. It demands that de-
used. It is advisable to apply colour. The scale of the sign teachers should know of the potential and the
model should be 1:100 on a 30x60 cm board. limitations of available computer facilities. At our
The starting point of designing was searching school, we introduced digital media to the Architec-
for the inspiration and preparing the sketches of the tural Composition course for the first time in 1997,
idea. Sketches were discussed and on their basis the after many years of experience with teaching tradi-
model of the composition was realized. Also, 'notional architectural composition. When we decided
function' (the traditional meaning) spaces were cre- to use the new digital media for modelling architec-
ated. The only function of space was evoking emo- tural forms, we wanted to investigate the new possi-
tions. Forms became more poetic and metaphoric. bilities of form creation. The digital course was based
(Figure 1) on the same exercises as the traditional course. After
five years of experiments, in 2002, thanks to the in-
Issues tensive usage of computer modelling techniques the
The design process of the façade was started by entire course was elaborated. This course included
searching for inspirations, which may help the stu- five groups of exercises: bas-relief (division, rhythm,
dents understand the design problem. As a medium façade); solid composition with specific formal; trans-
for searching all students used a sketch. After con- formation - from a cube to a parallelepiped; walk
sultation with teacher they started preparing card-

Figure 2
Examples of digital
compositions: A –
Façade, B – Solid
form, C – Walk
through.

through (desert, valley, and tunnel); walk through Issues
an internal space (space of celebration, contempla- Analysis of the possibilities of using computers for
tion and dynamic space). We have added informa- architectural compositions teaching has shown that
tion about required skills and software to the exer- using new digital media for modelling architectural
cise descriptions. In this course we permitted using forms gives new possibilities but at the same time
any possible geometrical elements and surfaces with creates new issues. The first is that little knowledge of
different colour and light. Practically in all exercises software capabilities at the beginning of the learning
an animation was the obvious element. (Asanowicz process made it difficult for students to achieve the
2003.) design goal. The problem that teachers faced were
The process of designing in both Traditional and the difficulties in the evaluation of complex space
Digital Course proceeded in the same way. The start- compositions as it was really tough to imagine how
ing point was searching for the inspiration. Each the form will look in reality, especially in the case of
student presented photos of existing architectural the last exercise - walk through linear space.
objects and a text, which explained the reasons of Usually, architects use sketches at the concep-
the choice. Next obvious stage was preparing the tual design stage. In our work the architect's tradi-
sketches of the idea. Sketches were discussed and tional tools such as a pencil, a sketch block and phys-
on their basis the model of the composition was real- ical models are effectively replaced by a computer,
ized. As a result, forms designed by students in digital which creates a new way of doing things. For ex-
environment were similar to traditional carton mod- ample, students applied various materials of differ-
els. The similarity concerned the degree of façade ent transparency and many projects concentrated on
complexity. At the same time, we observed that stu- playing with light and not form. As mentioned above,
dents liked to design variants of the form. In these students created animation which required prepar-
variants they tried to use different sources of light for ing some kind of a scenario, determining points of
achieving different emotions during process of per- perception, and defining which places and spaces
ception. were important. Thereby, the perception process
We observed larger differences in the process becomes dynamic and the onlooker's emotional en-
of creating a solid form. Forms became more com- gagement increases. It is possible now to convey the
plicated: curvilinear planes, surprising transitions emotional message of the designed spaces more pre-
among particular elements appeared, and light was cisely. This helps to better understand the relations
used in more conscious way. A simple computer ani- between composition elements as well as their influ-
mation allowed for a dynamic search of the best per- ence on emotions. At the same time, in early design
ception points and a better understanding of the vi- stages, when formal value is sought, computer mod-
sual frame. At the same time, we observed that better elling which can be done almost intuitively. Simple
knowledge of the software allowed students to cre- operations with basic forms enable students to de-
ate more and more complex compositions and many sign architectural sculptures, forms with specific for-
students preferred to design directly in the digital mal value. The level of freedom in designing space by
space without the use of a pencil. They claimed that computer methods differed substantially from tradi-
they were only able to present their project verbally. tional models. Most of the students claimed that the
It made the teacher and student co-operation much process of decision making was much more flexible
more difficult, since the idea traditionally accepted in and as a result more effective.
the graphic design convention was broken. (Figure 2)

Figure 3
From traditional
mock-up to Virtual
Reality
presentation.

DIGITAL ARCHITECTURAL COMPOSITION cal images, the reality of the illusion, and above all the
IN VIRTUAL SPACE perception of the invisible aspects of our world. The
Perception of a traditional composition was limited development of a virtual reality in which interacting
by the insufficiency of our imagination. It is ex- with representations of virtual objects can take place,
tremely difficult to imagine the space inside a mock- allows the use of new perceptual, cognitive and inter-
up, and then evaluate the correctness of the solu- active capabilities of man.
tions. Even great architects such as K. Melnikov con- Students may check the design solutions and de-
sidered that even the author of the project cannot cide whether the planned emotional values were re-
predict all. He claimed that what the author of a flected in the created space. The same also applies to
project sees and feels cannot be transferred into tra- the teacher who does not have to visualise what the
ditional means of presentation. The author's idea can student imagined designing a space. The evaluation
be understood up to a point, only to be perceived process becomes more objective, insofar as we can
completely in the completed building. His dream objectively assess the emotional values. (Figure 3)
was to see how people will perceive the unusual
form of internal space created by him in his design Issues
of a Columbus monument. It was a trimmed cone, This approach demands additional time for teaching
into which another similar but rotated cone was im- new software, which is fortunately not difficult. The
mersed at the top. K. Melnikov admitted that even he problem is rather in hardware part, as this way of pre-
could not have had imagined what impression this sentation requires powerful computers. The system
interior would have on others (Khan-Magomedov requirements were published on the Oculus web-
1990). site. The recommended settings for using the Rift
In our Digital Architectural Composition Course are as follows: NVIDIA GTX 970 / AMD 290 equivalent
we decided to use the Oculus as an evaluation tool. or greater, Intel i5-4590 equivalent or greater, 8GB+
As D. Gann has written in the preface to the J. Whyte RAM, Compatible HDMI 1.3 video output, 2x USB 3.0
book "Virtual Reality and the Built Environment": ports, Windows 7 SP1 or newer. Atman Binstock, the
"They make it possible to create virtual prototypes, to technical director for the Rift, explained why the re-
model attributes and to simulate performance char- quirements are as high and has written: "On the raw
acteristics without having to build full-scale mock- rendering costs: a traditional 1080p game at 60Hz re-
ups. By adding another dimension to the ways in quires 124 million shaded pixels per second. In con-
which space can be configured over time, they com- trast, the Rift runs at 2160×1200 at 90Hz split over
plement and enhance the value of using face-to-face dual displays, consuming 233 million pixels per sec-
communications and physical models." (Whyte 2002, ond. At the default eye-target scale, the Rift's render-
7.) ing requirements go much higher: around 400 mil-
Oculus allows one to enter through the monitor lion shaded pixels per second. This means that by raw
screen into "the computer world", where an interac- rendering costs alone, a Virtua; Reality game will re-
tion with the forms present there is taking place. It quire approximately 3x the GPU power of 1080p ren-
turns out that apart from these virtual forms, nothing dering." [2]
else exists. Designers and users of this virtual world
receive new experiences of space and new means DIRECT DESIGN
of interaction. Virtual Reality eliminated the bound- Over the last years, there was a discussion about the
aries between the spectator and the space. This cre- possibility of using Virtual Reality technology as a
ates the conditions for the perception of the meta- design environment where design actions would be
morphosis of forms, non-physical objects, paradoxi- possible. This idea was based on a full immersion of

the architect in the environment projected by them. S/WA/1/16).
Creation and visualization of design solutions follows
directly in virtual space. The architect is within the REFERENCES
projected space, defines a direction of changes and Asanowicz, A 2003 'Architectural Composition in Digital
in an interactive mode realizes these changes, mov- Space', Proceedings of ECAADE 2003, Graz, pp. 587-
ing forms in virtual space. Unfortunately, so far no 590
software offers real-time interaction. Preparing mod- Droste, M 2006, Bauhaus 1919-1933, Taschen
els for CAVE or virtual goggles requires a lot of time Khan-Magomedov, SO 1990, Komstantin Melnikov, Stroy-
izdat, Moscow
and this Virtual Reality model can be used for evalu-
Krinsky, VF et al 1934, Elements of architectural and spa-
ation only. tial compositions, GNTI, Moscow
Matz, I (eds) 1933, Sovetskoe Iskusstwo za 15 let (Soviet Art
CONCLUSION over 15 years), Izogiz, Moscow
Plowright, PD 2014, Revealing Architectural Design. Meth-
Correct learning by students of compositional cat-
ods, Frameworks and Tools, Routledge, NY
egories, formation of creative skills, and in general Schijf, R 1988 'Strategies for CAAD education - the Sin-
mastering basic methods and ways of architectural gapore way', Proceedings CAAD Futures 1988, Eind-
design process is of crucial importance for the devel- hoven, pp. 23-46
opment of their creative abilities. Whyte, J 2002, Virtual Reality and the Built Environment,
Our experience showed that the implementation Architectural Press
[1] http://www.bauhaus.de/en/impressum
of the new computer technologies is very promising.
[2] https://www.oculus.com/en-us/blog/powering-the-
What is more, students prefer to work in a digital en- rift/
vironment. When designing a façade, 17 out of 30
students begun work from a sketch, in Solid form de-
sign 5 students. In the final exercise, only 2 students
were using a pencil as a tool, while 28 designed di-
rectly using a computer. In Virtual Reality students
experienced a better understanding of the scale of
the composition than in traditional manual or digi-
tal modelling. Even a simple model without texture
and shadows provides more information about the
designed space. The only obstacle to the effective
perception seems to be a problem with navigation in
virtual space.
This does not mean that traditional models are
useless. Mock-up is one of the primary ways of devel-
oping spatial imagination, avoiding mistakes when
adjusting the scale of the proposed forms to the size
of the environment and avoiding mistakes or illogi-
cal/impossible to construct spatial forms. Based on
our 20 years of experience, we can conclude that the
best pedagogical effect is achieved by the parallel
use of both techniques.
This research was a part of grant financed by Min-
istry of Science and Higher Education (grant number

Flying Bricks
Algorithmic Design Studio
Wolfgang E. Lorenz1 , Gabriel Wurzer2

1,2
Vienna University of Technology | Institute of Architectural Sciences | E259.1
Digital Architecture and Planning
1,2
{lorenz|wurzer}@iemar.tuwien.ac.at
The design studio 'Flying Bricks' was held during the summer semester 2015. Its
main objective was to redesign an existing building with the use of facing bricks
algorithmically, utilizing algorithmic thinking and programming as a means for
form generation. The purpose of having students express their designs in terms of
code was to emphasize problem thinking over solution generation, which has
several advantages but also disadvantages which we would like to share in this
paper. Furthermore, we would like to show how our implementation process
worked, so that others can leverage that for their own algorithmic design courses.
Keywords: NetLogo, Digital Design, Bricks, Education
MOTIVATION stantly think about possible implementation of their

In this paper we discuss our experience of teaching ideas with the help of algorithmic methods which we
algorithmic thinking during a design studio using a taught them during the design studio. The actual
multi-agent simulation software. Moreover, we will programming work happened in NetLogo (Wilensky
have a look on how to promote problem thinking 1999), a multi-agent simulation which offers a rela-
over solution generation. tively "easy" scripting language plus built-in 3D visu-
'Flying bricks - algorithmic design studio' was alization. Most students had no background in pro-
held in the summer semester 2015. The objectives gramming, and thus we also gave an introductory
of our course were threefold: workshop on that subject beforehand.
The course was accompanied by the question
1. We wanted our students to digitally design fa- how to promote problem thinking over solution gen-
cades (or facade elements), eration, which is still underrepresented in architec-
2. to learning algorithmic thinking and tural practice. To illustrate that point further, we
3. to implement their ideas in form of a program wish quote Lawson (2005) who did an experiment
(script). in which students from architecture had to compete
against postgraduate science students. Their task
was to arrange Tetris-like blocks according to some
In a first step, students had to choose an exist- "hidden" rules which were only known to a computer
ing building which would have then to be redesigned which they could ask whether their proposed solu-
with facing bricks. In that process, they had to con-

tion was valid. In doing so, they showed strikingly studied. In that context, the students were faced with
different and consistent strategies: two questions:
"[...] while the scientists focused their atten-
tion on understanding the underlying rules, the ar- 1. How well would algorithms provide them
chitects were obsessed with achieving the desired re- with new perspectives on the problem and
sult. Thus we might describe the scientists as having a 2. to what extend would they be able to gen-
problem-focused strategy and the architects as having erate alternative solutions with algorithmic
a solution-focused strategy." (Lawson 2005, 43) help, in the same fashion that they had done
Thus, we can summarize that the problem which by hand in their previous design studios.
our algorithmic design course tackled was to shift the
students' attention - at the very least to an extent - As they had to find out, there is no overall answer to
from the solution space to the problem space. these questions. Algorithmic thinking means, above
all, to break down the design process into separately
EDUCATIONAL METHOD solvable subareas (Alexander et al. 1977; Alexander
In the first phase, students analyzed historic build- 1979). These areas included the overall form on a
ings and innovative brick systems in order to find broad level, colour schemes on a mid-level and the
new creative application areas. The design of fac- brick format itself on a low level.
ing brick facades is a complex issue in architecture. The third phase was occupied with the final de-
While facing bricks are still common and an integral signs. It required exporting the outputs of the stu-
part of architectural form in many countries of the Eu- dents' programs for further elaboration in a CAAD
ropean Union (compare The Netherlands, Belgium, (Computer Aided Architectural Design) tool, plus ren-
Germany, Italy), application in our own country (Aus- dering in a 3D modelling platform. All of these ma-
tria) has continuously decreased since the epoch of terials were then combined in presentable format
historicism. Such a development has been assisted (slideshow, poster, short paper) in a final step.
by the introduction of Exterior Insulation and Finish-
ing System (EIFS). The success of EIFS lies in cheaper Algorithmic design
constructions, light weight and thinner walls that re- We made the experience that from the very begin-
sult in an increase of space (and thus in an increase ning of the design process students tends to have an
of revenues). Thus, the first phase of the design stu- almost finished picture - or at least a sketch - of their
dio was characterized by the understanding of the design ready in mind. This picture may change dur-
brick format, its restrictions and the requirements on ing the process of analysis (location and task) and/or
a design regardless of actual implementation. Stu- design; however, an interim result is present. Our de-
dents analyzed important impacts on the algorithm sign studio 'Flying Bricks' focused on the problem it-
and on the design, such as the neighbouring build- self rather than on the result. The problem was, so to
ing, the shape of the building itself, the proportion of speak, within the design influencing factors (param-
the facade, the place and the orientation. "Wiener- eters). Thus, rules had to be defined that solve the
berger AG", a global manufacturer of bricks, acted problem and influence the result.
as our industrial partner. They let us visit their fac- The initial problem, to re-design an existing fa-
tory for clinker and supported us with information cade with facing bricks, was, first of all, divided into
about the technical requirements of brick construc- solvable sub-problems. In order to bring these sub-
tion from the very beginning. problems to the surface, the main focus was placed
In the second phase, algorithmic methods that on certain influence factors. 'Site analysis' by Edward
would support the students in their design were T. White (1983) provides an insight into the numer-
ous factors that influence architectural design. Some

of them were incorporated or expanded by the stu- 1. To generate an optimum; this includes less
dents. The existing building itself was recognized as material consumption or the claim for the
initial group of influences, including its geometry (cu- smallest number of different shaped bricks;
bic or free form), its proportion, its rhythm (sequence e.g. the Johnson Wax Building by Frank Lloyd
of openings) and the relation between open and Wright uses over 200 different shapes of bricks
closed spaces. The neighbouring buildings defined to form the angles and curves of the design
another group of influences, including the position (Levy 2004), and
of the cornice, zoning (base zone and roof zone), pro- 2. to generate a set of possibilities, from which
portion and size of openings (windows and doors). the "best" solution can be selected.
But also public space served as inspiration, such as
We hoped to find new aspects, that otherwise would
the course of the street. Location and orientation
not have been discovered or only with increased per-
were also considered as an important group of influ-
sonal effort. In summery the design process included
ences, including shadowing, protection from street-
several phases:
noise, wind, permission of outlook and insight, dis-
tance to the street or position above ground level. • Drawing of the existing facade in a CAAD pro-
Finally, the material used comprised another gram and obj-export,
group of influences with its own problems (and their • import of the obj-file and voxelization in Net-
solutions); e.g. the small scale of bricks and the re- Logo,
sultant specific need of static load-related arrange- • application of the own script to laying bricks,
ment. Bricks laid bonded in a structure not only in- • export of the result as obj-file and import in a
creases stability and strength but also leads to a var- CAAD program and
ied picture; furthermore, small scale provides flex- • further work on the result.
ibility of geometry. In short, size was discovered
to allow different possibilities of brick bonding that
ask for certain rules (minimum overlapping of bricks IMPLEMENTATION METHOD
for stability). The section about bricks was enlarged Learning from Lawson (2005) and White (1983) two
by colour psychology (different possibilities of bricks questions arises: What if the mentioned influences
colour), the rain security of bricks that makes further are stored as values inside a three-dimensional grid?
weather protection unnecessary (protection from en- And what if this stored information influences the
vironmental influences, such as sun, wind, rain and way of how an object evolves? This kind of, so to say,
their impact on geometry of the composition) and information driven design has successfully imple-
the extraordinary durability of bricks that provides mented during a workshop in Bialystok (Wurzer and
benefits in the preservation of buildings. Lorenz 2014). The advantage of an agent based pro-
In summary, the how and why was already gramming resides in the possibility to control agents
present in the approach of solving the problem, (entities) externally or internally. External controls
which influenced the final form. Therefore the def- comprise the information stored in the grid. This sup-
inition of design influences was most important. In ports the idea of problem thinking solution. The so-
short, problem solving was considered to take up in- lution is inherent in the stored information.
fluences to guide towards the solution. This yields
two possibilities for a program:

Agent Based Programming Writing a program encompasses two steps:
Agent based programming offers the advantage that
1. Setting up the world (assigning initial colours
results of a simulation can directly provide the infor-
to voxels, creating agents) and
mation comprised in the model (e.g. simulation of
2. calling a step function continuously, until
wind, the sun's course and noise dispersion). The re-
some stopping criterion is satisfied.
sults of these simulations are immediately available
to influence e.g. the path of agents that form a spe- The step function is occupied with moving agents
cific design, leave traces, or, as in our case, simply through voxel space, and letting them alter their en-
place bricks. Agents are, so to speak, seekers, driven vironment as they pass through (e.g. laying a brick).
by the places they pass. One could almost say, agent In that piece of the program, they also need to make
based programming provides an "all in one solution". use of sensory capabilities - i.e. "What is the colour
The key issue of our model is how and why of the voxel ahead", in order to decide what action to
agents move and who controls or influences the perform.
agent on his way from a starting point to its goal. In Our design studio was occupied with redesign of
the case that the agent does not know its path per se, existent building using bricks. Thus, we had to some-
we speak of external control. A well defined playing how load these into NetLogo during the setup phase,
field (the outer wall of certain dimensions) provides in order to have a "substrate" along the facade where
such a restriction. The agent may be free to walk (ran- agents could do their (brick-laying) job. In more de-
domly) but is thrown back before passing the border tail, we imported the 3D geometry as faces made out
line. In an advanced model the current location tells of three vertices, which we then subdivided until they
the agent where to move next, as in a pinball ma- were of voxel size (see left in Figure 1). Then, we sim-
chine. Relevant information that affects the path in- ply asked each voxel to colour itself if it contained a
cludes differentiation between solids and openings, vertex. The results of that process is shown right in
distances to specific points, heights above ground or Figure 1 (for more details, refer to Wurzer 2014).
shadowing by other buildings.
Figure 1
NetLogo Programming Principle of
Since NetLogo is a multi agent programming lan- voxelization (left)
guage with integrated modelling environment, it and result (right).
served as a basis for all of our programming work.
The tool is more than a programming language, since
it offers several useful metaphors that are already in-
built, ready to be used when implementing a design:
Voxel space. NetLogo's virtual world is a 3D reg-
ular grid made of voxels (i.e. cubes), each having an
extensible set of properties (e.g. colour). It was now the agents' job to lay bricks layer by layer,
Agents can move freely within voxel space. Each moving from one coloured voxel to the next coloured
agent also carries an extensible set of properties (e.g. voxel. In order to get into even more detail, we wish
position, heading shape, colour and type; the latter is to briefly present some examples that show how this
useful for defining different 'flavours' of agents, such was done (see Lorenz and Wurzer 2015 for an in-
as builder, seeker and so on). depth explanation).

DISCUSSION OF OUTCOMES set automatically and the custom property "offset" of
The playing field, as described before, was defined the agent is set to "right". At each step, the agent de-
by the outer walls of a specifically selected building. termines whether its offset is "right" or "left". In case it
In order to import a wall into the world of NetLogo, is "right", it tries to place a brick two centimetres to its
two possible ways exists: The wall can be discrete or right (opposite goes for "left"). If this position is not
continuous. The discrete world, a three-dimensional coloured, then the agent is beyond the facade and
grid, is preferable, since a division into voxels offers may not place a brick there. Thus, the offset is set to
the possibility of additional storage of information in its opposite value (left becomes right, right becomes
each voxel (which, in turn, influences the path of the left) and the process starts again. If this is not possi-
agent or properties of the brick). ble either, then the program stops. In all other cases,
Our participants used NetLogo's possibilities for the agent moves one step along the voxelized wall,
agent-based programming in different ways. Some adjusting its heading if it is necessary to turn (e.g. at
used voxels to store information, which then influ- corners). If, after taking a step, it reaches a voxel it has
enced agent movement along the facade and/or already visited, it changes its position to one level up.
properties of the built bricks. Other students simply If there is no coloured voxel there, the agent is be-
set bricks in a given wall without reacting to context. yond the facade and stops the program. In all other
Some projects even went as far as to incorporate ac- cases, the step function is repeated (see Figure 2a for
tual simulation within their design. We are now going a result).
to present a few examples in more detail. Viewpoints: The basic idea of this program was
Simply set bricks: This first program served as to create a reference oriented design that reflects cer-
inspiration. The design emerges exclusively out of tain viewpoints in a specific combination of headers
the algorithm itself. The basic idea is to simply offset and stretchers. At the beginning, the user defines
bricks in a stretcher bond along a given orthogonal specific viewpoints from outside the building. The
outer wall of a building. The particular challenge lies algorithm then automatically detects four main cor-
in well defined corners with bricks exactly positioned ners of the building (referenced diagonal to the four
one above the other (without any offsets). Therefore, corners of the NetLogo world). Initially, the facade is
first of all, a "spy" agent analyses the geometry of the covered with a stretcher bond. Subsequently, previ-
outer wall. Later, all bricks are arranged smoothly be- ously defined areas (e.g. above/below windows, lay-
tween the fixed corner bricks (let's call them "refer- ers with/without openings and n-th layers) are specif-
ence" bricks). The "spy" agent also checks if a full brick ically handled: Some stretchers are replaced by ro-
or a half brick can be laid at corners and at the edges tated headers and offset outward (leading to a pic-
of openings. The starting position of a single agent is ture similar to a course of diagonal bricks; Figure
Figure 2
‘Simply set bricks’
(a), ‘Viewpoints’
with rotation of
bricks (b) and 'Light
and shadows' (c).

2b). The replacing of bricks starts with a stretcher Simulation as input: The reasoning underlying
followed by a rotated header. The sequence contin- this approach was the facade's function to protect. In
ues with a stretcher and three headers, two stretchers more detail, areas of higher pollution should be re-
and five headers, four stretchers and seven headers flected by darker colour; furthermore, colour gradu-
and vica versa. The special feature is that all headers ation and different offsets should provide more va-
are oriented to one of the given viewpoints, which riety. A simulation of passing cars, leading to pollu-
were deemed to be "important". This means that the tion, serves as input for the design. The density is
design exclusively depends on the geometry of the added up and stored within each voxel of the facade
building (including openings), the viewpoints and (together with the maximal distance to the source of
the given sequence of rotated headers and straight noise). Agents (brick layers) ask for this information
stretchers. In other words, the layout is exclusively and reacts by altering their colour and their offsets
defined through the rules defined at the very begin- (Figure 3a-3b). More precisely, colour, on the one
ning. hand, depends on the maximal distance to the source
Light and shadows: This program based on the of noise, classified into three colour spectra. Offset,
idea to map light and shadow on facades depending on the other hand, depends on intensity, again, clas-
on orientation (and the surrounding environment). sified into three ranges. In general, displacement is
For this purpose direction of light and intensity is restricted to two centimetres to keep the maximal
calculated and stored in each voxel. The orienta- difference between two layers below four centime-
tion of the facade, the height in relation to the street tres (recommended by the manufacturer). In this ex-
level and the shading by neighbouring buildings in- ample NetLogo was considered to confer a large ben-
fluences the degree of offset (from one brick to the efit in terms of simulation, but also in terms of the
next). In other words, roughness depends on expo- possibility to export colour information.
sure. For the sake of simplification, sun-intensities The urge to turn: The basic idea was to cover
are summarized in definite ranges of offset. The off- the facade with continuously rotated bricks. Bricks
set of each brick operates like described in "Simply at corners of the building and of openings serve as
set bricks" with a maximum displacement of six cen- fixed points oriented in 45 degrees to the outside.
timetres (Figure 2c). This example demonstrates how All bricks between two corners change their orien-
simulation (of light) and design can take place in one tation in order to complete a rotation of 90 degrees.
and the same program. However, technical problems The bond's front view displays a continuous develop-
prevent current realization. Apart from that, the vul- ment from a rather "smooth" appearance, with the
nerability of the algorithm mainly occurs at corners only structure given by bed and head joints, to a
with a non-right angle due to imprecise transitions. rougher appearance of strongly protruding bricks at
Figure 3
‘Simulation as
input’ with noise
simulation (a) and
different colouring
of bricks (b), 'The
urge to turn' with
continuously
rotated bricks (c
and d).

the corners. Despite rotation, bricks of the next layer Chameleon: The basic idea was to take up the
have to centrically cover head joints of the layer be- bonds of surrounding environment and to colour
low. In a first step, a so-called "spy" agent walks along and offset bricks depending on their orientation. The
the facade in order to capture the geometry. This more a facade is oriented to the sun, the lighter the
includes to identify fix points (corners and edges of colour of the brick and the larger the offset (to in-
openings), to calculate the orientation of the angle crease self-shadowing). Both, calculation of the ori-
bisector of fix points and to identify successive levels entation and the setting of a simple header bond
with abrupt changes of orientation. The latter occurs are done automatically. Subsequently, single bricks
above and below edges of openings, when a layer of are offset depending on the orientation of the fa-
continuous rotation between corners meets a layer cade (Figure 4b). Bricks at the corner and the edge
with many rotating areas between openings. More- of openings are not offset.
over, the anchorage is carried out by horizontal con-
soles with straight stretcher bonds. These new layers DISCUSSION
influence again the rotation of the brick layers above During the course, scale differences between NetL-
and below. Again, all the information is stored in the ogo and CAAD programs turned out to cause the
voxels and the agent is guided by this information. main difficulties. While the latter have a well-defined
The algorithm is designed for rectangular buildings longitudinal scale (e.g. meter or centimetre), the
only. It is an example that reminds us of design intent scale of NetLogo is defined by the number of voxels
(which does not always need context; Figure 3c-3d). in x-, y- and z-direction. This is bound up with the fact
Brick patterns with bonds: An intensive exam- that voxelization entails the disadvantage of simplifi-
ination of different bonds led to the idea of this pro- cation. The original shape is transformed into boxes
gram. Patterns or pictures should be projected on of a three-dimensional grid with the resolution de-
facades, visualized by different structures. The tem- fined by the box size. The sizes of the outer wall and
plate might derive from an image or from the struc- of openings may be distorted (e.g. windows of the
ture of other facades. Both, the information about same size may turn into openings of different size).
the section of wall (opening or solid) and about the To increase resolution of the grid means to reduce
pattern are stored in the voxel. The agent just reacts the size of the boxes. However, this comprises per-
on the information it gets from the current position. formance. During the import process parts of the
A special challenge is the transition between differ- wall can be imported separately, e.g. parts of the
ent areas of specific bonds. They may not be abrupt wall, windows and doors. Each part can be associ-
but continuously. This is solved by creating new in- ated with a different colour of the voxel and a dif-
termediary bonds (Figure 4a). Additional feature con- ferent type (wall or opening). Since overlapping of
cerns colouring and/or offset of headers.
Figure 4
‘Brick patterns with
bonds’: transition
between bonds (a),
'Chameleon' (b).

these elements is possible, the order of importation ACKNOWLEDGEMENT
into NetLogo is important. We thank Mr. Staudinger and his team from Wiener-
The key benefit of NetLogo is the simulation- berger AG who provided insight and expertise. Fur-
based programming itself, since it can be used di- thermore, we would like to thank the following stu-
rectly as source of information (simulation of noise dents, since their contribution is part of this paper: B.
distribution). Moreover, agents (the placement of Arendt, M.A. Schiltz, A. Malhotra, B. Hanser, B. Blank-
bricks) are directly driven by forces contained in vox- Landeshammer and N. Jasinski.
els (a certain place) and not by the outside. This
opens up possibilities for future research. REFERENCES
Concerning problem thinking over solution gen- Alexander, Ch. 1979, The Timeless Way of Building, Oxford
eration it turned out that in most cases students still University Press
have certain results in mind. Sometimes the algo- Alexander, Ch., Ishikawa, S. and Silverstein, M. 1977,
rithm was even developed to fit this idea. What was A Pattern Language: Towns, Buildings, Construction,
helpful, on the one hand, was to think about certain Oxford University Press
Lawson, B. 2005, How designer think: the design process
influences and to implement their impact on the de-
demystified, Architectural Press, Burlington
sign. To store information in voxels that lead agents Levy, H.H. 2004, Artists and Architects (Famous Wisconsin),
and subsequently set bricks was, on the other hand, Badger Books Inc
the main aha-experience. We will focus on this aspect Lorenz, W.E. and Wurzer, G. 2015, 'Move in and out', in
in our future work. Lorenz, W.E., Wurzer, G. and Swoboda, S. (eds) 2015,
Flying bricks: algorithmisches Entwerfen, Digitale Ar-
chitektur und Raumplanung, Vienna, p. 31–38
CONCLUSION White, E.T. 1983, Site Analysis, Architectural Media Ltd.
In this paper, we have presented our algorithmic de- Wilensky, U. 1999, NetLogo, ccl.northwestern.edu/netlogo/,
sign studio 'Flying Bricks' which had the purpose of Center for Connected Learning and Computer-Based
teaching students algorithmic thinking and problem- Modeling, Northwestern University, Evanston, IL.
Wurzer, G. 2014, 'Loading Geometry into NetLogo3D', Re-
oriented thinking through agent-based program-
port for Vienna UT, Digital Architecture and Planning,
ming. The context of the design studio was build- Report No. 1, pp. 1-5
ing with bricks, supported by our industry partner Wurzer, G. and Lorenz, W.E. 2014, ''Multi-Agent From
'Wienerberger AG' who advised on practical ques- Simulation', Archivolta, 62(2), p. 57
tions in brick construction. Results obtained by our
students show that such an approach is feasible even
when starting from scratch, both programming-wise
as well as in brick technology. We are convinced
that similar lecture formats which let students as-
semble "simple elements" to form "complex shapes"
are didactically advantageous and guarantee a steep
learning curve.

Magnet-based Interactive Kinetic Bricks
Mohammad Kouchaki1 , Mohammadjavad Mahdavinejad2 , Parastoo Zali3 ,
Shahab Ahmadi4
1,2
Faculty of Art and Architecture, Tarbiat Modares University 3 Faculty of Art and
Architecture, Imam Khomeini International University 4 Faculty of Art and Archi-
tecture, Iran University of Science and Technology
1,2
{m.kouchaki|mahdavinejad}@modares.ac.ir
3,4
{paras2.zali|shahab.ahmadi.architect}@gmail.com
Brick has been used in construction since ancient times and has been respected
among other tectonic materials through out the history. Novel technologies
recently have opened new horizons in using brick in architectural design. This
paper investigates innovative implementation of bricks in kinetic architecture.
Kinetic structures usually employ complex and high-cost mechanisms to come
into force and their movements might be limited to some conditions. By the use of
magnet in digital design, this research examines new methods for performing
simple and affordable kinetic structures so as to create interactive relations
between architecture and human being. Magnetic energy is applied in two ways
to move a roof made of brick which is considered a heavy and masonry material.
Consequently, it represents the hidden potentials of magnet as a renewable source
of energy.
Keywords: kinetic architecture, interactive design, parametric design,

Bricklaying, magnet energy
INTRODUCTION to examine their ideas and innovations through de-

Nowadays explosive growth of digital technology signing the canopy. The coherence between theo-
has affected the core of architectural design. Cer- retical ideas and design to performance process was
tainly, digital tools offer new possibilities that were the main issue of this experience. The design team
inconceivable only a few years ago (Picon 2010). This did their best to combine endogenous ideas as well
research arises from a project in design studio of Tar- as indigenous materials with digital technologies in
biat Modares University (TMU) defined as "Interactive order to find a way to design a kinetic structure with
Transforming Canopy with Particular Reference to heavy and masonry materials like brick. It was an im-
Computational Design Thinking". The canopy seems pressive task to bridge local materials (such as brick)
to be a pavilion to exhibit cutting-edge technologies to new shape of interactive atmosphere which was
which developed by TMU students, understanding asked in the design terms of the canopy.
that, the canopy itself should show high level of inter-
activity. As a result, the design team were challenged

INSPIRING BRICKWORKS Cloaked in Bricks
Brick construction counts with a long and rich tradi- Cloaked in Bricks is a residential project designed by
tion in architecture, which can be traced back to the Admun Design & Construction Studio and located in
origins of our civilization and reminds us of soil and Ekbatan, Tehran, Iran. It consisted of a façade design
nature. Because of accessibility and unique features, and lasted from 2013 to 2015. This project is an at-
it has been employed widely in traditional architec- tempt to propose a prototype addressing the current
ture with variety of forms in each place. It was used in issues of residential architecture in its region through
different parts of a building as structure or decoration bridging between old and new, proving how local
in facade. Amazing samples of brick works around materials and patterns can be used in new ways cre-
the world exemplifies our ancestors' progresses in ating an architecture responding to both functional
producing and employing this material. As an exam- requirements and aesthetics.
ple, among Iranian ancient buildings, Zanbil Ziggu-
rat, porch Ctesiphon, Isfahan and Yazd Grand Mosque
Figure 1
and Rabat (Museum of brick work) can be mentioned.
The facade design
Despite the long history of bricks and masonry
of Cloaked in Bricks
in the built environment, they usually include lim-
project.
ited possibilities to be applied in the design. For that
reason, digital technologies can assist in developing
proper methodologies for masonry (Al-Haddad et al.
2011) to meet contemporary design requirements
and integrate new geometric possibilities. As noticed
by Campbell (2005, 13), brick's evolution overtime
has been grounded in two major areas: brickmak-
ing technologies and brickwork techniques. While
Brick appeared to be a proper choice for the façade
the first one sets the physical properties of the mate-
covering since it has always been used as a lo-
rial (i.e. weight, dimensions, resistance, appearance),
cal building material in Iran meeting environmental
the second one defines the space for design cre-
needs while creating numerous aesthetically beau-
ativity and efficiency of masonry constructions. The
tiful textures. The complex form of the façade, lim-
developments in brickwork techniques tend to be
ited construction period and economical conditions
driven by structural and aesthetic goals, supporting
of the project forced the project team to search for
the materialization of magnificent brick walls, arches
a new construction method for implementing the
or vaults, and also ornamental surface effects (Sousa
façade. The appropriate method seemed to be elim-
et al. 2015, 362). By implementation of parametric
inating mortar by punching the bricks (Figure 1). Fi-
design methods various forms can be created. This
nally, parametric design software facilitated the tex-
is what many contemporary designers are looking
ture design process. Despite the complex form of
for. In recent years, some innovative brickworks have
the facade the construction process was easily exe-
been created by designers which opens new hori-
cutable by workers through simple instructions pre-
zons toward adoption of brick in contemporary ar-
pared by employing a system of coding [1]. As a con-
chitecture. It is important to highlight two magni-
temporary praised building, it focused on dancing
cant monuments inspiring the design generators i.e.
brick to show flexibilty of masonry materials.
Cloaked in Bricks and Spris Café.

Spris Cafe One of the main issues of design was creating a vi-
Spris Cafe with 28sqm space, designed by Hooba De- sual variation of the form in a small space. In this
sign Group is located in Nejatollahi street in center concept, the situation of visitors in relation to the
of Tehran surrounded by Iranian handicrafts shops, project is significant in order to understand the form
neighboring the building of Iranian handicrafts Or- as they can differently perceive the composition of
ganization. The aim of the project was to renovate colors on the bricks regarding to their position (Fig-
a gift shop and change it to a cafe, considering the ure 3). The turquoise blue glazed side of the bricks
small size of the project and its location the main idea are facing south shaped with the monolithic geom-
inspired by the urban context to transform the tradi- etry of brick laying that was modeled by the 3D di-
tional elements into an architectural interior space. agram started from the pavement of the pedestrian
and continues inside of the café [2]. In this case, zoo-
Figure 2 morphic and transitive character of brick, makes spir-
The combination of itual space which remind traditional architecture of
bricks in Spris Café . Iran.
Case studies show that recent architectural
projects focus on innovative understanding of brick
while lack of movement made them conventional
rather than cutting-edge architecture.
DESIGN TO PERFORMANCE
Figure 3 It is believed that through implementation of para-
The interior of Spris metric design, we can create movement and rhythm
Café, facing the in brickworks making it much more dynamic. There-
roof. fore it was assumed that interactive approach to-
ward adoption of brick in architectural design pro-
cess might be pioneer answer to the design question.
The design team had two challenges concerning the
design process:
1. How to create movement in a number of ob-

jects simultaneously?
2. How to create movement in heavy and ma-
sonry materials like bricks?
In designing the spatial diagram, the materiality con-
cept is based on an integrated geometry continues In order to find an energy efficient source of
from outside to inside. The neighboring building, Ira- movement the design team decided to adopt Mag-
nian handicrafts Organization with brick facade, was net as a source of energy. Electromagnetic energy
the inspiration to use the same material for the cafe. is an extensive renewable resource that has been
Concerning the small size of the project, a brick with underestimated in many fields such as architecture.
5*10*20 dimensions sliced into eight smaller pieces We believe that using magnet as a source of en-
of 5*5*5 centimeters which one side of the bricks ergy will open new approaches and causes huge ef-
glazed in turquoise blue color. The terracotta bricks fects in future buildings. Additionally, there is a new
are also hygienic as they covered with antibacterial approach toward applying magnetic fields in archi-
layer (Figure 2). tecture namely "Magnetic Architecture". The cur-

Figure 4
The prototype
representing the
roof structure in
scale 1:10.
rent focus of magnetic architecture explores the de- the magnet would absorb them. After all, punching
sign opportunities of a new building process from the brick in minor section and reinforcing them with
phase one: simply using recycled iron-based material some iron-made bars seemed to be an appropriate
controlled with-in a magnetic field (Diaz and Dubor, solution.
2013). Magnetic architecture might be seen as a po- The design was modeled in Rhinoceros
tential to increase the flexibility of additive process to ((C)McNeel) with the parametric design plugin
reach the architectural scale. Grasshopper ((C)David Rutten/McNeel). The sur-
face was converted into a series of EPS bricks with
MECHANISM OF USING MAGNETS 250x100x50mm, which is one of the standard di-
mensions of commercial bricks. In total, the model
THROUGH KINETIC BRICKS
comprises 700 bricks placed vertically and ordered in
The idea of moving bricks is based on the interac-
one level. Firstly, each brick was punched from its mi-
tion between two energies: the electromagnetic en-
nor section and then a bar inserted and fixed within
ergy of the magnet and the weight of the brick. In
it to suspend the brick from the roof. Additionally, a
other word, magnet pulls the pieces of bricks up-
grid of holes was made within the roof as a base for
ward, while the weight of bricks pulls them to the
bars. The holes were created in the square shape and
ground. Using this interaction between these ener-
also bars were rectangular in order to fit in the holes
gies made the design objective to be focused on the
and avoid undesirable rotation while moving (Figure
Roof of the pavilion. Therefore, a specific detail was
4). The suspended structure of bricks enables them
needed to hang the pieces of brick from roof, so that

Figure 5
Moving the magnet
on top of the roof
and rhythm in
bricks.
Figure 6 to move vertically through the holes when a magnet

The algorithm of absorbs them. The length of bars is adjusted in a way
grasshopper plugin that enables bricks to move less than 20cm vertically
related to so as the integration of the roof structure would not
movement of brick be damaged. While the bricks move in a series, they
based on magnet. form some waves in the roof which attracts the user.
At this stage a system was needed to be defined
for generating magnet energy so that it can over-
come the weight of bricks to move them upward.
To produce the required magnetic energy, two ideas
Figure 7 were brought up and examined i.e. Permanent Mov-
3D modeling of roof ing Magnet (Magnetic Field) and Temporary Fixed
structure and 2-axis Magnet (Electromagnetic Field).
robat moving on
1. Permanent Moving Magnet (Magnetic Field).
top of that creating
The first idea is based on using a permanent mag-
rhythm in bricks.
net and moving it (the source of energy) on top
of the brick grid. The iron bars within each brick
would be absorbed by the magnetic energy gen-
erated by the moving magnet, creating a wave in
the different parts of the roof (Figure 5). To actual-
ize this idea, extensive investigates were done and
the results were examined by altering the factors in
Grasshopper model (Figure 6).. Finally, it ended up
with the idea that a 2-axis robot (like Motorizable

2-axis X-Y) is needed to receive the commands from CONCLUSION
Grasshopper and move on the roof surface based The results of the project show that magnetic energy
on those commands. The commands are transferred might play a crucial role in future of kinetic interac-
to the robot by Ardoino kits. As a result, the robot tive architecture. This research opens a new horizon
moves the magnet attached to it and enables the in computer-aided design while needs extensive re-
bricks to move in a special order (Figure 7). searches and practices to be recognized as a design
and construction method. Implementation of Mag-
net has the following benefits: Figure 8
1. Magnetic energy is a renewable source of en- Temporary magnet
ergy and is generated without damaging the made of steel with
environment. cable around it
2. It reduces the costs of a project during the de- carrying electrical
sign to construction process. As an example, current.
to construct kinetic structures by using mod-
ern technologies high-cost systems is needed,
Linking the robot with Grasshopper gives us the op- but using magnets seems to be an affordable
portunity to draw various patterns and move the method for creating movement in space, be-
whole structure based on that to create those pat- cause the energy generated for moving one
terns on the bottom view of the roof. Also, we can object practically causes several objects to
equip the pavilion with some sensors to catch the move.
presence of human in the space and transfer them to Endless potential of brick makes it a flexible material
the robot to move based on human movement inside for contemporary architecture. The results of the re-
the pavilion. Consequently, the parametric design of search emphasized on potential ability on combina-
pavilion leads to an space with which people could tion between traditional material and kinetic charac-
engage in interactive relations. ter of contemporary interactive monuments.
2. Temporary Fixed Magnet (Electromagnetic REFERENCES
Field). The second idea is based on using an electro- Al-Haddad, T, Gentry, R, Cavieres, A, Carpo, M, Cho,
magnetic grid with fixed components which includes J, Wagner, L and Zaitsev, A 2010 'Representation
some temporary magnets located on a regular ba- + Fabrication: Connecting descriptions and arti-
facts in the digital age', Proceedings of 1st Inter-
sis. Temporary magnet refers to a piece of steel with national Conference on Structures and Architecture
some cables turned around it. By exerting an elec- (ICSA), Guimarães, Portugal
tric current involving the pieces of steel, they would Campbell, JWP 2005, História Universal do Tijolo, Calei-
transform to magnets and produce electromagnetic doscópio, Portugal
field (Figure 8). As the position of magnets is fixed in Diaz, GB and Dubor, A 2013 'Magnetic Architecture: A
this method, by altering the amount of electric cur- New Order In Design', 1st eCAADe Regional Interna-
tional Workshop, Porto, pp. 237-245
rent, the amount of voltage is changed and subse- Picon, A 2010, Digital Culture in Architecture: An Introduc-
quently the power of electromagnetic field would be tion for the Design Professions, Birkhäuser Architec-
variable. If we place the magnets in a grid order, they ture, Basel
cover the whole parts of the roof and variability of Sousa, JP, Varela, PA and Martins, PF 2015 'Between Man-
electromagnetic field causes the bricks to move ver- ual and Robotic Approaches to Brick Construction
tically and create waves. in Architecture', Proceedings of eCAADe 2015, Austria,
pp. 361-370
[1] http://architizer.com/projects/cloaked-in-brick
s/

Aesthetics of Decision
Unfolding the design process within a framework of complexity and self-
organization
Alessio Erioli1
1
University of Bologna
1
alessio.erioli@unibo.it
Complexity-grounded paradigms and self-organization based strategies promise

enormous potential when channeled in a design process, but their current stage of
development (while delivering groundbreaking results in recent years) hasn't
significantly impacted yet the widespread architectural practice. Still, the
tendency (in the development of technology and society) is clearly towards an
increase in complexity and distributed intelligence, henceforth it is of primary
importance to adopt a design approach that allows the harnessing of such
potential and convey it in the creation of outcomes that favor a richer and
heterogeneous ecological entanglement. To tap this kind of potential in an
open-ended process requires a design approach that re-defines the distribution of
control, choices and information throughout the whole process (including
materials and fabrication processes).The paper explores the possibility of such
design approach in the territory that links education and research through a
series of Master Thesis developed at the University of Bologna and comparing
them to other case studies developed worldwide.
Keywords: continuity, tectonics, architecture, mereology, multi-agent systems,

theory, robotic fabrication, computation, simulation
INTRODUCTION AND BACKGROUND the role of the designer was that of a mastermind able
Complexity is undoubtedly one of the most powerful to conceive a complete model of the entire project;
paradigms of contemporary age; the idea of emer- an immediate consequence of such condition is that
gence and self-organization in particular, that com- the highest complexity achievable coincides with the
plexity can stem from the interaction and feedback one conceivable by the mind of the designer. Au-
over time governed by simple rules capable to pro- tomation of simple tasks (the hallmark of the indus-
duce stable, coherent organizations is of particular trial revolution) was only implemented in a linear
interest for its potential implications in architecture. fashion without the possibility of feedback, there-
Prior to the discoveries that led to the formula- fore precluding any form of endogenous adaptabil-
tion of this paradigm, it was taken for granted that ity or self-adjustment in the system, strengthening

further the divide between design and making first process requires a design approach that re-defines
theorized centuries earlier by Alberti. The design pro- the distribution of control, choices and information
cess thus sedimented a clearly hierarchical decisional throughout the whole process (including materials
pattern based on the division of labor where the di- and fabrication processes).
vide did not only involve design and making, but The work presented in this paper situates itself
branched further the design process itself separat- after precedents in the use of agency and complex-
ing (save for few remarkable exceptions) conceptual ity in architecture such as the work and research by
ideation and constructive thought as two consecu- Kokkugia ([3] and the recent independent develop-
tive moments. The technological acceleration and ments both in academia and practice by its respec-
the evolution of scientific-phylosophical paradigms tive founders Roland Snooks and Rob Stuart-Smith),
usually do not go hand in hand, with the first set- Jose Sanchez (Plethora Project) and his work with Al-
ting foot first in unknown territory while still carry- isa Andrasek at the DRL and the Bartlett (one exem-
ing the ways of the old ones: the first application of plar case: [4]), as well as other numerous other cases
new technologies almost always happens as an ex- in those and other relevant schools. What all those
tension of old paradigms (it happened with concrete cases share in common is an heuristic use of agency
in architecture, or with the desktop and file metaphor and multi-agent system as a way to embed performa-
in computers); it takes some time and adaptation be- tive coherence and navigate it in search of novel and
fore our mindset is changed by it and starts to envi- richer tectonic (spatial-structural, as well as percep-
sion the true, inner potential and its consequences. tual) expressions. The predictive use of multi-agent
So the advent of computers in architectural systems such as modeling user behavior in architec-
practices happened as a substitute for the drawing tural spaces is not of interest and while its reliability
board first and in the power of calculation slightly in some cases is acknowledged it is also not relevant
after (again, with few remarkable exceptions, see in this context.
[1]). Problem solving oriented processes were early
adopters, due to the inevitable osmotic exchanges IMPLEMENTATION
from computer science and reliance on measurable The methods adopted in the theses rely on iterative
performance, and contemporary instances already computational simulation implemented in several
benefit from agent-based strategies, genetic algo- programming and modeling platforms coupled with
rithms and neural networks (since they deal with material experimentations and (in most cases) multi-
measurable performance henceforth naturally evolv- agent system and robotic fabrication: such phases
ing towards greater efficiency through technology are interwoven with and not subsequent to compu-
and technique). Still, the tendency (in the develop- tational simulation.
ment of technology and society) is clearly towards an Although simulations might well start by try-
increase in complexity and distributed intelligence ing to reproduce certain processes or behaviors
that, for its own nature, exceeds and transcends any (using role models from biology/geology/physics/-
definition of "optimal" as always context-biased and chemistry for their inherently elegant complexity)
resists reduction to accomodate a plurality of func- their potential resides in their endogenous coher-
tional and communication patterns. ence. Once the relations regulating the system are
It is of primary importance to adopt a design ap- established so is the metric of the system's output,
proach that allows the harnessing of such potential despite little or none is known about the final out-
and convey it in the creation of outcomes that fa- come. They operate in a state of vagueness, driven
vor a richer and heterogeneous ecological entangle- by internal logic but open in determination and their
ment. To tap this kind of potential in an open-ended outcomes, being consequence of a coherent set of

operations, are endowed with the property of being of intensive processes of formation through the de-
a structure of communication in themselves. Internal sign of the material and geometry behaviors regu-
logic is not an overarching imposed principle, gov- lating them. The architect programs the decisional
erning form and function from the outside, but an pattern shared by the individual agents and makes
embedded feature of relations and objects that blos- decisions on the shape and topology of their body
som into organized complexity. Logic becomes an (thus defining their affordances) as well as their sen-
aesthetic operation that cannot be subjected to com- sorium, the range of accessible and processable data
plete synthesis or axiomatics, it operates in an irre- and its form (i.e. reading solar radiation in form of
ducible complexity. Simulations also do not need to vertex colors on a 0-255 value range, or a gener-
refer to a role model anymore, moreover their poten- ated data field not necessarily connected to a par-
tial lies exactly in the possibility to explore the un- ticular physical variable). Global form, collective
known by inventing alternate modes of reality, or hy- behavior and tectonics emerge from their interac-
perreality, as Baudrillard (1983) calls it. This capac- tion over space and time. There are two important
ity of abstraction is very different from the Albertian points that should again be stressed: the environ-
notation system that separated conception from re- ment is designed (when constraints are chosen and
alization, freezing all dynamics in the representation data to be read is organized) and the designer is
to which material processes should subdue; it is an a player in a more extended ecology involving ma-
abstraction of principles, one that conveys the dis- chines (analog, digital or biological) that operates
tance between the design intent and the artifact at not only through human-machine dialog but also
the level of instructions. Algorithms should then be by machine-to-machine communication or "the cor-
considered as actual modes of thought (which ex- relation of technical beings among themselves" (Si-
tends Baudrillard's precession of simulacra to the al- mondon 1958, 181). Ontologically speaking, the rela-
gorithmic world) that symbiotically evolve with us; tion between code and being is perpetual: differently
more concisely and to reconnect the notion of logic from the Deleuzian diagram/assembly where the for-
as aesthetic operation, what we need (in the words mer only provides a topological scheme for the lat-
of Shaviro) is "an aesthetic of decision" in a frame- ter but ceases to operate once the latter is formed,
work of self-organization and emergence [2]. The ex- the code continues to inform and operate on beings
pression touches a very important part of the design as they mutually interact and alter their own code
process that has to do with authorship, intentionality (Spuybroek 2008).
and the possibility, in the particular case of genera- This approach to computational simulation pro-
tive design, to drive self-organizing system in order motes an alternative understanding of control and
to better explore and harness their undisclosed po- interaction between top-down and bottom-up pro-
tential. According to Roland Snooks (2012) author- cesses, without giving in to the easy production of
ship in generative design can be traced in the pattern highly ephemeral systems that champion random ac-
of two contrasting kinds of decisions: criteria (which cidents nor reduce its exploration to the safe haven of
frame the possible realm of the output and are inher- algorithms with stable indexical relations to their for-
ently stable) and procedure (which regards the con- mal results. While computational design has straight-
dition of operation, therefore is inherently specula- forward advantages in the automatization of pro-
tive and processual); simulations enable the possibil- cesses and designing the structural and organiza-
ity of emergent outcomes through the interaction of tional aspects of a project, if our intention is to har-
design procedures. ness the inherent potential of non-linear complex
There is then a shift from the explicit design of systems it is not enough to simply apply a picture of
form and organization towards the choreographing complexity to essentially modern objects; complex

systems are equally adept at engaging with form, as and computes multiple discreet information in each
the emergence of form can be understood as the or- of its voxel cells, doing most of the heavy-lifting in
ganization of matter itself. the computational process. It acts as common sub-
strate for data exchange through which the multi-
CASE STUDIES agent systems is communicating by reading/writing.
It is the case of Oxymoron (Figure 1), a thesis project The information in the environment is then encoded
where Topology Optimization (usually employed as a as data map in the voxel field, selectively accessible
reduction device to optimize structure in given con- by the agents in 3 possible ways: data retrieval from
ditions) is used for its inherent rationality in comput- the nearest cell, value interpolation from n nearest
ing form according to stress distribution patterns as a cells (with different degrees of anisotropy), direct cal-
mean to explore opulence and a richer spatial expres- culation from the field general formula.
sion. The iterative exploration of non-reductive con- This common setup is then implemented (with
straints patterns (for which the condition to respect significant differences) in three more case studies:
stress-response criteria was necessary but far from The Red Queen Hypothesis, Ari3dne and Home-
sufficient) in the design of a tower for Shanghai was orhetic Assembles. In all of them the decisional
a key factor in maximizing the spatial expression and pattern is twofold: on one side the definition of
transcend the typical bone-like structural figuration environmental constraints and conditions (i.e. a
of Topology Optimization algorithms. Iterative sub- "pheromone" gradient that induces anisotropic be-
division of simple triangular façade elements in rela- havior) as well as specific interaction rules, the out-
tion to the distribution of stress patterns reinforces put evaluation and refinement of initial conditions
and further articulates the visual heterogeneity and on the other. Changes in these initial conditions do
complexity. not lead to a proportional response but a non-linear
In other cases, systems are modeled around propagation and feedback that might result in a het-
object-oriented strategies, extending the classical erogeneously densified phylogeny in terms of sys-
multi-agent system model theorized by Reynolds tem expression (highly volatile variations may occur
(1987) and comprised of the basic local rules of flock- in narrow parameter spaces whereas large subdo-
ing (direct communication between agents by cohe- mains might show a "flat" variation landscape). Func-
sion, separation and alignment) with the stigmergic tional performative criteria as well as unapologeti-
model (an indirect communication protocol that oc- cally aesthetic reason can drive the selection pattern
curs through the environment) based on the theo- that then feeds back into the algorithm in an increas-
retical premises in Jones (2010) and occasionally cus- ingly serrated and convergent exchange. These sys-
tom made rules for direct/indirect interaction, with or tems' internal logics provide solid coherence and the
through the environment. Interaction between dif- mapping of function onto form does not precede the
ferent agent "species" can be of collaborative or com- project development not privilege specific relations
petitive nature; the latter provides the conditions to over equally possible alternatives in an open spec-
strengthen and stabilize the former. trum.
Environmental perturbations do not alter di- In The Red Queen Hypothesis (Figure 2), the
rectly and/or proportionally the morphology of the competitive behaviors of two different sets of agents
system but affect the agents behavior, which pat- fighting for resources fosters the generation of an in-
tern of propagation through the system is not an interlocking spatial negotiation that reaches homeo-
dexical image of the perturbation field. The envi- static stability over time. The strategy is applied at
ronment itself is coded as an anti-object, (Repenning the scale of tectonic development of a structural skin.
2006) an active part of the simulation that carries

Figure 1
Oxymoron - thesis
project by Luca
Pedrielli - advisor
Alessio Erioli -
co-advisor Giovanni
Castellazzi.
Figure 2
The Red Queen
Hypothesis - thesis
project by Paolo
Alborghetti -
advisor: Alessio
Erioli.

Figure 3
Ari3dne - Thesis
Project by Daniele
Colombati - advisor:
Alessio Erioli -
co-advisor: Gabriele
Evangelisti.

Figure 4
Homeorhetic
Assemblies - Thesis
project by Roberto
Monesi - advisor:
Alessio Erioli -
Co-advisor: Paolo
Alborghetti.

Figure 5
Swarm
Constructability -
Thesis project by
Michele Semeghini
- Advisor: Alessio
Erioli - Co-advisors:
Martino Colonna,
Massimiliano
Cicorella.

Ari3dne (Figure 3) aims to embed constructor agents the single robot has been reduced to the bare min-
as part of the architecture, mutuating its tectonic ex- imum of movement (upside down thanks to mag-
pression from the ridged web formation process of netic wheels so to not interfere with the deposition),
the Euplectella Aspergillum; the agent have the ca- avoidance, signal emission and detection, stigmergy-
pacity to read and write situated data in the environ- based steering. The key parameter affecting how
ment and on the built material organization and op- the others play out is the minimization of energy
erate both as constructors or destructors according consumption for the single unit: the chosen ma-
to seasonal, use and circulation pattern variations. As terial works at room temperature, allowing to use
a result, the project blurs the traditional boundary for movement also the energy otherwise required to
between construction and maintenance process as keep it above the state change threshold. The peers
well as challenges the notion of a fixed object with coordinate with each other and choose when pour-
a "final" configuration while retaining as strong iden- ing material through a signal deposited in the envi-
tity and presence. ronment (light stored and slowly released by a sensi-
Homeorhetic assemblies (Figure 4) incorporates tive coating in this case).
the dynamics of beehive formation and simulates the
evolution of material viscosity and its hardening over DISCUSSION
time while constructor agents perform a stigmergy- The desire to talk to machines is not merely driven
based deposition in a data-rich environment. by a necessity to optimize a design to production
In both cases assemblies are dynamic and pipeline; rather, it is directed towards "the devel-
strongly redundant as they perform a variety of func- opment of reciprocity between designing and mak-
tions and exhibit several possible affordances that ing" (Maxwell and Pigram, 2012). Matter itself can
transcend the mere structural capacity.In the sim- produce tectonics, as it is able to process informa-
ulations that involve constructor agents, their be- tion and spatially organize as a consequence of those
havior is conceived with specific constraints and op- processes; material qualities are also translated into
erations tied to a machine (such as a robotic arm, properties of the interacting elements within the sim-
3D printers, etc.), its body topology and capacities, ulation; as system behaviors, they become rules for
and a material system. Thus, the relation with fab- the emergence of form as a consequence of material
rication and material goes both ways: simulations arrangement which cannot be included in the sim-
talk directly to machines as constructor agents (di- ulations. In the iteration among simulation, agency,
rectly writing machine code as output) and physi- constructive behavior and material effects, the aes-
cal outcomes and simulations are precious sources thetics of choice implemented by the designer af-
of information that then feedback into the simula- fects the trajectory and convergence of the design
tion programming as design constraint and param- process in ways that aim to challenge the established
eters (body size and movement constraints, material linear conception-realization sequence. While the
behavior). intentionality and subjective contributions can be
These principles constitute the foundation of an- traced as the pattern of choices and how these are
other thesis project, Swarm Constructability (Figure distributed into the process, the non-linear nature of
5). Here the research spanned low-cost robotic con- the process itself (and thus the impossibility of es-
struction, material system and behavioral program- tablishing clear cause-effect predictive relations on a
ming. In order to better detect the potential emer- global scale) leaves authoriality as an open question.
gence of self-organizing properties and artefact qual-
ities as a result of the system interaction and not of
the individual unit sophistication, the behaviour of

REFERENCES
Baudrillard, J 1983, Simulations, Semiotext(e), New York
Jones, J 2010, 'Characteristics of pattern formation and
evolution in approximations of physarum transport
networks', Artificial life, 16(2), pp. 127-153
Maxwell, I and Pigram, D 2012, 'In the Cause of Archi-
tecture: Transversing Design and Making', in David-
son, C and Roche, F (eds) 2012, Log n. 25: Re-
claim Resi[lience]stance//……R2, Anyone Corpora-
tion, New York, pp. 31-40
Repenning, A 2006 'Collaborative diffusion: program-
ming antiobjects', Proceedings of OOPSLA '06 Com-
panion to the 21st ACM SIGPLAN symposium on
Object-oriented programming systems, languages,
and applications, pp. 574-585
Reynolds, C 1987, 'Flocks, Herds, and Schools: A
Distributed Behavioral Model', Computer Graphics,
21(4), pp. 25-34
Simondon, G 1958; repr 1989, Du modes d'existence des
objets techniques, Aubier, Paris
Snooks, R 2012, 'Volatile Formation', in Davidson,
C and Roche, F (eds) 2012, Log n. 25: Re-
claim Resi[lience]stance//……R2, Anyone Corpora-
tion, New York, pp. 55-62
Spuybroek, L 2008, The Architecture of Continuity, V2_-
Publishing, Rotterdam
[1] http://www.danieldavis.com/a-history-of-paramet
ric/
[2] http://www.shaviro.com/Blog/?p=756
[3] http://kokkugia.com/
[4] http://drl.aaschool.ac.uk/portfolio/plugin/

Complexity and Simplicity
Tensions in teaching computation to large numbers of architecture students
A. Benjamin Spaeth1 , Theodoros Dounas2 , Joachim Kieferle3

1
Cardiff University 2 Xi'an Jiaotong Liverpool University 3 Hochschule RheinMain
1
spaetha@cardiff.ac.uk 2 theodoros.dounas@xjtlu.edu.cn
3
joachim.kieferle@hs-rm.de
This paper describes the challenges and approaches to introduce computational

thinking to a large and diverse group of architecture students during an
international workshop with 300 students from different cultural backgrounds and
educational levels, also integrating a diverse group of tutors whose computational
expertise varied extremely. The approach suggested articulating a design task
which enforced computational thinking but enabled different levels of engagement
with the computer as a tool. Hypothetically this would allow all participants to
engage with the computational thinking agenda regardless their computational
affinity even whilst applying analogue methods. Besides the intercultural
experience the workshop was successful in exposing a large group of students and
tutors to the concepts of computational design whilst accommodating different
learning preferences and engagement with the computer as a device.
Keywords: Computation Education, CAAD, Large Cohorts, Computational

Strategies
INTRODUCTION thinking and computational exploration within the

The paper presents the methods and results of a com- workshop for and by the students and faculty. The
putation and architecture international workshop in workshop was a one week design exercise where stu-
collaboration with Hochschule RheinMain, Germany dents from different years and levels of the two differ-
and Xi'an Jiaotong Liverpool University located in ent schools from Germany and China participated in
China. The workshop took place in Suzhou, one of vertically integrated group.
the biggest and fastest developing urbanisation ar-
eas in China in last 15 years, where the question of TEACHING COMPUTATION
fast and temporary urbanisation is immanent to pro- Teaching computation in Architecture is challeng-
vide living space for the thousands of migrant con- ing. It seems that everyone is expecting some fail-
struction workers building up the permanent settle- ure and some success in project based teaching in
ments. The number of students taking part in the Architecture, more so in CAAD (Brown 2002). CAD
workshop was large, numbering 300, creating an ad- teaching has developed from a learned skill-based
ditional challenge to the delivery of computational

approach where the target was the emulation of in the literature focusing on what exactly students
the productive capacities of a drafting department learn through courses in computation in architecture
of an architecture office, with few enlightened ed- schools worldwide (Hemsath 2010, Senske 2014).
ucators driving towards the design rather than the The educational workshop presented here is to
drawing fraction of CAAD. (Clayton et al 1999) Com- introduce computational thinking and designing at
putation has been used in various supporting roles the core of the curriculum, the design studio. The stu-
in architectural courses, from straightforward archi- dio was organised as a vertical studio including the
tectural design studio graphics and representation entire first and second year students of the hosting
(Iwamoto 2004) to structures, construction, technol- institution as well as all their tutors. The tutors in-
ogy and any other technical aids. The major chal- volved in the workshop though were in the majority
lenge in CAD teaching has been identified as to move not computational literate which put an additional
from the application of a digital tool employing inef- challenge to the pedagogical concept. Avoiding to
ficient routines to a deeper understanding of com- be caught in an ideological clash with the 'analogue'
putation and the inherit core strategies. In archi- mode of teaching which most of the participating
tecture this seems to be significantly different since staff was used to, the computational and paramet-
architects are very likely to apply the "production rical thinking was introduced rather through man-
bias" which detains them from getting deeper into a ual and analogue means than through using com-
subject than needed for the job to be done (Senske puters. This challenge was approached by shaping
2014). At the same time CAAD instructors have de- and restricting the design exercise as to computa-
veloped a variety of strategies to assist students to tional thinking and means are almost impossible to
understand and acquire CAAD skills and deeper un- avoid. Computational thinking was required to set up
derstanding of computation in architecture. These a flexible urbanisation system from containers which
strategies range from a fundamental position where could adapt to diverse situations by identifying and
CAAD is buttressing architectural composition (Flem- including recurrent parameters into the design sys-
ming 1989) to reflections of a linguistic nature (Cicog- tem. The brief required students to develop a com-
nani 2000, Cheng 1997); from technology being the putational approach not by using computers or soft-
driver (Bechthold 2007) to software seen as an envi- ware but by developing a computational and para-
ronment constructed for an ease of understanding by metric concept which was immanent to the design
novices (Gannon et al 2014); from CAAD as a driver for brief by its given boundary conditions and the avail-
energy performance analysis (Dvorak 1988) and from able variables.
integrating CAAD courses vertically in the curricu-
lum (Bollinger 1987) to offering CAAD as a separate Large Group Teaching
form of architectural design. For many CAAD courses This workshop was conceived to introduce not only a
there has been a chronic time-lag between the de- small group of selected students with an already ex-
velopment of a method or tool and its introduction isting particular interest in architectural computation
into the curriculum- (Clayton et al 1999). Apart from but to reach as many students as possible to intro-
the variety of responses this last point effectively ar- duce them to computational thinking in architecture
gues for the alignment of CAAD research with teach- and to rise their awareness of its potential, its appli-
ing it in the classroom, in the hope that such cross- cation to design where students might have not ex-
pollination would also allow optimisation of the state pected it and to degrade students potential reserva-
of the art in skills of architectural graduates. This tions against architectural computation.
notion will also hopefully advance the CAAD profes- Facing the large amount of students, numbering
sional practice. However there are very few studies around 300, along with a diverse educational level,

or lack thereof in computational tools amongst stu- 10 lengths across the axis. The crane was deemed ex-
dents and tutors it was a challenge to set up a ro- ceptionally high and could place up to six contain-
bust enough design task which demanded for com- ers in height. The stakeholders are people of high
putation related thinking but was not only based on mobility in the Chinese urban context (migrant work-
the use of computer or software so as also computer ers, students, early career professionals, freelancers
averse staff could tutor a student group. The work- or similar occupants). The mix and the needs of the
shop had tutorial groups of 20-25 students with one occupants are part of the variables students had to
tutor guiding each group. In each group teams of define in setting up their parametric system. Also
five or four were created, each one having to de- the different topologies which could be found along-
velop one project responding to the brief and using side the canal needed to be included into the algo-
computational concepts and tools. We had about rithm. The utilisation of a modular system consisting
12 tutorial groups formed, with only about half be- of containers was a means to support students' focus
ing versatile in classic CAD tools and only three tu- on computational thinking since the containers of-
tors being educated in the advanced use of compu- fer natural points of connection and embedded con-
tational tools. Therefore the container was conceived straints. They are units that facilitate bottom up de-
as to act as an invisible unit of computation, where sign concepts but at the same time they are architec-
the creation of diverse grammars and syntaxes would tonic through their inherently specific physical con-
emerge from the experimentation with this compu- straints and possibilities by design.
tational unit within the set boundary conditions even The high level strategy in developing an urban
without using the computer to model the systematic settlement should be inherently three dimensional
and algorithmic thinking. and allow for the connection to the plots/sites of
neighbouring teams. So the teams should agree on
The Brief and share their specific boundary conditions of their
The brief asked students to develop a small city hous- developed concept.
ing 2000 people across a water way between Suzhou
and the town of Zhou Zhuang, in China which locates Strategies and Computational Tools Em-
in one of the fastest and largest Chinese urban deployed
velopments at the moment. With a reference to his- We provided to all students introductory sessions to
toric urban developments along the main rivers and the computational tools and made the tools as sim-
canals in China and by identifying the water ways ple to use as possible. The computational platform
as efficient means of transportations one of the de- selected for the workshop was Rhino with Grasshop-
termining boundary conditions was to develop the per. The faculty prepared parametric definitions on
settlement along the great canal between the two grasshopper using separately cellular automata and
permanent settlements. The second determining L-systems, which allowed the students to design us-
boundary condition was the given size of standard ing prescribed definitions. We used the rabbit plug-
overseas containers and the immanent variations of in to be able to encapsulate high level computa-
stacking them. Furthermore it was determined that tion on cellular automata in simple parametric de-
the containers are shipped by the floating barges cisions. The parametric definitions were well docu-
only and can only be set by a specific container crane mented and sufficiently transparent so that the stu-
mounted on the barges. The ship, used for container dents would be able to employ them and under-
distribution, was assumed that it could place a con- stand the ramifications of their decisions in changing
tainer no further or higher than 4 lengths of a con- strategies or changing parameters. At the same time
tainer from the bank of the canal, and no more than the code was accessible enough for students to inter-

Figure 1
'Floating City'.
vene on a more fundamental level if applicable. At munity. Students demonstrated the use of compu-
the beginning of the workshop an instructional ses- tational thinking on an internal level organising the
sion on using the parametric definitions was given to islands as independent entities with the internal rule
the students in groups of about 50, repeating the ses- set applied to the elements and on an external level
sion 6 times. Essentially the tools were templates that where they identified interdependencies between
the students could use. They could stay within the the autonomous islands and the environment artic-
confines and constraints of the template or as some ulated into a parametric system to be applied to dif-
did tweak and explore more outside the frameworks ferent situations. Architectonically the work was also
we provided. successful in terms of completeness and aesthetics.
The students of the 'Circular High Mobility City'
RESULTS (Figure 2) approached the task by introducing a
The project 'Floating City' (Figure 1) employed not twofold zoning with a commercial zone facing the
only the computational tools provided by the teach- land side and the residential zone facing the canal.
ing team but the students also developed their own This provided a systematic different urban quality
narrative of a floating city and a container gram- to the different zones, introduced a public space
mar. It was noted that the grammar works on two between the zones and allowed for different pat-
scales, in assembling the containers into islands as terns within the two zones to emerge. Although the
autonomous entities and on a second level into as- project was presented as a closed ellipse the poten-
sembling these islands into a city as a larger com- tial to be populated along the canal in different ex-
tensions is self-evident.

The project 'Floating City 2' (Figure 3) is a representa- Having started from implementing and adapting
tive of a grid like approach. Although the computa- the provided L-system, combining it with the historic
tional approach would allow for overcoming a grid in quadratic elementary form of the Chinese traditional
favour of a flexible system of relational objects this city the group also ended up in a grid system (Figure
group of students approached the task by deploy- 4). They conceived their grid as a rail system, where
ing a topologically flexible grid onto the project site. the containers can be arranged according to occupa-
Determining service points and circulation routes the tional needs. However the automated mechanism is
grid is then populated respectively with containers. supposed to fill the grid in the one or the other way
Figure 2
'Circular High
Mobility City'.
Figure 3
'Floating City 2'.
Flexible grid with
parameterised
container
deployment.

Figure 4
Rigid grid with
flexible container
deployment
controlled by user.

Figure 5
'Nest City'. Two
overlapping system
patterns.
the basic principle remains the grid. This system here Inspired by the perforation of karst formations and
was conceived as an adaptable and controlled sys- anthills a student group developed a linear pattern
tem within the rigid grid, whereas the before men- which took into account daylight, topographic con-
tioned grid was adaptable to the found topology. ditions, functional aspects as well as density consid-
erations (Figures 5, 6). Based on a closed block defini-
Figure 6 tion the group loosened the block structure to aerate
'Nest City'. it but also extended it into a basically infinite ribbon
that could be knitted along the canal. The repetitive
pattern includes a conscious and also repetitive dis-
ruption formed by central public services. In terms
of computational thinking this represents an inter-
esting approach since the continuity is the standard
case and the disruption is the exception which takes
a deeper understanding since a continuous pattern
needs to be overlapped with a second layer defining
the disruptions.

CONCLUSION REFERENCES
The projects presented in the final presentation were Bechthold, M 2007 'Teaching Technology: CAD/CAM,
of a great diversity and reflected students' outstand- Parametric Design and Interactivity.', Predicting the
ing engagement with the task. Despite the large future, Frankfurt/Main, pp. 767-775
Bollinger, E 1987 'The New Studio: CAD and the Work-
number of students and the assumingly neutral
station: Implications for Architecture Education.',
interest to computational methods in architecture eCAADe Conference Proceedings
amongst the majority of the students and tutors the Brown, A 2002 'CAAD Complements: Art and Science,
engagement with the design task but also with the Real and Virtual | Cumulative Index of Computer
implicit computational agenda was on a high level. Aided Architectural Design', 20th eCAADe Conference
Although it was difficult to control the groups and Proceeding, pp. 214-218
Cheng, NYw 1997 'Teaching CAD with Language Learn-
in particular the tutors' interpretation of the design
ing Methods', Acadia 97, Cincinnati, OH
task the assignment was robust enough to engage all Cicognani, A 2010, 'Boxing architecture', Digital Creativ-
groups into a systematic and algorithmic approach, ity, 11(3), pp. 190-192
at least in a wider sense. Due to the container as Clayton, MJ 1999 'Stumbling, Backtracking, and
a computational unit and the set boundary condi- Leapfrogging: Two Decades of Introductory Archi-
tions most of the groups engaged with the intended tectural Computing', Architectural computing from
Turing to 2000, pp. 151-158
computational agenda. The consciousness of the
Dvorak, RW 1988 'Designing in the CAD Studio', ACADIA
agenda amongst the students may vary but we re- Conference Proceedings
ceived feedback from students that explicitly referred Flemming, U 1990 'Syntactic Structures in Architecture:
to the workshop and how it was inspiring for a regular Teaching Composition with Computer Assistance',
housing studio in the course of their studies. The Electronic design studio, Cambridge, Mass.
The simplicity of the interface of the computa- Gannon, M and Brockmeyer, E 2014 'Teaching CAD/CAM
Workflows to Nascent Designers', Proceedings of the
tional agenda to the participating students and tu-
19th International Conference on Computer-Aided Ar-
tors facilitated the engagement of the large majority chitectural Design Research in Asia, pp. 801-810
of participants, students as well as tutors. The level Hemsath, TL 2010 'Searching for Innovation Through
of complexity behind the computational agenda was Teaching Digital Fabrication', Future cities, Z\"urich,
to be uncovered by the groups themselves; on an in- pp. 21-30
tellectual level as well as on the level of engagement Iwamoto, L 2004, 'Translations: Fabricating Space', Jour-
nal of Architectural Education, 58(1), pp. 35-38
with the computational tools provided.
Senske, N 2014 'Digital Minds, Materials, And Ethics:
On a next workshop we hope that we can use the Linking Computational Thinking and Digital Craft',
massive number of students as 'live design agents' Proceedings of the 19th International Conference
to solve architectonic problems. Instead of pro- on Computer-Aided Architectural Design Research in
gramming computer agents we can use the inte- Asia, pp. 831-840
grated power of human and analogue computing to
attempt complicated solutions to design problems.
This of course can only happen if we still employ the
same strategy as here: common computational tem-
plates, where the common platform is an advantage
leading to diverse solutions rather than a disadvan-
tage constraining the participants.

From Designing Buildings from Systems to Designing
Systems for Buildings
Heike Matcha1
1
Aachen School of Architecture
1
Matcha@fh-aachen.de
We study the novel possibilities computer aided design and production open up
for the design of building systems. Such systems today can, via individualized
mass production, consist of a larger number and more complex parts than
previously and therefore be assembled into more complex wholes. This opens up
the possibility of designing specialized systems specifically for single buildings.
The common order of starting with a building system and designing a building
using this system can be reversed to designing a building first and then
developing a system specifically for that building. We present and discuss
research that incorporates students design projects into research work and fosters
links between research and teaching.
Keywords: Building Systems, Parametric Design, Parametric Modelling,

Structuralist Architecture
Systems not only allow us to think in an orderly fash- ularities and made more difficult by irregularities or
ion, following a certain plan, but to be able think at all. differences between the parts or components. So in
Georg Christoph Lichtenberg, 1776 a way, all building is systematic building and always
has been - starting from simple stone huts where
Introduction and Background: Monotony the individual stones are selected for similarity to
vs. Variety in Systematization straw huts where the strands of straw are as equal
Every building has to be assembled from parts that as possible, via the pyramids assembled from simi-
can be handled and transported. And bearing the lar stones. The question is not whether a building is
advent of large-scale on-site 3D printing or compa- systematic, but how much. In modern architecture,
rable building technologies, the assembly of build- much design started from industrialized production
ings from individual components (i.e. bricks, pan- methods, idealizing Henry Ford's automobile con-
els, beams, girders) will remain an integral part of veyor belts. Such Fordist mass production, though,
architectural production. More generally speaking, resulted in extremely regular, even monotone struc-
building designs have to be subdivided into com- tures, because only parts equal to one another could
ponents that can be produced, transported and as- be produced. Systematizing production and there-
sembled. The more those components are related to fore design meant minimizing variety, thereby maxi-
one another, the better, because production, trans- mizing monotony and anonymity. This, we propose,
port, handling and assembling is simplified by reg- was and is an important reason for the public dis-

missal of much of modern architecture. As the ad- Buildings from Systems or Systems for
vent of computer aided construction and production Buildings ?
today allows for the cost-neutral mass production of The possible conception and production of complex
variety, building systems can be seen in a new light: building systems allows to reverse the relationship
systems can be much more complex, and therefore between direction and design that has been preva-
monotony and anonymity avoided in favour of diver- lent in modern architecture (and still is in much of
sity. We see the mass production of diversity in ar- current architectural production): the systems are de-
chitecture as a desirably goal because it can bring ar- veloped first, aiming for greatest versatility, and ar-
chitecture closer to the diversity found in the natural chitects design their buildings using those systems.
environments (and to some extent, medieval town- The most famous example of such a building proba-
scapes) humans have evolved to feel comfortable bly is the Mero Construction System. This procedure
in. (Matcha 2015, Matcha and Barczik 2009, Matcha has monotony inscribed in its core: through the avail-
2010, Matcha and Karzel 2013.) ability of only one or merely a few types of compo-
nents. Today though, this 'classical' way can be re-
Figure 1
Designing with
Systems:
Experimental
preliminary student
studies, Aachen
School of
Architecture.

versed. A system may be derived from a building de- Research project Overview
sign. This would have been non-sensical under the We study these novel systematization possibilities
paradigm of fordist mass production where the eco- through a series of hypothetical student design
nomic goal was to reduce the number of parts as projects. Building designs of varying complexity are
much as possible. When variance can be mass pro- developed and then studied for possible segmenta-
duced, the parts can be different, and the systems tion into parts. In order to be able to be produced, the
more complex and therefore individualized to a spe- parts need to be related to one another, like mem-
cific design project. Furthermore, instead of one sys- bers of a family. Such relationships can be devised via
tem having to be able to generate different types of parametric variation. Students are therefore familiar-
buildings (e.g. housing, schools, administrations), a ized with parametric design concepts, tools and tech-
system may now be devised for ones specific build- nologies. The back-and-forth between systems and
ing only. Even if every part is different, and only used buildings is studied in both directions. Different com-
once, for just one building, the production would still plexities and sizes of parts are studied together with
be cost-neutral in comparison to the 'classical' sys- different means of connecting them. Assembly and
tems of identical parts due to the benefits of cus- modular building systems from the 1960s and 1970s,
tomized mass production as long as the parts are re- when many projects attempted to stretch the possi-
lated like relatives in a family. Instead of designing bilities of mass production to the limits, are studied
buildings from systems, we can now design systems together with possibilities to extend them for more
for buildings. variability.
Figure 2 A closer look at the research project and the

Designing with incorporation of student projects
Systems: The course consists of two parts, the first of which
Experimental has three lines of study running in parallel: First: be-
student studies for coming acquainted with parametric CAD tools. We
lamp designs, give an overview and choose to work with Rhino
Aachen School of and Grasshopper as they are widely used and easily
Architecture. contain the functionality we require. Second: Ana-
lyzing existing buildings for which building systems
have been used. The students understand how the
Figure 3 systems work, how the have been used and which
Experimental shortcomings they have - mostly, with 'classical', non-
pavilion design, customized systems, the architects striving to escape
Physical and Digital monotony, thereby working against an aspect that
Models, Student lies in the systems' principle. Third: Playfully creat-
Project by Markus ing a simple design from a readily available simple,
Schöps and Ramon even primitive, building system: matches (Figures
Pardo Vaquez . 1 and 2). The students experience the possibilities
Aachen School of and restrictions of a tightly constraining building sys-
Architecture. tem literally at first hand. For us, it is very important
that the students work in several media in parallel.
Using their hands to immediately create something
physical facilitates a much richer understanding than

purely mediated work. Especially the spatial possibil- of building. Apart from versatility with current design
ities are - as a matter of course, perhaps - much easier tools, conceptual understanding and a reversal of the
to explore in a three-dimensional medium (physical direction of thought are necessary.
models) than in a multi-dimensional medium (CAD)
that is, however, forced through a two-dimensional Figure 4
means of representation (computer screens). Experimental
The second part of the course consists of the de- pavilion design,
sign of a simple pavilion (Fgures 3-5). It should be sit- Physical and Digital
uated near our architecture faculty's buildings, con- Models, Student
tain a summer café and multifunctional gathering Project by Zeinab
space for students and visitors and serve as a physical Shehin and Jasmin
display for the ongoings with the faculty. This design Wilkens, Aachen
should consist of individual components that are var- School of
ied, and these variations should be systematized. The Architecture.
students are to start with a geometrical shape that
Figure 5
fulfills the desired functions and then develop a sys-
Experimental
tem for the components. In order to be able to fo-
pavilion design,
cus on working on and with the systems, we restrict
Physical Models
the building's functionality to that of an enclosure, a
showing
shell, without the need for internally separated indi-
parametrized
vidual rooms. In the design of the pavilions the stu-
system component
dents employ their newly learned skills in parametric
and advanced geometrical modelling and work back
REFERENCES and whole
Buri, B. and Weinand, Y. 2008 'Origami - Folded Plate assembly, Student
and forth between building design and systems de- Structure, Architecture', Proceedings of the 10th Project by Hanieh
sign. Eventually, the designs are build in medium- World Conference on Timber Engineering, Miyazaki Erden and Hanieh
scale physical models, and CAM methods used. Engel, H. 1997, Tragsysteme I Structure Systems, Gerd
Hatje, Ostfildern Yousefipak, Aachen
Outlook Iwamoto, L. 2009, Digital Fabrications - Architectural And School of
In the future, we aim to extend the work in two spec- Material Techniques, Princeton Architectural Press, Architecture.
New York
tra: Firstly the size and performance of the individ-
Matcha, H. 2015 'Parametrized Systems: Conceiving of
ual building components. So far, they are merely able Buildings as Assemblies of Varied Parts', Proceedings
to be assembled into two-dimensional enclosures of of eCAADe 2015, Vienna
one specific size and use. We want the systems to Matcha, H. and Barczik, G. 2009, 'Productive Processes', in
create more than one shape, more than one size of Agkathidis, A. (eds) 2009, Modular Structures in De-
enclosure that can contain more than one type of sign and Architecture, BIS Publishers
Matcha, H. 2010, 'Regelbasierte Planung - Parametrik',
function. Secondly, we want to study the possibili-
in Hauschild, M. and Karzel, R. (eds) 2010, Detail
ties of systems where the components already con- Praxis: Digitale Prozesse, Institut für Internationale
tain functional space. Larger-scale CAM methods like Architektur-Dokumentation
printers or robots for example weaving a material al- Matcha, H. and Karzel, R. 2013, 'Lehre als Praxis als
low for this. Our overall goal is to free the producers Forschung: 1:1 Entwicklung von Messeständen
of architecture from the mental constraints of fordist unter Einsatz parametrisierter Software und com-
putergesteuerter Herstellung', in Pahl, K.A. (eds)
mass production and the restrictions of pre-existing
2013, ECHT?! Zum Bezug von Praxis und Lehre in der
building systems without increasing the actual cost Architekturausbildung, TUDpress

Ubiquitous Digital Repositories In the Design Studio
A Case study
Theodoros Dounas1 , Benjamin Spaeth2

1
Xian Jiaotong Liverpool University 2 Cardiff University
1
theodoros.dounas@xjtlu.edu.cn 2 SpaethA@cardiff.ac.uk
The paper investigates the usability and effect of a ubiquitous digital repository in
the architectural design process. Acknowledging the post-digital era where
students work with diverse media either digital or analogue, the project explores
the suitability of a digital log in augmenting conceptual thinking, feedback
provision and intellectual exchange by means of a studio in an architectural
undergraduate course. Students integrate a digital log into their workflow
resolving a design task of an architectural studio. A server-based repository
serves as students' individual archive as well as a share-point for peer-students'
informal exchange and tutors' feedback. The conclusion of the study is that
sketching and organization habits from the analog media the students have
learned persist even with a more digitally inclined generation. The use of digital
tools that obliterate the analog-digital division, holding the best of both worlds
are still subject to the constraints of timely introduction in the curriculum,
cultural resistance in terms of organization of a project and more so void of
experimentation in their use by students.
Keywords: digital repositories, Design Studio, hybrid media
INTRODUCTION in the project. The students were assigned a typical

The main goal of the case study was to examine the studio project representing the final year project of
effect of a ubiguitous digital log to record feedback their bachelor degree. The design tasks were generic
during design studio tutorials plus the experimen- and formed exclusively a vehicle to conduct the pre-
tation on digital sketching as a generative systems sented research; one studio explores solutions for
during design production. The paper presents the a futuristic museum in 2050 concerned with contri-
first part only of the case study concerning the use butions of foreign and migrant workers to the lo-
of the system as a digital log. The case study took cal society and the second studio develops propos-
place in Xian Jiaotong Liverpool University in the deals for a community and surf centre at the Australian
partment of Architecture in the final year of the BEng coast. The students were encouraged to integrate
architecture programme. Two groups of 9 architec- the portable tablet and the ubiquitous digital repos-
ture students at undergraduate level 3 participated itory into their design work. Students were not given

special advice how to use the tablet and the reposi- ture, organisation, plus the potential for augment-
tory, but certainly were introduced into the general ing a user's capacity to think via generative processes.
functionalities. The digital repository (EvernoteTM) Thus the project proposes to augment formative vi-
and tablet (Apple iPad AirTM), were used as students' sual feedback to the students but also introduce gen-
design log during the whole process, offer multiple erative processes through an interface that is more
functions, such as original content creation, organi- accessible than classic cad systems. Researchers in
sation features, document annotation and informa- the past have conducted protocol studies (Abdel-
tion sharing opportunities. Apart from the students mohsen et al. 2007), studies in automated under-
also the two tutors supervising are equally equipped standing of design drawings (Achten 2005) Stud-
with a tablet, pen and access to the ubiquitous repos- ied drawing sequences to deconstruct the manner
itory for each student. The platform was able to in which architects draw, (Cheng 2005) and earlier
record all sketching data, whether performed of the have tested CAD platforms as a means for architec-
tablet or not, original image files, recording of con- ture 'sketching', leading to a distinction between cre-
versations, video and photographs of models etc. ative & productive sketching and 're-creative or re-
As such it was able to capture all possible informa- productive' sketching (Daru 1991) The innovation in
tion and feedback loops during tutorials with the stu- this case does not lie in the employ of digital tools
dents. per se (Hannibal et al. 2004), but in situating the
digital sketching in an orchestrating role, early in ar-
BACKGROUND AND CONTEXT chitectural production where both teacher and stu-
Studio/Learning design through digital sketching dent use it formatively. As such the current paper
has been well documented (Cheng 2005). Digital presents the results of the ubiquitous digital repos-
sketching is a production tool already established in itory side of the project rather than the mechanics,
other disciplines like animation or graphic design, learning and output of digital sketching per se. To our
however architects continue to use analogue tools knowledge from literature the integrative approach
in their education studio mostly because of the in- to digital sketching has not been attempted before
herent conservatism of studio education, reproduc- and has the potential to provide innovative, new, in-
ing models of the late 19th-early 20th century. Ca- sights into studio education. We take the position
pable architecture educators focus more on their stu- that the analogue-digital threshold is no longer clear
dents' capacity to think rather than merely being pro- in our era in both architectural design and education
ductive in a digital environment, as some times is re- (Martens et al. 2006). The students and practitioners
quired for example in the profession. This makes the of architecture are employing hybrid blended media
architecture studio an educational setting hostage to from the beginning to the end of a project in the anal-
the skills that students learn on their own, at times ysis, composition, generative and production phases.
enabling learning or at times constraining it within
the confines of specific software. Sketching however Digital repository and log
is an interface that most students know and develop The main platform that we tested in the project is
within their education, while tutors persist in provid- EvernoteTM, as a 'capture everything' repository of
ing formative visual feedback using sketches. Sketch- the design studio. Examining sketching as an or-
ing, even of questionable aesthetics or skill, provides ganization and production tool, we equipped the
an immediate interface to express thoughts and con- students also with Ipad tablets and pressure sensi-
cepts in architecture, making it ideal as the platform tive pens from Wacom. The tablets were not essen-
of communication in the tutor-student relationship. tial for input as all other devices compatible with
Digital tools, from journals to CAD provide struc- Evernote were used such as mobile phones, desk-

top and laptop computers etc. The inclusion of based on. The first hypothesis we developed was that
hand drawing was one notable feature in the sense the amount of digital sketching would increase with
that students could either take high resolution pho- the availability of the tablet and the ubiquitous digi-
tographs of their sketches or scan their hand drawn tal repository. We also as a second hypothesis spec-
sketches an upload into Evernote. An external ser- ulated that the students of this generation are digital
vice provider hosting unlimited space for clients' use natives so they would integrate not only the tablet
provides the server based digital log. The tablets but also other digital devices at their hands such as
have a client software package (EvernoteTM 5.8.8.x) mobile phones. Furthermore we assumed that the
installed. The client is integrated into other software students are handy in the integral use of different
packages on the tablet thus information and docu- types of media and their hybridisation. As a conse-
ments processed in other applications can be trans- quence of the ubiquitous availability of information
ferred, stored and edited within the digital design and the unlimited opportunity of collecting informa-
log. The server-client system can handle all types of tion, inspiration and feedback we assumed that stu-
media such as pictures, text documents, audio and dents will take advantage by increased frequency of
video recordings. The Evernote clients provide direct feedback requests and shared work in progress, in
annotation opportunities to text or picture/image peer groups. The possibilities in organising informa-
documents. Pictures from the camera i.e. can be dition on different levels and the implicit opportunities
rectly transferred to Evernote and edited and shared to structure the design process explicitly in different
within the application. Evernote can also directly ways suggest that this might have impact on the de-
process screen shots and webcam captures. The soft- sign process or on the awareness of the non-linear
ware also offers organisation opportunities to struc- nature of design processes. In essence we hypothe-
ture project information and design processes. Notes sised that we would see many different modes of use
can be titled and assigned to subject areas. Students of digital sketching and many different models of or-
also can tag specific items with different tags. The ganising information and output in the design stu-
graphic interface allows filtering and organising the dio.
information according to the multilevel structure ap-
plied to the project and its related notes respectively. RESEARCH METHOD
A chat room allows students and tutors to discuss up- The method we used for validation was a triangula-
loaded notes online. Sophisticated sharing options tion method from the experience and opinion of the
enable students and tutors to control accessibility for students participating in the project, the grounded
notes and chats individually. Students were offered a observation of the tutors through the project and
general introduction to the system to understand the the evaluation of the full data set produced through
working principles and potentials of the proposed the project. 'Methods triangulation' (Perone et al.
systems. Students were not advised specifically how 2003) refers to employing different qualitative and
to use the system since students developing individ- quantitative research methods so that the hypoth-
ually different approaches and uses of the system is a esis is examined from multiple vantage points and
major objective of the research. Therefore only a gen- dimensions. The purpose in using a methodolog-
eral introduction to the functionalities of the system ical triangulation to examine the hypothesis is to
was given. compensate and overcome the weaknesses inher-
ent in each single social qualitative method of re-
HYPOTHESIS search. (Maginn 2008) It also leads to more reliable
At the beginning of the project we made a series of research findings in social sciences as triangulation
hypotheses which the project and its methods were allows for the design of a multifaceted, cross vali-

dated research project (Carey 2009). This entails a design stages is rated with an average of 4.1. Qual-
combination of qualitative and quantitative meth- itative comments reveal sketching to be mostly un-
ods. The first method was based on running a de- derstood as a tool of communication. Most students
tailed questionnaire through the 18 participants in emphasise the importance of sketches to make ideas
the project. (Maginn 2008). The second method was or processes explicit to others; in particular to share
based on grounded observations of the researchers their work with tutors. Some also understand sketch-
during the tutorials and the development of the fi- ing as a way to structure design thinking in partic-
nal year project using the platform (Carey 2009). The ular and design processes in general for their own
third method was based on the quantitative and purposes. Few students relate their sketching with
qualitative validation of the first two methods, but inspiration. Four students explicitly refer to sketch-
due to space constraints is only present as sum- ing as a mean to keep record of their design ideas.
marised evaluations in the discussion. It is essen- In reflection of how this communicative aspect of
tially an examination of the artefacts produced by the sketching was captured into evernote according to
students in terms of qualitative classification of their the students, figure 3 shows that the main uses were
work, verification of their understanding of the use to maintain a feedback loop with the tutor and but
of the platform by looking into the combined results also as a repository of all files.
of their project work and the sketching feedback, the 9 students used the Evernote log as a diary where
measurement of frequency of commit logs in terms the design process and tutor feedback was organised
of the devices used etc. continuously time based. 5 students organised their
log not according to timelines but implemented dis-
STUDENT FEEDBACK tinct design ideas as the underlying organisational
Retrospectively students are asked to reflect on their structure. The remaining 4 students were not ex-
experience and on the impact of the proposed plicit in their reflection on their organisational use
approach to their design approach and sketching of the digital repository.On the comparison between
habits. The questionnaire was completed by 18 stu- the organisational model of digital repository and
dents (100% of the participants). The online ques- process to the common manual approach students
tionnaire included 17 questions. Qualitative ques- referred more to the difference in the nature of the
tions are set with ratings from 1 to 5, where the rat- materials rather the organisational model. However
ing always is qualified with a description respectively. students appreciated the feasibility and ubiquitous
All quantitative questions are followed up with a field availability of the produced material if included in the
for qualitative specifications on the qualitative ques- digital repository. Others emphasise again the qual-
tion. Figure 1 presents the answers to the first ques- ity of the repository as a means of communication.
tion, on what kind of log method the students used Few reflected on the repository as a means to ease
before they participated in this study. exploration of "fragmented design ideas" which may
All students reported that their previous studio be lost in a physical archive, as it can be seen in figure
was predominately organised by a physical sketch 4.
book where students either bound loose collections Physical work was mostly included into the
of paper or directly sketched into the book. On the repository through photographs or scans. Only one
question what role sketching plays in the design pro- student reported on editing and working actively
cess, without distinction between digital or manual with the scans and photographs retrospectively the
sketching, most students underline the importance upload. One student seems to completely omit the
of sketching in design. (Figure 2) use of physical models and reports that all his work
The importance of sketching in early conceptual was only created digitally. In general the students

Figure 1
Dominating log
method used
before the case
study.
have not identified a significant difference in their installed software accordingly. All students have in-
organisational model applied in previous physical stalled software to ease sketching on the pad. Some
repository to the now digital one. All emphasise have several applications other just went with the
again the ubiquitous and non-physical character of recommended one. 5 students admitted that they
the repository enabling easy transport and availabil- have installed also software for entertainment or per-
ity of the design material. All students have com- sonal use. Only 2 students have installed technical
plemented the tablet software selection according drawing software packages and only 1 person report
to their needs. Seven students explicitly report that on mind map software to support the structuring and
they have use the tablet for reading and therefore organisation of ideas and thoughts. Almost two third
Figure 2
The role of
sketching and
drawing in the
design process.

Figure 3
How students used
Evernote.
of the students (11) used the repository also for self- would take place. These uploads where mostly pa-
organising purposes such as to-do-lists, reminders or per sketches or photographs of models. Very seldom
timetables. 4 students even reported to use the dig- the sketches or photographs were worked over in the
ital repository not at all for self-organisation. The re- digital sketching application. We could also observe
maining 4 made use of the self-organising capacities that, despite the students' contrary confirmation in
only occasionally or seldom. On a scale from 1 to 5 the questionnaires, students had difficulties to actu-
("not helpful at all" to "extremely helpful") the aver- ally orient well in their repository. We have observed
age rating for the helpfulness of the proposed to the that in most cases the students had such difficulties
design process was established at 4 ("very helpful"). because of missing organisational structure where
Qualified comments on this question again empha- tags, titles and subject assignments were not used.
sise the feasibility and all time availability of the cre- A restricting factor in the whole process was indeed
ated material. Students who didn't rate high on the the internet connection. Since the sketches drawings
helpfulness unfortunately weren't able to articulate and pictures demand a minimum level of quality an
the shortcomings in a qualified way. Also sometimes immediate upload and editing process was mostly
qualitative comments are not or not completely in not possible. Only students who were well prepared
accordance to the qualitative numerical rating. An in advance, uploading their material constantly and
overall overview of how physical design utput was in- not just immediately before the tutorial could ben-
cluded into Evernote is presented in Figure 5. efit from direct comments, annotations and correc-
tions in the digital repository. Where digital interac-
TUTORIALS OBSERVATION tion was not possible, tutorials took place in a usual
As reported by most of the students in the ques- way on paper and the output was retroactively then
tionnaire the digital archive was foremost used as to scanned and fed into Evernote. This interruption in
share their work in progress with the tutors. It was the process and the reluctance of some students to
evident from discussions, the tutorial work, and con- actually continuously use the repository as a working
firmed by the questionnaires that students very often platform concluded in some cases in a fragmented
uploaded their work imminently before the tutorial record of works and feedback. The use of the tablet

Figure 4
Modes of
organisation in
Evernote.
Figure 5
How was physical
design output
included into
Evernote.

as the main means in and the basis for tutorials in- tive if introduced early and en mass in architectural
deed needed some time for adaption not only for the education. On the other hand the experiment did
students but also for the tutors. Doubtlessly the dig- show that the students responded positively when
ital repository represented a great advance in moni- provided with a repository and documentation sys-
toring students' progress, when upload was actually tem that captures all information in their design pro-
conducted. And the absence of uploads was in most cess. Even if they did not experiment extensively they
cases also a helpful indicator for students facing chal- found the fact that no information was lost, even go-
lenges and obstacles in their design advances. In the ing back to discarded sketches that otherwise would
face to face tutorials we observed that student and be lost during the design development, an occur-
tutors equally referred not only to the digital and digi- rence that was rarely happening in the past. We
tised material but also and constantly to the physical also found that since the students worked individu-
representations, making jumps between the digital ally and not in teams peer feedback was not picked
and the physical. up within the system, but relied in classic fashion in
verbal feedback between the students. One other as-
RESULTS AND DISCUSSION pect where the students did not take advantage of
We started the project with high hopes that the tools the features the platform offers is exactly the record-
were now in place to develop completely digital pro- ing of voice and sound. As such there are almost
cesses in the architectural education studio. How- no sound recordings of tutorials or discussions in the
ever the students were also too entrenched in modes repositories. The provision of feedback stayed within
of operation and organization of production suited the institutionalized tutorial system, where 60% of
in analogue processes. This created the situation the students prepared their work right before the tu-
where a percentage of the students experimented in torials rather than in a continuous fashion. However
some areas of content creation and feedback, some the remaining 40% who did take advantage of a con-
others in making efforts to incorporate a linear as- stant feedback loop provided by the digital reposi-
pect of organization or a folder like organization in tory system considered this an advantage as they re-
their projects. None made a totally effective use of ceived more continuous feedback and were not re-
the software in a manner that took advantage of the liant only on the contact time. Another positive of
complete potential of the platform: For example us- the system is that the tutor has the ability to overview
ing links between notes to organize relevant ideas the whole production of the student, and situations
and sketches rather than folders, to create a tagging where the student falsely hides information thinking
system that then would rely on search to find exter- they are not important are avoided. In the same vein,
nal references and relevant sketches or images. Al- students commented positively on creating their fi-
though we expected this generation, as digital na- nal presentation by having a consistent format and
tives and familiar with touch interfaces and mobile archive for all of their work, as it was already digi-
devices from an early age on, to grasp intuitively the tized and ready to be placed into production. Further
potential of an ubiquitous platform, and to engage afield we do see the value of an 'always on, always
and create novel pipelines of information and organi- accessible' digital design repository and it would be
zation in the studio, the students remained in a kind great if we could use one with even more capacity
of traditional studio behaviour. This effectively com- and capability in integration. Some ironically could
municates that habits are entrenched from early on call this a BIM however BIM products do not currently
in the design studio and are difficult to shift, thus have the directness, ease of use and lightness of a
suggesting that the introduction of afore described note taking system like Evernote. Concluding the
methods in the design studio would be most effec- observations the advance of ubiquitous and mobile

digital devices in our everyday lives do not allow to Cheng, N.Y.W 2004 'Stroke sequence in digital sketch-
assume their use in the professional or educational ing', Architecture in the Network Society [22nd eCAADe
environment. Even when the hardware is provided Conference Proceedings], Copenhagen, Denmark,
pp. 387-393
there seems to be a certain hesitation to use these
Cheng, N.Y.W and Andrew, M 2003 'Digital Curricula: Ef-
technologies i.e. in the design studio. Whether this fective Integration Of Digital Courses - A Delicate
is due to already established habits and design rou- Balance', Connecting >> Crossroads of Digital Dis-
tines or whether the methods are considered inferior course [Proceedings of the 2003 Annual Conference of
to the established methods could not be uncovered the Association for Computer Aided Design In Architec-
in this study. Yet the fact that although the digital ture], Indianapolis,Indiana, USA., pp. 129-130
Cheng, N.Y.W and Andrew, M 2005 'Learning Design pro-
devices were used but only to emulate traditional de-
cesses with digital sketching:copying graphic pro-
sign methods to a digital media demonstrated the ur- cesses from animations and storyboards', CAAD fu-
gent need to integrate a professional use of ubiqui- tures 2005, Vienna Austria, pp. 291-300
tous media and tools into the design teaching. Daru, R 1991 'Sketch as Sketch Can - Design Sketch-
ing with Imperfect Aids and Sketchpads of the Fu-
ture', Experiences with CAAD in Education and Practice
ACKNOWLEDGEMENTS [eCAADe Conference Proceedings], Munich, Germany
We would like to acknowledge the Teaching Devel- Hannibal, C, Knight, M and Brown, AGP 2004 'How Useful
opment Fund at Xian Jiaotong Liverpool University is the Digital Sketch?', Architecture in the Network So-
and colleagues at the Academic Enhancement Of- ciety [22nd eCAADe Conference Proceedings], Copen-
fice for supporting the project 'Digital Sketching as hagen, Denmark, pp. 374-379
Maginn, JP (eds) 2008, Qualitative Housing Analysis:
a Generative System' that this paper is based on.
an international perspective, Studies in qualitative
We would also like to acknowledge the support of Methodology, Volume 10, JAI Press, Emerald Publish-
our colleagues Austin Williams and Marian Macken ing group, Howard House, Wagon Lane, UK
for providing critical comments and support on this Martens, B, Mark, E and Cheng, N.Y 2006 'Thresholds be-
work. The project 'Digital Sketching as a Generative tween Analog and Digital Representations.', Commu-
System' was recognised by Xian Jiaotong Liverpool nicating Space(s) [24th eCAADe Conference Proceed-
ings ], Volos, Greece, pp. 372-383
University with the 'Most Innovative Teaching Prac-
Perone, PJ and Tucker, TL 2003, An exploration of triangu-
tice Award 2014-2015' after it was nominated by the lation of methodologies: quantitative and qualitative
students. methodology fusion in an investigation of perceptions
of transit safety, National Center for Transportation
Research, Tampa, FL, USA
REFERENCES
Abdelmohsen, S and Yi-Luen, E 2007 'Tracking De-
sign Development through Decomposing Sketch-
ing Processes', IASDR 2007, Emerging Trends in De-
sign Research, Hong Kong Polytechnic University
School of Design, Hong Kong
Achten, H 2005 'Resolving some Ambiguities in Real-
time Design Drawing Recognition by means of a
Decision Tree for Agents', Computer Aided Architec-
tural Design Futures 2005 [Proceedings of the 11th In-
ternational Conference on Computer Aided Architec-
tural Design Futures ], Vienna, Austria, pp. 311-320
Carey, M 2009, The Social Work Dissertation, Open Univer-
sity Press, McGraw-Hill House, Berkshire, England,
UK

CAAD EDUCATION | Design Tools
Computational Thinking and the Architectural Curriculum
Simple to Complex or Complex to Simple?
Guzden Varinlioglu1 , Suheyla Muge Halici2 , Sema Alacam3

1
Izmir University of Economics 2,3 Istanbul Technical University
1,3
{guzdenv|semosphere}@gmail.com 2 smhalici@hotmail.com
Recent trends in architectural education and practice have encouraged the use of
computational tools and methods for solving complex design problems. Newer
technology can augment the design process by applying progressively
more-advanced computational tools. However, the complex nature of these tools
can lead to students getting lost at the skill-building stage, they can become
trapped in computational design terminology, leading to designs of limited spatial
quality. This paper introduces a pilot study from Izmir University of Economics
(IUE) for the integration of computational design technology in the
undergraduate architectural curricula, based on a workshop series using a
top-down teaching strategy.
Keywords: Basic design, learning outcomes, keyword analysis, visual scripting

environment (VSE)
INTRODUCTION and Yazar (2007) present a graduate course entitled

With the widespread use of digital media, the inte- "Designing the Design" focusing on computational
gration of digital tools with architectural education thinking and the new emerging language and meth-
and discourse is a common topic for design educa- ods of design. However, few studies display the out-
tors. Open-source parametric and algorithmic design come of student learning via digital design and fabri-
tools are pushing the architectural design process by cation courses which focus on the junior years of ar-
challenging traditional pedagogical methods. Sev- chitectural design education. Early interaction with
eral studies have been conducted in search of the digital tools in the early stages of architectural cur-
extents and methods by which digital tools are inte- ricula can serve to shape the remainder of the edu-
grated to the architectural curriculum. Oxman (2008) cation. To evaluate the computational approach for
presented an experimental design research studio on foundational and architectural design education, this
"the exploration of digital architectural concepts" as paper aims to make a discourse analysis based on
a pedagogical framework for educating the digital the interrelations between the process and the out-
architect. Ozkar (2007) offers a design teaching ap- come of innovative student work from a workshop
proach of "learning by doing" by integrating the no- series for first year basic design and architectural de-
tions of design thinking and computing to the first sign students at Izmir University of Economics (IUE).
year of architectural design education. Colakoglu
CAAD EDUCATION | Design Tools - Volume 1 - eCAADe 34 | 253

OBJECTIVES AND METHODS With this in mind, we explored two directions of skill
For several decades, design schools have incorpo- learning: top-down and bottom-up. As defined in
rated CAD with design studio and non-studio sub- Sun et al. (2001), top-down learning goes from ex-
jects (Mark et al. 2001). Some schools introduce plicit to implicit knowledge, while bottom-up learn-
digital technology within isolated courses in order ing is the opposite; going from implicit to explicit
to computerize drafting skills, while others integrate knowledge. In our case we wanted to explore the in-
computation to design studios. Kvan et al. (2004) teractions between the two types of learning, partic-
identify four levels of digital media interaction in ularly in terms of one type giving rise to the other.
design curricula: general computer use, application As part of an internal review of teaching practices
of digital media, engagement of design computa- within the architectural department of IUE, we im-
tion, and advanced exploration of design theory and plemented a set of intensive one-day workshops for
methods. Concerning these levels of interaction digital design, requiring no prior knowledge of soft-
with digital media, case studies at several universi- ware language. For the first workshop, we posed
ties for graduate (Schieck 2008) and undergraduate a simple first-year design problem to be solved by
programs (Duarte et al. 2012) expound the teach- relatively complex digital means including the digi-
ing strategies and their merits. In Duarte, Celani tal fabrication of each model. Not only did the end
and Pupo's article, a comparison of two universities results prove more than satisfactory, the students
with alternative methods for the integration of dig- themselves quickly lost their initial nervousness and
ital tools in design education display which strat- reported enjoying the tasks.
egy to follow while changing the architectural cur-
riculum. Referring to Duarte's two examples, he de- BACKGROUND
fines a discrete studio model with an updated com- The Faculty of Fine Arts and Design at IUE has a
puter curriculum versus the traditional curricula with first year curriculum based on an interdisciplinary
a bottom-up approach. However, it is not clear studio set-up for all departments; fashion, architec-
how the basic structures of design curricula need to ture, interior architecture, industrial design, and vi-
change in order to effectively incorporate state-of- sual communication design. Students in the first year
the-art technologies. of their design education will experiment with mate-
How do we allow for transformations that rials and tools for the first time as, in Turkish educa-
would satisfy current advances in digital technolo- tion, there are no basic design classes prior to univer-
gies within an established curriculum? One school of sity. The design studio is at the core of architectural
thought is that such additions be piled-on towards education, where students are given design prob-
the end, when the basics of the traditional curricu- lems which have to be solved holistically through
lum have been covered thoroughly; the logic being background research and explorations in two and
that students are sufficiently equipped in traditional three dimensional media. In the basic design curricu-
design methods to begin applying advanced digital lum, shapes, forms, materials and tools are abstract
tools. More and more however, as these technologi- rather than referring to professional conventions. Ba-
cal advances are constantly being further developed, sic design courses explore abstract formal composi-
we find students struggling to comprehend, never- tion, while technical courses emphasize new mate-
mind execute, the surrounding language. It would rials and processes as inspiration for innovative ar-
appear, regardless of the connectivity between architectural forms and structures. Independent from
chitectural design and its digital methods, that the professional disciplines, the curriculum is based on
skills of one are in fact rather remote from those of teaching the designerly way of thinking. This peda-
the other. gogy "learning by doing" is based on hands-on expe-
254 | eCAADe 34 - CAAD EDUCATION | Design Tools - Volume 1

rience for deeper understanding of creative skills and to the emphasis on interdisciplinarity and transdis-
designerly knowing (Cross 2001). ciplinarity for first year, the second year design stu-
Learning by doing, or experiential learning, is a dio focuses on spatial relations, functional require-
pedagogy based on hands-on experience rather than ments and structural integrity, on a more architec-
uniform second-hand lectures to form a creative indi- tural design language. Following the same Bauhaus
vidual (Ozkar 2007). Ozkar's pedagogical framework, pedagogical methodology, second year gradually in-
based on the Middle East Technical University (METU) troduces the architectural thinking. However, digital
basic design program, describes the pedagogy that tools are rarely introduced in the studio classes, but
we follow in IUE. Following Bauhaus pedagogical ap- rather as separate supportive courses.
proaches developed by Walter Gropius in 1922, the To date, computational thinking is not fully in-
first year of basic training in the Art and Design Studio tegrated to the curricula. The interdisciplinary ap-
FFD 101/102, students experiment with color, shape proach has brought interactive pedagogical teach-
and materials, with no specific functional or spatial ing methods and a need for independent profes-
requirements (Kvan et al. 2004). Running parallel sional classes for each discipline leading inevitably
with this studio class, another studio approach is In- to a discrete studio model. This pedagogical model
troduction to Architecture (ARCH 101/102) which fol- brought the disadvantage of CAD/CAM courses be-
lows the Ecole des Beaux Arts approach. Often, stu- ing conducted independently of the main design stu-
dents are given a group of precedent buildings to dio, leaving computing as an add-on. A good ex-
analyze through plans, sections, and elevations from ample would be the FFD Computer Aided Techni-
which they are expected to build scale models. The cal Drawing course which teaches technical drawing
students learn from precedents, not only evaluated skills to students via computer. Based on student
by skills and mastery of representation tools, but also evaluation of the separate first year CAD classes and
by articulating the formal, spatial, and material con- observations made in their second-year architectural
tent of architectural ideas. In FFD 101/102, students design studio, we decided to search for ways of in-
are encouraged to produce their own creative de- tegrating CAD/CAM tools into the design studio. To
signs based on their subjective perceptions, as op- avoid the disadvantages of the discrete studio model,
posed to ARCH 101/102, where students draw inspi- computational design tools and methods were ap-
ration from existing designers and architects. plied in one-day workshops.
In terms of course schedule and credits, the em-
phasis is placed on FFD 101/102. A deeper insight WORKSHOPS
to "learning by doing" will help us to explore the The first-year workshop "magnetospheric(de) forma-
teaching strategy in our university. A design studio tions" was conducted as a single unit basic design
is an atelier "environment where students test out studio curriculum for all design students (Varinlioglu
theories, ideas, materials, constructions, and similar et al. 2015). The challenge was to offer the same
productions, as part of their design process" (Ozkar complex computational design problem to all stu-
2007). Directly introducing tools and materials, this dents, having no prior experience in VSE or with
pedagogical model aspires to develop the craft that digital fabrication. For this workshop, we explored
a designer needs by testing visual and spatial quali- cyclic relations among the computational definitions
ties. Following the discrete model curriculum, tradi- of magnetic force, their applications to the form find-
tional design tools such as technical drawing, card- ing process, and the fabrication of digital models by
board modeling and free-hand sketching, are intro- laser cutter. Given the simple design scenario of an
duced to students in separate first-year classes which outdoor park, students were asked to design a chil-
enable them to work on basic design tasks. Due dren's playground by deconstructing and exploring

Figure 1
Process of student
work from the first
workshop.
the holistic algorithm of a magnetic field. With a continued with parametric modeling as an augmen-
scenario borrowed from the more complex gradu- tation of the first workshop. This intensive one
ate level, we introduced the definition of magnetic day workshop focused on altering part of the Izmir
behavior with few contextual requirements or er- cityscape, the silhouette of the historical "Asansor"
gonomic concerns, within the Grasshopper VSE. Pre- region. This site is well-known for its urban elevator, a
cooked Grasshopper definitions and the given de- functional historical structure that connects two lev-
sign task allowed students to explore computational els of the city. Today, besides functional usage, the
analytical thinking and designing methods, while Asansor also serves as a touristic destination, but the
pushing them to experiment with forms by changing bare cliff still remains as a background to the site. The
parameters. The workshop would only accept digital objective of this study was to design vertical circu-
fabrications created with the CAD/CAM tools avail- lation elements in urban scale and to decide on the
able in the modeling lab. These constraints allowed level of intrusion with respect to topography, land-
students to focus on exploring iterations of the free- scape, urbanscape and historical cityscape (Figure 2).
form design, while introducing a level of freedom as For this workshop, requiring a little experience
the craft of model making would be augmented by in VSE's and digital fabrication tools, students were
the CAD/CAM tools (Figure 1). given a slightly more complex design problem. Us-
The second-year workshop "VERTICalgorithm" ing the same VSE as for the first workshop, the fo-
Figure 2
Design and
fabrication process
from the second
workshop.

cus was on making renovations to the vertical back- The goal of the questionnaire was to emphasize dis-
drop of the historical region. In contrast to the first course analysis based on keywords. We asked par-
workshop, students had to deal with a given topog- ticipants to evaluate all eleven team projects based
raphy within an urban context related to their ar- on the presentation boards shared on the workshop
chitectural design studio project. Borrowed from a blog. The presentation boards included hand-drawn
graduate level computational design studio, five key- sketches and models, the initial grasshopper defini-
words were given for this second workshop: force tions, and the final CAD/CAM models which were
(magnetism), seed, fold, parasite, and melt. The pre- superimposed on a physical model of the histori-
cooked Grasshopper definitions were given for three cal landscape. In addition, open-ended questions
keywords (force, melt and parasite), while the rest of were included in order to reveal patterns of the stu-
the groups searched and created their own VSE defi- dents' learning experience, as based on the key-
nitions. Besides five compulsory keywords from com- words. The students were then asked to analyze
putational design, we mentioned additional archi- their own work, as well as each-others work with
tectural/design keywords to help students develop these given keywords. Further questions focused
initial design ideas. These keywords were paired with on analyzing the work through basic-design versus
similar keywords from computational design termi- architectural-design terminology. The jury evalua-
nology, as in: cityscape/urban context, architectonic- tion of students' work included four dimensions of
s/form, swarm behavior/pedestrian movement (Fig- the aesthetic, functional, environmental and compu-
ure 3). tational aspects.
INSTRUMENTS RESULTS
To observe the level of students' learning, data was The expected result of the first workshop was to
collected on their final products by four methods: equip students with analytic thinking through CAD/-
CAM tools, without colliding with their design skills.
• Observing students' work process This was achieved by a top-down teaching approach
• Completing questionnaires after students' fi- for computer tools and a bottom-up approach in de-
nal presentations signerly thinking. Simple design problems with few
• Conducting informal interviews with partici- constraints allowed students to concentrate on the
pants at the end of the desk critique skill building process. The second workshop concen-
• Having jury members evaluate students work trated on design thinking with relatively complex de-
sign problems. Thus, from complex to simple, the VSE
allows the flexibility of dealing with more complex
Figure 3
Keyword analysis
matrix based on the
questionnaire of
the second
workshop.

Figure 4
Two model
outcomes
generated from the
same pre-cooked
definition.
design problems in second year. could accurately infer that they used the keyword
Figure 4 shows the pre-cooked Grasshopper def- "force". The second group, resulted in an evolution to
inition given to two different groups. The first group the pre-cooked definition, and the reviewer students
(group 1 "blocks" in Figure 3) researched various al- were unable to say that they used the keyword force.
ternatives within the same pre-cooked definition, A further analysis was based on two keywords refer-
while the second group (group 5 "experience" in Fig- ring to the same concept, from basic design termi-
ure 3) used minimum modification to this definition nology and complex computational thinking respec-
by simply changing parameters and focusing on the tively. As expected, students were seen to use famil-
digital fabrication outcome. As a result, the first iar terminology rather than those newly learned or
group augmented the definition to a more spatial implemented from the workshops. In all, the tech-
level within the urban context, but had difficulty in nology helped produce fairly good computational re-
the digital fabrication stage. The second group tar- sults even if, at this initial stage, some students were
geted the outcome much faster but ended up with using the tools without appearing to build a great
poor spatial quality. Based on the learning outcomes deal upon the architectural discourse.
of the second workshop, the variety of the keywords
given lead to free-form exploration of design ideas, CONCLUSION
however creates no spatial diversity. By means of a series of one-day intensive workshops,
Later, the questionnaire proved that the first without prior training, groups were introduced to
group with little modification was a stronger exam- the basic concepts of parametric design thinking and
ple of targeting the keyword as the reviewer students CAD/CAM tools. These groups then had the opportu-

nity to apply this knowledge to developing solutions ACKNOWLEDGEMENT
for the design problems posed. Having neither the We would like to thank, Bahar Durmaz, head of the
architectural nor computational experience to solve department and the design studio / CAD instructors,
design problems independently, the combination of Burkay Pasin, Ozgur Genca, Hugh Clarke, Lale Basarir,
newly acquired skills was applied to great effect and Burcu Kor, Filiz Keyder Ozkan, Ozge Hazer, Ethem
satisfactory solutions were created from this bottom- Gurer.
up approach. However, when asked to choose from
more than one computational design keyword, they REFERENCES
struggled to develop their architectural design so- Colakoglu, B and Yazar, T 2007, 'An Innovative Design Ed-
lutions or to fabricate good physical models. Thus, ucation Approach: Computational Design teaching
starting designs from single keywords and VSE defini- for Architecture', METU Journal of Architecture, 24(2),
tions resulted in more competitive and creative out- pp. 159-162
comes during the design process. In that sense, we Cross, N 2001, 'Designerly Way of Knowing: Design Disci-
pline versus Design Science', Design Issues, 17(3), pp.
propose to integrate the bottom-up approach which
49-55
proved so successful in the first workshop into our fu- Duarte, J P, Celani, G and Pupo, R 2012, 'Inserting Com-
ture curriculum. putational Technologies in Architectural Curricula',
Looking towards the future, we aim to continue in Gu, N and Wang, X (eds) 2012, Computational De-
with these workshops alongside the normal architec- sign Methods and Technologies: Applications in CAD,
tural curriculum in an effort to ease computational CAM and CAE Education: Applications in CAD, CAM
and CAE Education, IGI Global, pp. 390-411
design methods into the everyday thinking of stu-
Kvan, T, Oxman, R and Martens, B 2004, 'Ditch-
dents. To this end we are updating our curriculum to ing the Dinosaur: Redefining the Role
include digital design classes at a basic design level. of Digital Media in Education', Interna-
Our initial effort of including computational thinking tional Journal of Design Computing, 7, p.
into the architectural curriculum was through an up- http://www.arch.virginia.edu/arch541/dinosaur.html
date in the computer aided classes, run separately Mark, E, Martens, B and Oxman, R 2001 'The Ideal Com-
puter Curriculum', 19th ECAADe Conference Proceed-
from the design studio. Within our curriculum, an
ings, pp. 168-175
initial stage which includes computational tools to Oxman, R 2008, 'Digital Architecture as a Challenge for
first year students is achieved by the FFD Computer Design Pedagogy: Theory, Knowledge, Models and
Aided Technical Drawing course. As the name im- Medium', Design Studies, 29(2), pp. 99-120
plies, the course is intended to begin technical draw- Ozkar, M 2007 'Learning by Doing in the Age of Design
ing via computer. Besides design classes, this format Computation', Computer-Aided Architectural Design
Futures (CAADFutures), pp. 99-112
initiates the process of building computational skills.
Schieck, A F 2008 'Exploring Architectural Education in
By using a basic emerging form and by following the the Digital Age: Learning, Reflection and Fexion',
parametric modeling and technical drawing tech- 26th ECAADe Conference Proceedings, pp. 861-870
niques, we experimented with inter-representational Sun, R, Merrill, E and Peterson, T 2001, 'From Implicit
tools of modeling, VSE, coding and representation Skills to Explicit Knowledge: a Bottom-up Model of
tools in the digital environment. Based on the learn- Skill Learning', Cognitive Science, 25(2), pp. 203-244
Varinlioglu, G, Halici, S M and Alacam, S 2015 'Computa-
ing outcomes of the students, a simple design prob-
tional Approaches for Basic Design Education: Ped-
lem was shown to help students explore computa- agogical Notes Based on an Intense Student Work-
tional thinking. In the coming years, a more thorough shop', 19th SIGRADI Conference Proceedings, pp. 576-
analysis on students' learning outcomes will be re- 580
peated with this sample group of students until their
graduation.

Teaching Programming and Urban Complexity to
Architecture Students
Lukas Treyer1 , Daniel Zünd2
1,2
Chair of Information Architecture, ETH Zürich
1,2
http://ia.arch.ethz.ch
1,2
{treyer|zuend}@arch.ethz.ch
In this paper we present a framework that allows to introduce architecture

students to agent-based simulations in the context of urban planning. It provides
them with an understanding of how such simulations work by instructing them to
learn how to program and develop an agent on their own. Along with the
framework we explain our didactic concept of teaching
complexity-science-methods to students from other fields such as architecture. In
the discussion we report on that theory and practise should be alternated at very
short intervals. Additionally we emphasize the importance to teach a good
understanding of the capabilities of modelling and simulation tools, since
uneducated students tend to trust them too blindly.
Keywords: Programming, Crowd Dynamics, Urban Planning
INTRODUCTION ulation reflect, amongst others, visitors of the festival

Urban planning is a task dealing with many uncer- with different needs.
tainties (Mustafa et al. 2015). It is hard to predict the In the course of the lecture series, students learn
impact of decisions and the consequences of deci- how to use an agent-based modelling and simulation
sions made. With modern technology, it is possible framework, and all the tools that are needed to use
to use the increasing computing power to reduce unit. This means that students first are taught the pro-
certainties and to better understand the urban sys- gramming language that is later used in the agent-
tem of interest. based framework. Later in the lecture series, basic
This paper introduces a course held to architec- concepts of complexity science are taught and the
ture students from different semesters and with dif- change of perspective, introduced by it, is explained.
ferent backgrounds. The aim of the course is to intro- In the last part of the lecture series, the students have
duce the students to simulations and give them in- to implement their own agent.
sights into how simulations can improve the urban The implemented agents are finally run in the
design process, but also rising awareness of the dis- framework and compared against each other. To mo-
advantages of simulations. To be able to run the sim- tivate the students to invest some time into their im-
ulations on a student's computer we limit the scope plementation, the simulation framework is set up in
of the urban planning perimeter to the area of a latin a way that it is possible to directly compare the dif-
music festival in Zürich. The agents in the crowd sim- ferent implementations of the students. This is used

to run a competition with all agents in one simula- puter science, comparable to drawing plans in archi-
tion. The students with the strongest, longest living tecture; even though way less expressive when com-
agents receive a price. paring the immediate media "code" and "drawing".
So we need to prepare the architects to deal with a
DIDACTIC CONCEPT textual representation of a language that is by far less
In this section we describe the structure of the expressive than a spoken language or the graphical
course and explain the interplay between the three language they know from architectural drawings and
main parts "programming for beginners", "design of that on top of that depends on many very strict syn-
the festival layout", and "simulating complexity with tactical rules.
agents". The course breaks down into ten lessons, After explaining the building process in pro-
four of which dedicated to the introduction of programming, its concepts are introduced by focusing
gramming for beginners, two to the design of the fes- on the key aspects "data types", "control structures",
tival layout, three to the agent programming and one "grammar rules". The taught programming language
to a final presentation. is Processing/Java [1] and the students are also in-
The course addresses architecture students as an troduced to Eclipse [2] in the first interactive step-
addition to their regular design courses. It focuses by-step exercise. In the second lesson we provide
strongly on the programming rather than the dean alternative approach. Having heard about the
sign aspect. Nevertheless, we conceive it as impor- concepts that are key to the flow of a program, stu-
tant that the course includes a short design exercise dents are presented the syntactical elements: a list
to relate the simulation tools to traditional planning of keywords and operators, an introduction to blocks
methods. Also, this exercise introduces the few stu- and statements and an explanation of the meaning
dents from other disciplines such as Mechanical En- of selected characters and keywords such as ":" and
gineering and Integrated Building Systems to basic "class". The third lesson introduces object oriented
concepts of architecture and design. programming concepts since we heavily depend on
it using the simulation framework later in the course.
Programming For Beginners The third exercise accompanying this lesson asks stu-
We start the introduction to programming by point- dents to animate an object in the Processing frame-
ing out the different way of how programming tasks work. The fourth lesson on programming introduces
are tackled, compared to the traditional way archi- the simulation framework, its structure and how to
tects cope with a task. By comparing computer sci- use it in the course. In the sequence of the course this
ence with architecture we implicitly mention the de- is actually the fifth lesson with the fourth lesson giv-
bate on the relation of design and science (Cross, ing an introductory overview on simulation methods
2001) in the sense that computer science emerged and complexity science.
out of physical / mathematical science as an engi-
neering discipline while architecture in the 20th cen- Simulating Complexity With Agents
tury gradually got more and more backed by en- Before introducing the simulation framework, which
gineering sciences such as civil engineering. This is later used in the course, students are introduced to
means that both disciplines require a domain knowl- modelling, simulation, and complexity science. The
edge in order to design (software or buildings). For goal is that the students start to understand how sim-
the beginners in the course, we think it is important ulations work, and how to interpret and use them in
to mention, that the skills they learn are not suffi- urban planning. A key aspect of this part is to en-
cient to do software design. Yet they start to learn courage a critical attitude towards computational ur-
the art of programming, the craftsmanship of com- ban simulation methods and make students aware

of what complex simulations are able of and what the course and as discussed in the discussion section
not - similar to weather forecasts. They are taught this exercise is also a valuable test for the critical atti-
to treat simulations like sketches: imperfect images tude we hope to induce earlier in the course.
that communicate an idea. And we raise the aware-
ness of the verification problem of simulation mod- SIMULATION FRAMEWORK
els. Thus they learn how to interpret the results of a In this section we introduce the agent-based frame-
simulation and not just blindly trust the computer. It work for crowd simulations that was specifically im-
became clear during the design exercise, that this is plemented for this course. We are aware that already
a crucial part, when teaching modelling and simula- many agent-based modelling frameworks exist [4]
tion to laypeople. This crucial part of educating ar- that simulate crowds, but we decided against using
chitectural students and introducing them to digital an existing one for several reasons.
simulation tools is further discussed in the discussion The main reasons to implement our own agent-
section. based framework are due to its educational pur-
The last part of the course consists of implement- poses. With our own implementation we are able
ing agents for the agent-based simulation. Each stu- to show and explain the students important con-
dent has to implement his own agent. Implicitly it cepts of software architecture, without the overhead
should become apparent that they develop an agent of already existing implementations in terms of num-
not as close to a real person as possible but an agent ber of classes and configuration files. The simplicity
that deals with the constraints set by the framework of the framework and the number of lines of code
in the most efficient way. gives the students a chance to understand the whole
framework, and motivates them to do so. The frame-
Design of a Festival Layout work was implemented using Java/Eclipse [2] with
To introduce the usefulness of modelling and simu- the Proclipsing plugin [3], using Processing [1] as the
lation to support design and planning task, students visualization library.
have to plan the festival in a traditional way. The stu-
dents have to place 20 drink stands, 20 food stands, 8 Software Architecture
toilets, 3 music stages as if they would in urban plan- The architecture of the framework is implemented
ning when defining land use zones. Additionally, the to directly reflect important concepts of Object Ori-
students are divided into four groups, each of them ented Programming (OOP). Because the course did
trying to optimize the festival layout according to one not require any prerequisites from the student, the
of the following criteria: (i) flow of people; (ii) over- focus was mainly on inheritance and polymorphism,
all profit of all stands; (iii) number of happy people; which we understand as the two most important
(iv) security. As not only architects may attend the concepts of OOP. If the festival visitors are not
course and since the design task is very simplified, counted, the whole framework consist of 15 classes,
the design can be visualized using office tools such divided into three packages: the geometry package,
as MS PowerPoint or a scanned hand-drawn sketch; it which contains all the classes that are directly related
does not need to be a precise drawing. The designs to geometry; the agents package, which contains all
are then presented by explaining why they think their the different types of agents that are used in the sim-
design fulfills the assigned criteria best. The idea ulations; and the main package, which only contains
of this exercise is to relate the traditional planning the main class.
methods to the subsequent task of programming an The main class sets up the whole program and
agent that is competing against other agents in the implements the functions that are needed for pro-
given simulation framework. As it turned out during cessing, such as the setup and the draw methods,

which set up the visualisation window and draw the Figure 1
state of the simulation, respectively. Pseudocode of the
The agents of the simulation are all implemented behavior of festival
within the agents package. The implementation of visitors.
the simulation is such, that all entities that are present
in the simulation are agents, i.e. the wall elements
(Walls), the entities that provide the needs (Needs),
and the visitor of the festival (Visitors). The base class
for all the agents in the simulation is the Agent class.
It defines the basic functions that all the agents in the
framework must implement.
The Walls are agents that are immobile and they
only limit the liberty of action of the Visitors. They are
used in the simulation to imitate the street canyons
and force the visitors to stay within them. The Needs
are agents that are immobile too, but have the capa-
bility to serve certain types of needs the visitors have, The direction the visitors walk during the simulation
and, in the here described version of the framework, is calculated with a simplified version of the social
represent drink stalls, food stalls, music stages, and force model (Helbing and Molnár, 1995). We drop
toilets. The Visitors represent the people that visit the most of the terms and use only the ones that are
festival and are the only mobile agents. They have needed to calculate the before described behavior.
the four needs which are served by the four Needs Each agent i has to solve the following system of
agent type. The needs have to be fulfilled within equations to know his new direction of walking, thus
a certain time, otherwise they will leave the festival his location after the next step:
again. d~xi (t)
= ~vi (t) (1)
dt
Simulation Dynamics d~vi (t) ∑
The framework we implemented models in the micro =F ~l + ~a
F (2)
j
dt
level. This means that every agent is simulated atom- (j6=qi)
ically. We do not assume any social interaction be- where ~ xi is the location of agent i at time t, ~vi the
tween the agents, they are only self-serving and try ~l the force that pulls the agent towards his
velocity, F
to fulfill their needs. When the agents enter the fes- desired location, and F ~ a the force that repels one
j
tival, they choose the location they want to go and agent from another. F ~ a models the mass of the
j
then walk into the goals direction until they reach it. agents and ensures that they cannot walk through
Only then they choose a new location, to which then each other, but are compressible. This is an impor-
they walk . This process is repeated until one of the tant property for crowd simulations.
needs can not be fulfilled in a certain time. The ba- Figure 2 depicts the landscape of what a visitors
sic algorithm of the visitors is depicted in Figure 1. sees if he wants to go to the center of the bottom
The variables that are initialised at the beginning de- left edge. Ten other agents are present in the square
scribe the happiness score of the agents with the four (peaks) and four needs providers (hills). The layper-
needs. At the beginning they are satisfied with all son can read the image as if the agent was a marble
the needs, but they slowly decrease as the simulation put at any location on the square, and gravity pulls
progresses. on it until it reaches an exit.

Figure 2 rectly attracted by vertex 29, without going through
The forces vertex 28. The reason to have this is twofold. First
expressed by it gives the agents more degrees of freedom when
potentials in a they walk on the streets, because they are not at-
square with ten tracted by too close vertices and thus are more agile.
visitors (peaks) and The second reason is that if the agents walk strictly
four needs along the graph, also within squares, it can happen
suppliers (hills). that they walk around their final destination or along
walls. This can happen when the closest vertex to a
final destination is not directly on the shortest path
to it, in an Euclidean sense.
Figure 3
Analysis Methods
Section of the
To teach the students the value of simulation, we
festival, with the
implemented some basic analysis methods into the
walking graph for
framework. The framework allows to have real-time
the agents
feedback on how dense the crowd is at any point in
visualised. The
the festival, visualizes the locations of where visitors
numbers are the
decided to leave the festival again, counts the num-
identification
ber of customers served at each service provider, and
numbers of the
visualises the walking paths the agents took.
vertices and the
The problem that arises with the above described To get an overview on festival security it is im-
lines between them
method is when the geometry gets more compli- portant to visualize congestions. It allows to plan
the vertices of the
cated. The festival addressed in the lecture, is not in advance and understand the possible risks a lay-
graph.
only on one square but consists of two squares and out of the festival has. To understand the locations
several streets that connect them. If the agents which do no serve the visitors good enough, we im-
would walk only according to equations 1 and 2 they plemented a visualisation of the locations of where
would get stuck in corners. To circumvent this prob- people decided to leave the festival. It helps the plan-
lem, we implement the agents in such a way that they ners of the festival to identify the locations where vis-
don't walk directly to their desired location. They itors are not satisfied anymore, be it by the lack of
rather walk along a graph that guides them through needs providers or due to congestion.
the streets, i.e. the graph is set in such a way, that The third visualisation the frameworks provides
it is always possible to directly walk from one ver- is the walking paths of all the locations all the agents
tex to another, without hitting a wall. After choos- have been to. This visualisation technique gives the
ing a new location yet before starting to walk they planners an idea of where people walk and highlights
calculate the shortest path from their current loca- high frequented paths and streets. The last visuali-
tion along this graph, and then go from one vertex sation method depicts the number of customers the
to another, until they reach the desired location. To need providers served. It is of high interest for the fes-
prevent that agents walk to not needed vertices, we tival organiser to know the frequency of customers
additionally implemented an algorithm that removes for one location, for economic reasons. It allows them
vertices on the path if they are not needed. For exam- to estimate the rent prices for the locations.
ple in Figure 3, if an agent is at a location close to ver-
tex 27 and chooses to go to vertex 29, the agent is di-

EVALUATION Figure 4
For the evaluation of the course we asked the stu- Box plots of the
dents to answer a questionnaire. It consists of 33 seven discussed
questions, out of which 16 are standard questions questions showing
taken from our university's official course evaluation all survey
survey. If not otherwise stated the questions are an- participants.
swered using a scale from 1 to 5, where 1 means to-
tal disagreement and 5 total agreement. Out of the
34 students 14 participated in the survey. The sur-
vey was conducted 5 months after the end of the
course. Out of the 17 specific questions we show the
results of the following 7 in Figure 4. As shown by the
answers to D3 students had heterogeneous previous
knowledge. Also, five of the 14 survey participants Figure 5
where no architecture students. In addition to Figure Box plots of the
4 we show the results categorized by beginners/no seven discussed
beginners in Figure 5 and Figure 6. All the beginners questions showing
where architecture students. the answers of the
The results of the questions concerning the qual- seven students who
ity of the course, its hand-outs and its teachers can be took another
summarized as rather good. Hence any bad results programming class
regarding the teaching goal correlate with the course before.
design rather than its execution.
• D2 Did you profit from the Q&A sessions after
each lecture?
• D3 Did you take another programming course
before this one? (1 = no, 2 - 4 programming
was involved a bit or you didn't finish the Figure 6
course, 5 = yes) Box plots of the
• D4 Can you imagine to use the learned pro- seven discussed
gramming skills in another project in the fu- questions showing
ture? the answer of the
• D5 Do you conceive programming more like a seven beginners.
language or more like mathematics? (1 = lan-
guage, 5 = mathematics)
• D7 How would you rate your skills when con-
fronted with a new programming task such as
editing a webpage or filter data in a csv/excel
file?
• D14 Are you able to solve programming prob-
lems by yourself / by searching in the internet?
• D15 Do you prefer to learn programming by
understanding the concepts underneath or
by trying out and follow syntactical rules?

DISCUSSION mentatively, in a qualitative way. Even though the
This lecture series tries to reach the challenging goal categories do align with the values calculated by
of combining three different topics into one story- the simulation framework, we thought the students
line. By conducting the course in fall 2015 we found would understand that they should find qualitative
a few details worth discussing. arguments for their layout, that in the best case
Generally, the students were very motivated would not be covered by the values calculated by the
throughout the course. Which can be seen in the framework. Nevertheless, when doing the exercise
number of drop-outs during the semester. By expe- almost all the students ran the simulation they were
rience, about fifty percent of students do not finish introduced a lesson before, and then stated that they
courses that are teaching programming. In the case chose the layout, because it produced the best simu-
of this course, the fraction of people finishing the lec- lation results. This shows that, even thought we dis-
ture series was above 75%. cussed the significance of this kind of simulations in
Nevertheless, during the course of the lecture sean earlier lecture and pointed out that they should
ries, it became clear, that our strong division of fun- design the layout in a traditional way, the students
damental concepts of programming and the syntac- still used the simulation to justify it.
tical form in Java code is not the best approach for
beginning programmers. As Aguiar and Gonçalves CONCLUSION
stated in 2015 (Aguiar and Gonçalves, 2015), the two The lecture series, introduced in this paper, offers a
should be introduced to early programmers more in way to introduce architecture students to complex-
connection to each other, so that the theoretical con- ity science, programming, modelling and simulation.
cepts can immediately be tested and hopefully bet- We were able to keep the students interested in the
ter understood. This also complies with Lahtinen topic, until the end, even though it was an elective
et al. (Lahtinen, Mutka, Järvinen, 2005) finding that course.
students perceive programming structures and lan- The evaluation of the course shows that not all
guage syntax equally hard to learn. goals of the lecture series were met. Students with an
It is also shown by answers D5 and D15 (see Fig- already basic knowledge in programming were able
ure 4), that the alternation between a focus on funda- to strengthen their skills. For students with no prior
mental concepts of programming and a focus on its knowledge in programming, the course seemed to
pure syntactical rules is appreciated by a heteroge- be too difficult, even though most of them were able
neous group of students. While beginners prefer to to solve all the exercises without any help from the
hear about concepts as shown in Figure 6, more ad- assistants.
vanced students prefer to learn the syntax first, see The didactic concepts, in its general form, was
Figure 5. Yet, the division of a lesson into one hour able to lead most of the students to the goal of under-
lecture and one hour solving the exercise in an inter- standing basic concepts of complexity science, pro-
active Q+A session was highly appreciated by both of gramming, modelling and simulation; with the need
the groups as indicated by D2. for only few adaptations.
The main goal of teaching programming to ar-
chitects must be considered as failed since the an-
swers to D7 and D14 are very low among beginners,
REFERENCES
Aguiar, R and Gonçalves, A 2015 'Programming for Archi-
see Figure 6.
tecture: The Students’ Point of View', Proceedings of
A special finding we would like to highlight are the eCAADe Conference, pp. 159-168
the results of the design exercise. We clearly stated, Cross, N.C. 2001, 'Designerly ways of knowing: de-
that the students should justify their design argu- sign discipline versus design science.', Design Issues,
17(3), pp. 49-55

Helbing, D and Molnár, P 1995, 'Social force model for
pedestrian dynamics', Phys. Rev. E, 51, pp. 4282-
4286
Lahtinen, E, Ala-Mutka, K and Järvinen, HM 2005 'A
study of the difficulties of novice programmers', ACM
SIGCSE Bulletin, pp. 14-18
Mustafa, E.S., Mohamed, A.M., Saadi, I.S., Cools, M.C. and
Teller, J.T. 2015 'Modelling Uncertainties in Long-
Term Predictions of Urban Growth: A Coupled Cellu-
lar Automata and Agent-Based Approach', Proceed-
ings of CUPUM 2015, p. 18
[1] https://processing.org
[2] http://www.eclipse.org
[3] https://github.com/ybakos/proclipsing
[4] https://www.openabm.org/page/
modeling-platforms

Experience in CAAD Education Using a MOOC System
Ádám Tamás Kovács1 , Mihály Szoboszlai2
1,2
Budapest University of Technology and Economics
1,2
{kovacsadam|szoboszlai}@arch.bme.hu
This paper describes some of the challenges of using a Massive Open Online
Course (MOOC) framework system with a variety of digital content. Situated in
the 'efficiency' paradigm of digital design methods and Computer Aided
Architectural Design (CAAD) education, we allow participants to set their own
schedules, meet demands that are appropriate for their abilities, and determine
their own path. The content within this framework motivates students through
life-like tasks and examples. This paper shares our experiences in CAAD
education through a course curriculum developed by applying a variety of digital
content. We have focused on resolving the problem of inefficient teaching of
CAAD systems by developing a blended learning curriculum.
Keywords: CAAD, education, b-learning, MOOC, curriculum analytics,

mind-map
INTRODUCTION not and falls behind, while the rest become bored.
In ancient Greece, pedagogues and tutors dealt with Moreover, this absurd situation cannot be solved by
only a few students, so they could tailor the schedule, traditional means, since we cannot place a teacher
the difficulty, and the focus of the curriculum to the beside every student.
students. This fully personalized education allowed With the latest technology, we are able to resolve
the students to return to a topic which was not clear this problem by developing blended learning - that is
enough and deal with it until it was completely un- to say, b-learning curriculum which may be the next
derstood. With the spread of public education, teach- stage in engineering education after getting to Com-
ers began to handle 20-30 students, which precluded puter Aided Design from pure geometry. (Szobos-
personalized education. The only goal could be to zlai et al. 2015) With a Massive Open Online Course
reach as high a percentage of the class as possible (MOOC) frame system and with a variety of digital
through knowledge transfer. This meant that those content (such as educational videos, tests, and other
who were less talented, as well as those who were interactive content), we can allow all participants to
above average, were left on their own. Another dif- set their own schedules, to meet challenges that are
ficulty arises in the university setting - namely, that appropriate for their abilities, and to determine their
members of a class come from vastly different back- own focus. This content motivates students through
grounds, so even fewer students can be reached effi- life-like tasks and examples. If the knowledge we
ciently by teaching, since they cannot all understand teach is too abstract, or if its practical value is unclear
at once. In practice, the instructor shows something for students, then only extrinsic motivation can urge
that one part of the class can follow, another part can- them to score more points or achieve better grades.

This is less effective than stimulating their inner mo- This is an advantage, because this way the educators
tivation to work for their own benefit. do not have to know coding, and if someone drops
To introduce our new curriculum, which could out for whatever reason, one can be more easily sub-
be applied to basically any CAD or CAAD course, we stituted. From another standpoint, less energy is nec-
chose "Digital Concept Planning" in the course group essary for getting use to the system.
of "Applied Informatics for Architects", where we use In a proper educational frame system, communica-
Trimble SketchUp. The reason of our choice is that tion plays a great role. In Openlearning system, the
this is the course which students usually encounter three layers of information exchange are supported:
first, so they are fresh subjects who have not acquired student-student, student-teacher, all of the students-
any bad habits by learning such applications in the teacher. On the main page, there is a news session,
past. where we can share all the course news with all of
the students simultaneously, for instance, if there is
THE OPERATION OF THE FRAME SYSTEM a test in sight or a deadline coming. They get email
We wanted to use an existing MOOC platform, thus notification when we make an announcement. Stu-
we made a research and it turns out there are several dents can talk to each other or the teachers via instant
platforms on the market that one can use [2]. In or- message. Hence, if they have a brief question, we can
der to choose one, our main criteria were free usage, answer them immediately. They can also use the fo-
interactive and audiovisual content support, user rum, which works as a Q&A session, where students
friendly interface, and intense personalization. As a can help each other. Meanwhile, teachers can view
result, we launched our course in Open learning [1]. the conversation as well and can correct information,
There are several features which facilitate putting the if needed. This system works much better than com-
entire curriculum into an easy-to-understand struc- municating via emails, because usually students have
ture, and this provides a quite good site to organize similar problems, so we do not have to answer the
and manage the course. same mail several times.
The curriculum has its menu page, where the stu- Figure 1
dents can see all the lessons with their scheduled Heat-map of the
dates. One can only access the lesson content after curriculum
its scheduled date. It is important to have the op- webpage
portunity to share different types of content, from
the traditional types to the latest interactive ones -
e.g., videos, animated GIFs, Prezi presentations [3],
Cl3ver 3D presentations [4], etc. Several different
types of tests can be integrated into the curriculum,
which while makes the exams more interesting and
also makes correcting them faster. It is good to visu-
alize the requirements of the whole semester or the It also has a built-in analytics system, which can be
lesson. That way, students always know where they used to see how students use the contents. Thus, we
are at the moment and what they have to accom- can change it accordingly for the next semester. First
plish. The opportunity for diverse content also sup- of all, we can check every student's account and see
ports this purpose. which course content he watched or what activities
Under-the-hood management of curriculum has he has performed. In this manner, we can communi-
a very user-friendly visual content editor which can cate with him in time and possibly prevent a falling
be edited by those who have permission to do so. behind or a dropping out. Secondly, we can follow

how many people clicked on each page and what Every lesson is based upon an interesting "life-like
was the average time spent there. This way, we can situation". For example: We are attending a summer
tell which content was more useful or harder to un- workshop at an architect studio which has won a com-
derstand and can make corrections for next semester. petition concerning the bus stops in Budapest. First, we
Thirdly, we can make a heat-map (Figure 1) analy- have to model the bus stop according to the preliminary
sis for every page, which shows us where users have standard plans; then, we have to adapt it to different
clicked while on each one. Finally, we supplement site situations. (Figure 2) We practically simulate an
these by making personal surveys to measure the office environment. That way, the students are mo-
subjective part of education efficiency. From first to tivated and remain focused on the curriculum, be-
last, these are the techniques which give us a strong cause they see the practical value of the information
base for curriculum measurement and development. we teach throughout the semester.
After giving the task, we do not leave the stu-
A STRUCTURE OF A LESSON dents on their own to complete the job; instead,
First, we review the knowledge that we already know we have introduced the "strategy planning phase".
and introduce the new knowledge needed in order to When they see a problem which they think is unsolv-
solve our "life-like situation" in the lesson. This latter able, it is usually because they see it as a huge com-
material is demonstrated through smaller tasks and plex mass, and they cannot get a grip on it. They have
problems. This is the "sandbox stage", when students to break apart the problem into smaller pieces that
can experiment with the tools, try all of the functions, they can deal with one by one, thus executing the so-
and get acquainted with the limits. lution step by step.
Once we have acquired the necessary knowl- In practice, we give some time to think about
edge, the next stage is to solve the situation chal- the problem, analyze the materials, and identify the
lenges. We believe in learning by doing, because model parts and their relations. We encourage them
what students do and find out on their own lasts to visualize it on paper using the mind-map tech-
longer and is embedded deeper in their minds, so nique, so it will help them understand the whole
this second half of the lesson is an individual part. modeling process better. (Zampetakis et al. 2007)
Figure 2
Modeling strategy
mind-map

When the time is up, we show our mind-map ver- In 2016, authoritarian education seems out-dated
sion of the modeling strategy (Figure 3). We discuss it in our case. The CAD/CAAD softwares are becom-
out loud, so students can compare the two versions. ing increasingly complex. They have so many func-
They can refer back to this at any point during the les- tions and keep constantly changing with each new
son if they cannot determine the next step. Finally, version. It is hard to catch up with it, and that is
they begin working on their own, at their own speed. why sometimes, when students investigate a prob-
lem, they delve into it so deeply that their knowl-
edge outlasts the teacher's. Hence, the teacher is no Figure 3
longer necessarily the greatest expert in the class- Life-like situation
room. Nonetheless, he knows how to handle prob-
lems, where to get the required information, and can
tell if students use the tools in the most professional
and optimal way. Consequently, the role of teacher
has changed from being the one who has exact an-
swers for everything, to being rather a moderator Figure 4
of problem-solving. That is why during the individ- Smart video
ual life-like situation stage, in addition to answering directory
questions, the tutor also observes, evaluating how
professionally and how efficiently the students work
and correcting them accordingly.
Video tutorials are widely used in many fields of ed-

RESULTS AND EXPERIENCES
ucation. When one would like to explain a complex
If we want to edit a hard-hitting curriculum, we have
process with many steps, the length of the tutorial
to build it the way software development works. As
will exceed the watching tolerance of many students.
an article in WIRED magazine explains "We live in a
If they track an explanation - say, in 30 minutes - stu-
beta world." [5] which is true in our case as well. It
dents will speed up the video. They will try to jump
is an iteration process through versions which makes
to the portion of the long content which they seek.
the curriculum better and better. That is why we have
Sometimes it is hard to find the most relevant part
to get feedback constantly, which is now possible.
of an explanation. The user jumps back and forth to
Users and teachers are connected, not only in class-
stop at the required detail. Using video tutorial in our
rooms, but virtually after lessons as well.
project, we have introduced a so-called "smart video
We had 137 students complete the course with
directory" system (Figure 4). We list the entry points
the new curriculum over 2 semesters. There was a
with time code, practically labeling these indicators.
survey taken, which was filled out by 75 students,
This way, re-playing certain parts of the whole expla-
where they could evaluate and tell us their opinion.
nation of a procedure can be accomplished quickly.
We find that the content was a success, substantiated
The effectiveness of using these tutorial videos has
by the fact that students gave it 4.7 points out of 5.
increased. In practice, if someone is stuck on a given
As for the usefulness of the frame system, they gave
step, the student does not have to watch the full
it 4.4 points. Over the course of the semester, they
video. Simply by clicking on the title of the step, stu-
seemed to welcome this curriculum development.
dents can jump ahead to that point in the time code.
When we introduced the strategic planning
At the end, we summarize what was learned in
stage at the first lesson, the students were surprised.
the session and preview the topic for the next lesson
They looked confused as to why they could not open
to stimulate interest beforehand.

the software and begin the task. Instead, they had or who make slower progress followed the video tu-
to put the computer aside and plan the entire exe- torial from the start in the classroom as well. This
cution process. Later, they became used to it and is a positive development, because they are usually
understood the reason. It strengthened their visual afraid of asking questions and are capable of getting
thinking, and the learning process became more ef- stuck on a small problem. In this case, however, they
ficient. When asked if they thought this strategy was can move on more easily and experience more suc-
useful, all of the students answered in the affirmative cess.
both semesters. Students also took advantage of the various
We wanted to know how much they would use means of communication; they commented and sent
the video tutorials. When asked, 75% of students instant messages. Hopefully, though, we will im-
replied that they did use them. Nonetheless, that did prove these numbers in future semesters. As one stu-
not mean that everybody watched every video. Ac- dent expressed it nicely, "I think we are not used to
cording to our analysis, 60 people out of 137 watched such a system at the university, and that is why we did
each video for a shorter or longer period of time on not use this opportunity as much as we could." Another
average. With an audience-retention analytics tool student opinion was that the whole faculty should
(Figure 5), we can see on a diagram how many times use the same frame system, because it is not efficient
each video frame has been played, compared to the to log in and follow courses on different platforms.
number of those who started to watch the tutorial.
This diagram, as we can see, has many peaks, and it is KEY RESOURCES
very uneven. While discussing the multiple benefits of MOOC in
From these facts, we conclude that students do architectural education, Martijn Stellingwerff (2015)
not approach the situation task by watching the full notes that creating such curriculum content de-
video and following it step by step; rather, they only mands plenty of effort. First of all, all the stake-
use it if they get stuck with a step. I find this very pos- holders must understand the MOOC frame system,
itive, because they use their problem-solving skills and all faculty members need to be motivated to use
better. Instead of mechanically following the steps it.
demonstrated by the tutor, they figure them out The course content has to be re-thought as well,
on their own. What they accomplish by themselves and the variety of content needs to be edited - for
leaves a more lasting impression, so the acquired example, texts, illustrations, tests, video recordings,
knowledge will be more effective in the long term. and other interactive content. Of these, video con-
On the other hand, those who are not self-confident tent creation is very challenging. The audience has
Figure 5
Audience-retention
diagram of a
tutorial video

to concentrate hard on listening to new information CONCLUSION
while watching the video tutorial. If the audio-visual Based on our experiment, developing the curriculum
content is lacking in quality, it tires and distracts in such a way is worthwhile. The benefits are appre-
them. It is the same case in an auditorium where ciable, already in the short term. Moreover, it makes
the acoustics or the projector is not good enough, the learning process more efficient, enhancing the
and people can only listen for the first 5 -10 min- reputation of the faculty and the university as well.
utes before growing tired and giving up. This makes It improves the students' respect for the institution,
the learning process less efficient at the end of the since it employs the latest and most developed tech-
day, so we have to strive for quality in ways that nology to serve them.
require plenty of professional and expensive equip- One glowing example of this is the Delft University
ment (cameras, lights, tripods, microphones), a room of Technology, where a brand new faculty was estab-
suitable for recording, and people capable of using lished to create MOOC content as well, in a separate
the equipment and editing videos as necessary. building especially built for that purpose. How can
After the curriculum content has been com- we achieve such results within the framework of a tra-
pleted, we are still not finished, because maintaining ditional education system? We believe it is attainable
the system is just as important. Such a system must in small steps. Therefore, we have begun to establish
live, and that life must be sustained, which demands and experience this system slowly but surely, in order
time and effort. It obliges us to be in constant con- to arrive at the most efficient mode of operation.
tact with students, answering incoming questions,
following their progress, and holding their attention. REFERENCES
In order to manage this increased amount work, we Stellingwerff, M. 2015 'The MOOC-ability of Design Edu-
introduced "student tutors." The students who supply cation', Proceedings of eCAADe 2015, Vienna, pp. 57-
correct answers to other students' questions in the 60
forum become "student tutors." This is mutually ad- Szoboszlai, M. and Kovács, Á.T. 2015 'Development
vantageous for everybody, because the student who of Computer-based Visual Demonstrations in Engi-
neering Education', Studies of the University of Zilina
asks gets a correct answer, so he learns. The student
2015, pp. 75-80
tutor develops a deeper understanding of the topic Zampetakis, LA., Tsironis, L. and Moustakis, V. 2007, 'Cre-
while teaching, and the teacher can conserve his en- ativity development in engineering education: the
ergy, investing it in curriculum development, for in- case of mind mapping', Journal of Management De-
stance. velopment, 26(4), pp. 370-380
[1] http://www.openlearning.com
[2] http://www.moocs.co/Higher_Education_
MOOCs.html
[3] http://www.prezi.com
[4] https://www.cl3ver.com/
[5] http://www.wired.com/2013/04/beta/

Dynamic Light as a Transformational Tool in Computer-
aided Design
Ellen Kathrine Hansen1 , Michael Finbarr Mullins2 ,
Georgios Triantafyllidis3
1,2,3
Aalborg University
1,2,3
{ekh|mullins|gt}@create.aau.dk
New lighting technologies may fulfill a need for holistic design methods by
offering opportunities for both architects and engineers to apply methods and
knowledge from media technology that combine daylight and interactive light, in
order to complement and deepen an understanding of context. The framework
combines daylight and interactive light and includes human needs analysis,
spatial understanding, qualitative analysis, qualitative tests and visual
assessments. A transdisciplinary model termed the "Architectural Experiment" is
applied in a specific case by combining serial, parallel and iterative processes
which include contextual analysis, architectural design, simulation, C++
programming, implementation of the dynamic smart-film diffuser, programming
of voltage ranges on Arduino boards, rapid prototype construction and lighting
technology.
Keywords: Design Tools, CAAD Education, Design Concepts , Lighting Design
INTRODUCTION the built environment, which can be applied to meet

New lighting technologies may fulfill a need for holis- human biological, psychological and physiological
tic design methods by offering opportunities for needs. The implementation of these technologies
both architects and engineers to apply methods and transforms the functional split between engineers,
knowledge from media technology that combine technologists and architects into a complex transdis-
daylight and interactive light, in order to comple- ciplinary design process (Hansen 2014).
ment and deepen an understanding of context. The A criterion for success of this endeavor will be, as
framework combines daylight and interactive light Nobel Prize laureate Herbert Simon wrote of complex
and includes human needs analysis, spatial under- systems, that a large numbers of parts interacting in
standing, qualitative analysis, qualitative tests, visual a non-simple way are resolved so that the whole is
assessments, more than the sum of its parts. Simon distinguished
Architecture and lighting technology has under- between two modes of apprehending structure: be-
gone a revolution in recent years, as new sensor tech- tween the world "as sensed" (or description of state
nologies and computational skills enable 'intelligent and the criteria for identifying objects) and the world
materials', interactive control of light, energy-saving "as acted upon" (the description of process and the
measures and new LED technology to be applied in means for producing or generating objects having

the desired characteristics) (Simon 1962). The task hagen, Denmark, which draws on knowledge and
then is to employ a sequence of processes and tools skills within architecture, lighting engineering and
that support each other and lead to innovative de- media technology, in teams of researchers, the light-
sign solutions based on a theoretical model of 'state'. ing industry and graduate students of lighting design
Simon's theories may have relevance to complex sys- (see [1] ).
tems encountered in architecture, however, where The model is applied in a pedagogical curricu-
he expected complex systems to be hierarchies in a lum, in the context of multi-level learning competen-
world in which complexity evolves from the interplay cies.
of single elements, the goal here is to achieve a re-
solved design of integrated simplicity, which evolves METHOD
from an initial state of complex diversity. The initial statement of the "Architectural Experi-
Towards the end of testing this relevance, there is ment" is a theoretical proposition of how to design
a need to develop mixed research and design meth- with light as a multi-dimensional design element,
ods, where computer aided design tools used in by integrating scientific, technical and creative ap-
physical and virtual experiments assist the transdis- proaches to light in 5 domains. It proposes that dis-
ciplinary design process. To study this in practice, it tinct qualitative and quantitative criteria and meth-
can be instructive to focus more narrowly on a single ods from different disciplinary traditions can be in-
aspect of architectural intervention. This paper takes tegrated successfully, despite disparate technical/-
lighting as the specific focal aspect. scientific, social scientific and art/humanities back-
Toward describing and developing such meth- grounds (see Figure 1).
ods, this paper focuses on the potential of lighting to
act as a primary means of renovating and transform- Figure 1
ing architectural environments. In this regard, light Model for the
as a transformational design tool is perhaps one of Architectural
the most non-destructive, reversible and gentle ways Experiment.
to transform and redefine architecture and spaces
(Magnusson et al. 2015). The case studied is a partic-
ular urban and architectural challenge represented
by an urban church building, which had fallen into
disuse. The building was acquired by a private buyer,
who sought innovative ideas for its transformation.
The process to integrate digital tools and com-
bine architecture, lighting engineering and media
technology builds on a trans-disciplinary model pre-
viously developed in the context of 'smart homes'
(Hansen 2013) and presented as the "Architectural The case study method is useful to investigate con-
Experiment" (Hansen 2014). This paper will docu- textual conditions pertaining to the phenomenon of
ment and analyse the application of this process from study, in this instance the processes employed in ap-
a graduate student group project, dealing with inte- plication of the transdisciplinary model. Based on
grating computer-aided design and new technolo- this theoretical proposition, briefly described in the
gies to transform an architectural context through preceding sections, the method relies on multiple
light. The work was carried out at a lighting de- sources of evidence, drawn from a single case (Yin
sign programme at Aalborg University in Copen- 1994).

The findings from the case study can then be com- multaneous variety of transparent to translucent to
pared to the initial proposition in order to refine the opaque.
theory, for use in future applications.
Figure 2
Three phases and
illustrative light
CASE: THE ABSALON CHURCH: COMPUTER-
settings were AIDED DESIGN OF DYNAMIC LIGHTING
defined (Source: Objectives and research questions
Magnusson et al. The intention of the project is to transform a city
2015). church building, situated in an older, dense, hous-
ing area close to the city centre of Copenhagen, from
its traditional religious functions to a more dynamic,
contemporary and multifunctional space, using light
as the main design element. The initial question
asked in this context is therefore 'how light can trans-
form the interior space of the de-sanctified church to
meet needs of the local community for new activities'.
Drawing on knowledge and skills within architecture,
lighting engineering and media technology, gradu-
ate students of lighting design teamed up to provide
a design solution for a non-religious space, where fo-
cus is on an essentially neutral ceremonial volume.
The needs for a non-religious ceremony were sought
initially by proposing a dynamic, programmable vi-
sual spatial layering, adjustable from the start to the
end of the ceremony.
By defining and combining the concepts of
emotion/mood/atmosphere and light effects, three
phases and illustrative light elements were defined:
1. light from pendants hanging from the ceiling, 2.
highlights in the centre of the space and 3. poly-
mer dispersed liquid crystal (PDLC) (Baetens 2010)
sheathing placed in front of the windows and walls,
thereby allowing dynamic diffusing of daylight and
artificial light. Using these lighting elements, three The process of designing with dynamic
phases of the ceremony, arriving, transition and de- lighting
parting, are illustrated (see Figure 2). The design process will be described using in terms
The diffuser was constructed of a middle layer used by the "Architectural Experiment" model.
of PDLC, surrounded on each side by a transparent Step 1: Transfer. Knowledge from different disci-
conductive layer for the purpose of applying voltage, plines is collected, transferred, reviewed and coded
with an outer-most layer of non-conductive isolating into a common language to define a common prob-
material (see Figure 3). To slow the perceived trans- lem statement for the design.
formation of the material, the diffuser was built up of The very first step was to define a common imag-
independent triangular components, enabling a si- inative research question starting by asking "Why is

Figure 3
The diffuser was
constructed of a
middle layer of
PDLC, surrounded
on each side by a
transparent
conductive layer for
the purpose of
applying voltage,
with an outer-most
layer of
non-conductive
isolating material
(Source:
Magnusson et al.
2015).
a church for sale?" Collecting knowledge on peo- the large windows give the room unique features and
ples need for at ceremonial space and not a religious allow much daylight to enter. The space presented
space led to the imaginative research question: What an unique and solemn atmosphere.
if a now defunct church could be transformed into a
neutral ceremonial space, satisfying new ceremonial Figure 4
needs in our society? Design elements for
Knowledge on transformation and societal satisfying
needs, religion, neutrality, similar projects, and stake- ceremonial needs
holders was collected to define the overall problem of society, through
statement and the three success criteria. A reflec- the use of light
tion on transformation, heritage and light was made (Source:
through literature survey. This knowledge was used Magnusson et al.
to define the differences between the light in a typi- 2015).
cal church and a religion-neutral ceremonial space.
This led to the problem statement (see Figure
4): Can we transform a religious space into a religion
neutral ceremonial space, satisfying new ceremonial
needs of our society, through the use of light?
1. Using light to achieve a new religiously unbi-
ased identity.
2. Ensuring that the ceremonial contents and
functions are supported and facilitated by the
lighting design.
3. Ensuring that neutrality does not compromise
the solemnity required of a ceremonial space.
Qualitative analysis was made based on architectural
observations of the design space. It was noted that

Quantitative analysis of the daylight in the building conducted.
was done through illuminance measurements (see Images of light attributes were replicated in a
Figure 5). digital environment, defining a certain ceremony and
placing 3D models of people in the space of the
Figure 5 church building. An online survey was conducted to
Knowledge from compare pre-defined variables.
different disciplines From the results the following design guidelines
is collected, were defined:
transferred,
reviewed and • A compromise between appropriateness and
coded into a neutrality is a compromise between warm
common language. and white light: ideally between 4000-6000
Kelvin.
• Some level of contrast is needed to achieve
solemnity and intimacy.
• A very neutral and preferred result seems pos-
sible with diffusion of daylight.
• Many small luminaries are preferred, but must
avoid heavenly references to be neutral.
Step 2: "Translate". The knowledge from the differ- The visual hierarchy of the design is achieved using
ent fields is translated and investigated through ex- three different lighting elements that together con-
periments. stitute the design proposal:
A literature survey of the meaning of ceremonial
rituals referred to theory from sociologists, philoso- • Pendant lights, floating over the benches in
phers and psychologists and formed a fundamental varying heights.
design concept: Ceremonies are rituals that mark a • A highlight of the floor in the ceremonial cen-
change of state of being. This idea was combined ter.
with the question of how to satisfy the practical func- • Dynamic diffuser panels that control and alter
tional needs of the ceremony, as well as to add value the intake of daylight
to the ceremony. This led to the following assump-
tion that became the core of the design concept, to The results of the attributes-test suggested that diffu-
be explored through a design proposal: sion of daylight is a neutrally perceived element. The
diffuser offered opportunities for the sought after dy-
1. "If the ceremony is dynamic then the lighting namic element of lighting, by controlling how warm
could also be dynamic." light of the pendants is perceived and amount of day-
light present in the space.
For example, if a funeral is charged with emotions, A second experiment was conducted to test how
which evolve through the ceremony in phases, light- people perceive light with different fading phases of
ing could support these emotions by supporting the the dynamic diffuser, viz.: 60, 30, 10, 5, 2 and 1 sec-
transition between phases and fulfilling the spatial onds.
and atmospheric needs attached to each phase. A predominant consensus was achieved from
For a better understanding of how different this experiment, and a 30 second fading time for each
lighting attributes can support a transformation of triangle was chosen as a basis for the dynamic dif-
religious space into a neutral one, an experiment was fuser.

Rapid prototype simulations of daylight and the ferent scenarios and moods of the changing scenar-
smart-film diffuser were essential to study the design ios and their reaction to different solutions was ob-
potentials of the double interior skin, appropriate to served.
the three phases in the ceremonial space.
Figure 6
Step 3: Transform. The findings from the above 3D model
experiments were transformed into a design solu- simulations of the
tion using 3D models, scale models and simulations. diffuser-panels
The effects of the dynamic diffuser together with the were used for
other lighting elements were rendered in 3D mod- qualitative testing
els, renderings and photometric simulations (3DS- in focus groups
Max, Rhino, V-ray, IES files) to examine the quantita- (Source:
tive relationship between daylight intensity, translu- Magnusson et al.
cency of the interactive interior wall, as well as time- 2015).
lapses in relation to the ceremony.
In accordance with the design concept, daylight
entering the ceremony space was to be filtered in dy-
namically to support emotions present in the course
of the funeral ceremony. The solution to these speci-
fications was a setup comprised of a PDLC smart-film
material, an Arduino controller using open frame-
works and customised electronic architecture. To
perform experiments with each of these technolo- Step 5: Share and learn. Video, renderings, ani-
gies and to test the final functionality and compati- mated materials were prepared to communicate the
bility, a down-scaled prototype was developed. design idea under different lighting conditions (see
The 12 individually controlled PWM outputs sit- Figures 7-9). The time aspect was essential in this
uated on the Arduino Mega offered individual out- context and the video was therefore an important
put voltage levels in the range between 0 and 5 volt tool.
direct current (VDC). To meet the higher voltage re-
quirements of the PDLC material, additional electron- DISCUSSION
ics were needed between the Arduino outputs and The transdisciplinary model for the "Architectural
PDLC composite. Observations of the quality of the Experiment" is applied to the case study, which
light transmitted through the diffuse and transparent combines serial, parallel and iterative processes and
panels were made and measurements of the values which included contextual analysis, architectural de-
of the transmittance and reflectance of light in the sign, simulation, C++ programming, implementation
different stages of the material were measured. of the dynamic smart-film diffuser, programming of
In order to test different lamp sources with differ- voltage ranges on Arduino boards, rapid prototype
ent luminous power and lighting distribution a sys- construction and lighting technology.
tem was needed, which would provide a flexible ap- The case demonstrates the potentials of com-
proach for access to the space and the different light bining digital, engineering and architectural design
sources. Dialux software was used for this process. tools and methods to transform the meaning of
Step 4: Test. 3D physical models and digital simula- spaces, by coding intelligent materials and dynamic
tions were used for qualitative testing in this case (see light to meet the changing needs of the urban archi-
Figure 6). A user focus-group was questioned on dif- tectural environment.

The case focuses on the creation of a neutral cere- 2. Translate. The knowledge from the different
monial volume, using a complex system of resources. fields is translated and investigated through
It attempts to create, through lighting, an influence experiments. Rapid prototype simulations of
upon the sensed or perceived world. The case study daylight and the smart-film diffuser were es-
is used to examine the process it employs to achieve sential to study the design potentials of the
this aim. double interior skin, appropriate to the three
phases in the ceremonial space.
Figure 7
3. Transform. The findings from the experiments
Arriving. Rendering
are transformed into a design solution using
of the solution:
3D models and simulations. The effects of
video capture
the dynamic diffuser together with the other
(Source:
lighting elements were rendered in 3D mod-
Magnusson et al.
els, renderings and photometric simulations
2015).
(3DSMax, Rhino, V-ray, IES files) to examine
the quantitate relationship between daylight
intensity, translucency of the interactive inte-
rior wall, as well as time-lapses in relation to
Figure 8 the ceremony.
Ceremony. 4. Test: The 3D model and simulations were
Rendering of the used for qualitative testing in this case. A
solution: video user group was questioned on different sce-
capture (Source: narios and moods of the changing scenarios
Magnusson et al. and their reaction to different solutions is ob-
2015). served.
5. Share and learn: Video, renderings, animated
materials, communicating the design idea
with different light settings. The time aspect
Figure 9 was essential in this context, and the video
Departing. therefore an important tool.
Rendering of the
solution: video While luminance levels can be predicted and mea-
capture (Source: sured relatively easily, several of the success crite-
Magnusson et al. ria for the project revolved around achieving a dy-
2015). namic architectural atmosphere of neutrality suitable
to a ceremonial space where perception and time
is subject to the inherent ambiguity of visual per-
ception, subjective feeling, emotional response and
Following the "Architectural Experiment" model, the personal preferences. An attempt was made to ap-
process and tools can be grouped in the following proach this question both analytically and with a
five domains: qualitative approach and established results that in-
dicate that lighting attributes can have an important
1. Transfer. Knowledge from different disci- role in affecting whether a space is perceived as reli-
plines is collected, transferred, reviewed and gious or not.
coded into a common language. The process of combining complex combina-

tions of daylight and electrical light, by employ-
ing the initial model has confirmed for the authors
its usefulness in integrating many disciplines and
computer-aided technologies in a design studio set-
ting. The method of moving from an analytical
phase centered on research and testing, into cre-
ative design-proposal phase, followed by a series of
smaller experiments to develop implementation and
detailing was successful in incorporating the three
academic fields of lighting engineering, architecture
and media technology and in achieving a resolved
design of integrated simplicity, evolved from an ini-
tial state of complex diversity.
AKNOWLEDGEMENTS
The student team comprised Tim Jesper Ivan Mag-
nusson, Andrei-Ducu Predescu, Joaquim André Mor-
eira Maia and Christian Henrik Ankerstjerne. Super-
visors for the project were assistant professor Paolo
Burelli and professor Marc Fontoynont.
REFERENCES
Baetens, R, Jelle, BP and Gustavsen, A 2010, 'Properties,
requirements and possibilities of smart windows for
dynamic daylight and solar energy control in build-
ings: A state-of-the-art review', Solar Energy Materi-
als and Solar Cells, 94(2), pp. 87-105
Hansen, EK and Mullins, MF 2014 'Lighting Design: To-
wards a synthesis of science, media technology and
architecture', Fusion: Proceedings of the 32nd eCAADe
Conference, Newcastle UK, pp. 613-620
Hansen, EK, Olesen, GGH and Mullins, MF 2013, 'Home,
Smart Home: A Danish Energy-Positive Home De-
signed with Daylight', Proceedings of the IEEE : Special
Issue: The Smart Home, 11, pp. 2436-2449
Magnusson, T, Predescu, A, Maia, J and Ankerstjerne, C
2015, 'Seeing the Light – Transforming with light.
Redefinition of a ceremonial space', in surname
missing, initials missing (eds) 2015, Report:, Aalborg
University Copenhagen
Simon, H 1962, 'The Architecture of Complexity', Proceed-
ings of the American Philosophical Society, 106(6), pp.
467-482
Yin, R 1994, Case Study Research, Sage, London
[1] http://www.light.aau.dk

IT Driven Architectural Design for All?
Sebastian Białkowski1 , Anetta Kępczyńska-Walczak2
1,2
Lodz University of Technology
1,2
{sebastian.bialkowski|anetta.kepczynska-walczak}@p.lodz.pl
This paper discusses teaching parametric design as a supportive method of

introducing design logic. Two case studies have been described, analysed and
concluded. The first case study focuses on a workshop based design of a
parametric pavilion, which resulted of building 1:1 scale object. The second case
study concentrates on the academic compulsory course providing parametric
design knowledge based on a particular topic imposed by tutors. In both cases
the main purpose was to get students being accustomed to a different way of
thinking, to open their minds to new approaches to design process and to
demonstrate a connection between programming skills and imagination
capabilities. Each of the cases returned valuable guidelines for design studio
pedagogy which has also been revealed in this paper.
Keywords: design logic, parametric, design pedagogy
INTRODUCTION ing scripting or programming, to fill in the gaps

There is no doubt, computers have become main of the software feature set"(Boeykens and Neucker-
aiding tools for designers nowadays. Fast speeding mans 2009). That possibility of software customiza-
of computational technologies brings new toolsets tion by applying programming abilities is becoming
for architectural practice.Actually, the IT driven de- increasingly appreciated. In consequence such situ-
sign process is not a novelty and has been graduation, on the one hand enforces the universities to
ally adopted in architectural design domain. How- make shift in education, but on the other hand opens
ever, there is a great shift observed lately in the di- domains of experimentation and research. There is
rect influence of programming technologies on de- a great discussion on the scope and methodologies
sign practice. To fully exploit these abilities archi- of teaching these emergent technologies since they
tects have to obtain new skills and knowledge. This change the paradigm of design process. The ques-
can be done by learning programming or scripting tion how to support creative thinking with digital
languages which are opening new domains in ar- tools still needs further investigation (Kepczynska-
chitectural creativity and "manufacturing the ideas". Walczak 2014). "However, visual programming ap-
It has been widely discussed if there is no direct proaches that promise to fit the graphical mindset
link between a designer's mind and designing tool, of aspiring architects better are also gaining ground,
a designer becomes rather a reviewer than a cre- e.g. Bentley's Generative Component and Grasshop-
ator (Kepczynska-Walczak 2008). As claimed by Ste- per for McNeel Rhino 3D. The reason for this is re-
fan Boeykens and Herman Neuckermans " (...) soft- ported to be that algorithmical thinking requires a
ware can only provide the desired functionality us- problem-centered approach, while architecture usu-

ally takes a solution-centered strategy" (Wurzer et al. for students and their imagination. Besides the de-
2011). As the title of the paper suggests there is a cision about material the three axis CNC milling ma-
question to be answered if the information technolo- chine was chosen for the main manufacturing pro-
gies driven architectural design should be a compul- cess. Both a material and a fabrication method were
sory part of the curriculum. the key constrains imposed on students.
CASE STUDY Figure 1

In this chapter two chosen case studies are described, P3 Workshops -
analysed and concluded. Some guidelines for de- Introduction to
sign studio pedagogy is also be revealed. In both parametric design.
cases the main purpose was to get students being
accustomed to a different way of thinking, to open
their minds to new approaches to design process and
to demonstrate a connection between programming
skills and imagination capabilities.
WORKSHOPS - PAVILON P3
The first case study focuses on a workshop based de-
sign of a parametric pavilion. A group of students Figure 2
from Architecture field of study, driven by a need to Teamwork in
learn and explore new digital tools more and more groups.
commonly used in a worldwide architectural prac-
tice, organised a workshop and named it P3 - Pro-
cess Product Purpose. Its main goal was to learn para-
metric design methods and use that new knowledge
to create a full-scale pavilion not only virtually but
also in reality. The workshop was planned as a two
weeks long event, including building up the physical
1:1 scale model. However, due to the need to pro-
vide more introductory tutorials, and additionally, to
a technical delay, it was extended and lasted three The workshop began with an introduction to para-
weeks. During that time students experienced, in metric tools in the form of an intensive course of
a very intensive and compressed way, a whole de- Rhinoceros and Grasshopper (see Figure 1). The se-
sign process - from a design concept to realisation ries of tutorials were planned to cover a wide range of
on a building site. From the very beginning, a 15 differentiated knowledge of parametric design field
millimetre plywood was chosen as a basic material including basics of attractors, dynamic relaxation or
to construct the physical pavilion. The relatively low data managing. During the workshop students' cu-
price and easiness of tooling were the primary factors riosity led to more involving problems such as big
in decision making process. Another argument was data structures, complex geometry dependencies of
that the chosen material was highly accessible on manufacturing or tagging tasks. It is also worth to
the market as a standard construction material. The mention here that for majority of participants, this
budget provided for the workshop allowed to buy 78 workshop was the first opportunity of gaining knowl-
sheets 2,20 meters long and 1,20 meters wide. It gave edge about architectural design process driven by al-
over 200 square meters of plywood in total available gorithms.

Figure 3 design logic was a correlation between three points,
Prototype testing. freely chosen on a flat plane which represented the
ground.
Figure 4
Base form creation
process.
Participants were divided into five design teams (Fig-

ure 2). Each group was asked to design a vision of the
pavilion, concerning a type of material, its limitations,
Figure 5 amount, and available manufacturing process. Apart
Simulation from a pavilion's architectural form, the high empha-
principles. sis was placed on the construction and structural is-
sues. During design process, students were asked to
omit any traditional joints applied in physical models,
such as glue, metal plates or nails. The whole struc-
ture was intended to be made of plywood only.
Furthermore, they were encouraged to make
mock-ups to evaluate their design and ideas. After a
few days of work the midterm presentation was or-
ganised, where each group presented their design
logic and structural solutions. A few joints prototypes
were prepared in CNC milling machine to test their
durability and stiffness. The best idea of each group
was extracted to become the basis for one common
final project.
The main structure of the pavilion was based
on a dynamic relaxation algorithm driven by Kan-
garoo solver for Grasshopper. The Frei Otto stud-
ies on tensile (soap bubble) structures and Antonio
Gaudi study of inverted gravity logic led students to
Figure 6
think of self-supporting structures as a base idea of
Form finding -
their projects. A mesh as a starting point for simula-
tweaking
tion was the result of combination of two algorithms:
parameters.
Voronoi tessellation - to find a general outer shape of
a pavilion, and the magnetic field behaviour for a di-
vision of the resultant space into triangular mesh (see
Figure 4). For both algorithms a starting point for the

sign of the pavilion was ready, additional scripts were Figure 7
prepared for further fabrication purpose. During that Marking out the
stage, participants developed identifying system for pavilion on the plot.
all elements for future pavilion assembly.
Figure 8
Assembly process.
A simulation of inverted gravity was run on basic

principles of string behaviour. On each node of the
mesh an uniformly acting force against gravity was
applied. To keep the structure on the ground, de-
signers chose the most characteristic points to be
fixed. To enrich the form finding process additional
constrains were applied. Creation of overhang en-
tries to the pavilion, on the each site of the structure,
required additional structure movement limitations
(see Figure 5). Particular nodes in the pavilion were Figure 9
hooked to the ground by strings behaviour to imitate Digital model of
tents strand. The improvement of structural stiffness pavilion.
was achieved by applying supplementary forces act-
ing between plates. That method generate origami
effect on the surface of pavilion enhancing it tecton-
ics. Thanks to parametric tools, dynamic relaxation
could be running many times with different param-
eters or additional constrains, which leads to divers
outcomes (see Figure 6).
Figure 10
In parallel when part of a team was working on
Final result - overall
general design of the pavilion, second part of partic-
view.
ipants worked on 1:1 scale prototype of few adjoins
panels, developing the best solution for their connec-
tions and joints. Tests on the prototype allowed stu-
dents to evaluate their ideas and prevented design
from a variety of construction errors. When a final de-

Figure 11 was supported by digital instructions prepared by
Final result - birds students as a part of the script. Three weeks of in-
eye view. tensive work resulted with a structure containing 570
unique panels joined by 1008 connectors mounted
by 2016 wedges. This fully digitally designed struc-
ture (see Figure 9) became a kind of architectural
manifesto for University authorities and other faculty
students by showing capabilities of IT tools in con-
temporary architectural practice (see Figure 10 to 13).
COMPULSORY COURSE
A successful result, strong interest and high stu-
dents attendance at the parametric design workshop
confirmed the growing need to introduce a regular
course on advanced CAAD tools such as paramet-
ric design methods. The first trials to apply these
methods were made for the fourth year students of
the Bachelor degree. The participants of the course
Figure 12 were both, regular enrolled and Erasmus students.
Final result - detail. The whole course lasted fifteen weeks and was di-
vided into three parts. During the first three weeks
students were acquainted with new technologies
and approaches in architecture and prepared pre-
sentations on chosen topics. A wide-ranging vari-
ety of themes was covered, such as an application
of 3d printing technology in architecture, BIM pro-
cess, robotic's arm aiding fabrication etc. That part
Figure 13 aimed to introduce participants to a new environ-
Final result - ment of computational design, and make aware of
interior. recent trends in technologies. Then, subsequent six
weeks (the second part of the course) were dedicated
to a new software and tools exploration. Students
learned about NURBS modelling toolset provided by
Rhinoceros3D with additional plugin for Graphic Lin-
ear Programing named Grasshopper3D. During this
part many examples and practical exercises related
to architecture were conducted to make participants
With the help of the faculty surveyors and data gen- familiar with the computational process of design-
erated by Grasshopper, the pavilion was precisely ing. It is worth to mention, that none of students
marked out on the plot (see Figure 7). As soon as had any knowledge or skills in parametric, genera-
all panels, joints and wedges were cut out, students tive or computational design before. Last part of the
began pavilion assembly process (see Figure 8). The course contained a project task. The main assump-
construction took about five days to complete and tions were to design courtyard roofing in our main

faculty building. As shown in Figure 14, three options hind it. In that scenario, participants were developing
of roofing were providing for participants: to design their skills faster than through understanding gen-
whole building roofing, cover one of the courtyards, eral problems contained in standard tutorials. In our
or design a cover for both patios. Students were free course, tutorials (second part) appeared only to help
in form of the design concerning exploration of new students rapidly understood a basic workflow of par-
parametric tools. During these weeks participants ticular tools usage and were limited to minimum. The
were concentrated on their individual response to a principal emphasis was put on a final project. A de-
given topic. The results were summarized during the sign task for the final project was intended not be
final project presentation. complex. Students were encouraged to focus mainly
on a building form, its appearance and design logic
rather than its functionality. The reason for that was Figure 14
a desire to divert thoughts of students from straight- Initial boundaries
forward design studio problems solving towards the for final project.
process itself in order to open a potential for enrich-
ing a design development. The whole process of
achieving a final form of a building or a structure was
as important as the final appearance.
RESULTS
The case studies represent different approaches in a
multifaceted meaning. Nevertheless, some aspects
are clearly observable. It is worth stressing here stu-
To sum up, the aim of the course was to make students' attitudes were highly diversified in both cases.
dents accustomed to a different way of thinking, to The participation in the workshop was free for all
open their minds to new approaches to design pro- faculty students who were interested in the partic-
cess and to demonstrate that by connecting pro- ular topic. That freedom attracted highly motivated
gramming skills with imagination capabilities. The and interested participants. In opposite to the work-
first task - a presentation of up-to-date technolo- shop, the compulsory course as a different pedagog-
gies was intended to be a preface to individual ap- ical method, gathered all students. The first case def-
proaches stated in a final design problem. Students initely demonstrated a high motivation of learning
presented their research to exchange knowledge and new tools for architectural design. Despite the atmo-
inspire each other with new solutions and opportuni- sphere of work with a slightly competing factor, par-
ties given by a modern technology to reveal how con- ticipants inspired and encouraged each other. What
temporary architectural design might be successfully is more, the aim of creating and building physical
aid by digital tools. 1:1 scale object occurred a significant incentive. The
From the authors' experience, the learning pro- second case revealed different conclusions, though,
cess of parametric design methods is much more re- it also gave a valuable feedback for teaching para-
quiring than traditional methods used before. The metric design methods. Not all participants per-
best results were possible to be achieved during formed an ability and developed logical or mathe-
solving individual problems appearing in a project. matical way of thinking to fully appreciate paramet-
Focusing on individual projects, students improved ric design. What is more, during working on their fi-
their consciousness of parametric design and devel- nal project, a majority of students was focused only
oped an ability to understand the whole logic be- on final result of their work (see Figure 15). The de-

sign process was consciously ignored, threaten by the feeble preparation of bachelor students to com-
the students as a transitional stage without signifi- plex and demanding designing process such a para-
cant meaning. metric design. In a contrary to the academic course,
the workshop revealed a radically different outcome.
Figure 15 All students involved in pavilion development were
Outcom form very proud of the outcome but at the same time they
compulsory course were highly conscious about design process which
- Student work. led them to it.
Figure 16
Outcom form
compulsory course
- Student work.
The maturity of students can also influence their de-

sign process and the way of thinking. The difference
in projects' quality was clearly visible during a final
presentation of the compulsory course Projects deliv-
ered by the Erasmus students were much more elab-
orate (see Figure 16). Their works diverged from the During the exhibition made after the workshop, the
other, by higher sensitivity of design logic and pro- process was the main product exhibited by them. Ad-
cess responsible for a final project. What is interest- ditionally, neither the academic course nor the work-
ing, those students were older from the regularly en- shop assumed the needs of existence of any theoret-
rolled ones, and additionally they were participants ical part of parametric design. That put our method
of Master Program in their Alma Mater. This example in contradiction to methods presented by Aguiar and
shown the importance of design maturity of students Gonçalves (2015) in their paper. All theoretical prob-
when Parametric Design course is implemented in a lems and questions were answered during computer
studying program. Two editions of the course show labs. The usage of graphical, node base programing

application such as Grasshopper3D gave beginner
students better visibility what was happening during
a script execution. What is more, clear and pleasant
graphical user interface did not deter potential users
as opposed to the classical written syntax.
CONCLUSIONS
Both case studies enriched authors didactic experi-
ence and allowed them to evaluate their teaching
methods. As the results of the workshop and com-
pulsory course show, students' predispositions to un-
derstand and use parametric designing methods ap-
peared highly diversified.Architecture as a field of
study, due to its combination of art and engineer-
ing, attracts people with varied skills and aptitudes.
Parametric tools require from the user to be more
conscious about designing as a process. From teach-
ing point of view, parametric design should be intro-
duced at Master degree program where students' at-
titude to learn complex tools and her/his design ma-
turity is highly developed. Moreover, it is suggested
parametric design classes should be only proposed
as elective courses.
REFERENCES
Aguiar, R and Goncalves, A 2015 'Programming for Ar-
chitecture: The Students’ Point of View', Real Time -
Proceedings of the 33rd eCAADe Conference - Volume
2, Vienna, pp. 159-168
Boeykens, S and Neuckermans, H 2009 'Visual Program-
ming In Architecture Should Architects be trained
as programmers?', Joining Languages Cultures and Vi-
sions: CAADFutures 2009 CD-Rom, PUM
Kepczynska-Walczak, A 2008 'Contemporary Renais-
sance Architect – Yet Architect?', eCAADe 2008 Artesis
University College, Antwerp, pp. 445-450
Kepczynska-Walczak, A 2014, 'The Act Of Design
– Beyond The Digital?', Architecturae et Artibus,
6(1/2014), pp. 24-28
Wurzer, G and Alacam, S. L. 2011 'How to Teach Archi-
tects (Computer) Programming: A Case Study', Re-
specting Fragile Places 29th eCAADe Conference Pro-
ceedings, 2011, pp. 51-56

Implementing Biomorphic Design
Design Methods in Undergraduate Architectural Education
Asterios Agkathidis1
1
University of Liverpool
1
a3lab@liv.ac.uk
In continuation to Generative Design Methods, this paper investigates the

implementation of Biomorphic Design, supported by computational techniques in
undergraduate, architectural studio education. After reviewing the main
definitions of biomorphism, organicism and biomimicry synoptically, we will
assess the application of a modified biomorphic method on a final year,
undergraduate design studio, in order to evaluate its potential and its suitability
within the framework of a research led design studio, leading to an RIBA
accredited Part I degree. Our research findings based on analysis of design
outputs, student performance as well as moderators and external examiners
reports initiate a constructive debate about accomplishments and failures of a
design methodology which still remains alien to many undergraduate curricula.
Keywords: CAAD Education, Strategies, Shape Form and Geometry, Generative

Design, Design Concepts
INTRODUCTION ponents have included Imre Makovecz (Kuhlmann

The term Biomorphism ultimately derives from 1998) and Frank Lloyd Wright, who, as Aldersey-
Goethe and was firstly introduced by the British poet Williams (2003) points out, took the approach to new
and writer Geoffrey Grigson in 1935, in relation to the heights. Even though organicism and biomorphism
work of the sculptor Henry Moore (Kosinski 2001). It are related (and are often used - incorrectly - synony-
is composed out of the Greek words βίος (life, living) mously), there are significant differences between
and μορφή (form). Strongly related to Surrealism, it them. Organic architecture, as Wright himself de-
was used to describe the creative synthesis between fined it, does not necessarily resemble natural forms,
cubism and abstract art. It has been often associated but rather relates to materiality and integration into
with fluid, organic shapes in art, architecture and de- a natural context.
sign. Particularly in architecture, the term has been Finally, the term 'bionics' - combining 'biology'
related to nature inspired forms, naturally occurring and 'technics', or 'electronics' - was invented by US Air
patterns and shapes. Force colonel Jack E. Steele and presented at a 1960
The notion of organicism promotes harmony be- 'Bionics Symposium', entitled 'Living Prototypes - The
tween architecture and nature to the point where Key to New Technology' (Nachtigall 2005). Aiming to
form and natural context merge into one. Key pro- inform engineering and technology with knowledge

and aspects of performance as they have evolved in ous other attempts to incorporate nature into design
nature, bionic design can be seen as an aspect of and architecture have been taking place from ancient
biomorphism. However, bionic, or biomimetic, archi- times until our days, in almost all architectural move-
tecture implies not only the form-related aspects of ments. Starting with traditional architecture around
mimicry, but the inherent qualities of construction as the world, continuing with Arabic ornaments or ar-
well (Gruber 2011). Gruber also points out that Frei chitects and designers during Art Deco and art Nou-
Otto's research studies on membrane surfaces, and veau, Antonio Gaudi in the beginning of the 20th
Buckminster Fuller's tessellation techniques for dome century, until Oscar Niemeyer, Frei Otto in the 50's
geometry, were most probably the first attempts to and 60's continuing with Sandiago Calatrava and Sir
integrate bionics into architectural design. It is in the Norman Foster in our days.
last ten years - by incorporating emerging technolo- Through the rise of digital design and fabrication
gies and tools, such as parametric, algorithmic and techniques, bio inspired design is becoming more
generative design methods - that architects and en- and more important in today's architectural practice
gineers claim to have moved beyond a mimicry of ge- and academic framework, calling for innovative ed-
ometry and order, and into enhancing environmen- ucational concepts in both undergraduate and post-
tal, structural and material performance by learning graduate curricula.
from the mechanisms and properties found in nature.
New terms such as zoomorphism, geomorphism and AIMS AND OBJECTIVES:
anthropomorphism have arisen in order to specify Main aim of this research paper is to evaluate the
the particular source of inspiration or mimicry in each suitability and implementation of biomorphic design
case (in these instances, animals, geology and hu- methodologies into undergraduate design studio.
mans, respectively). Furthermore, we will investigate a variety of contem-
Even though these terms appeared only in the porary biomorphic design methods by monitoring
20th century, nature has always been a paradigm for design case studies, using different approaches and
artists and architects. Looking back in the origins techniques, practiced in the academic framework of
of western architecture, such as the design of an- Studio 04, one of the five graduate year's design stu-
cient Greek temples, nature always plays an impor- dio units in our University. The studio encourages the
tant role, not only in terms of shape and appearance use of digital design and fabrication tools, as well as
but also in defining proportions and structure. Vari- physical modelling and drafting techniques.
Figure 1
Diagram of design
method applied in
the design
experimentations
presented.

Our main research questions can be described as fol- the final paper, are offering an analytic evaluation
lows: overview of biomorphic design methodologies inte-
gration in undergraduate education. In addition, we
• How can biomorphic design methodology will present a categorisation of the various design ex-
be integrated in an undergrad design studio perimentations using nature as a source of formal in-
module? spiration, spatial / typological innovation, organiza-
• Which are its strengths and weaknesses? tional method or as a tool for performative integra-
• Does the applied design methodology tion into structure and geometry.
achieve its objectives in producing innova- We will present four characteristic design exper-
tive design solutions and increase students' imentations, which were produced by applying the
design skills and future employability? method described previously and proceed to discus-
• Does the integration of biomorphic design sions and conclusions based on the analysis of the
methodology in undergraduate level seem collected data.
appropriate?
DESIGN EXPERIMENTATION 01: HILL AND

RESEARCH METHOD AND FINDINGS
VALLEY
Our design method was based on three main
Our first design experimentation initially analysed hill
phases, Analysis - Morphogenesis - Metamorphosis
and valley formations resulting from the movement
(Agkathidis 2015), aiming the production of com-
of water (Figure 2). They emerge through the tidal
plete architectural proposals, in a research led edu-
currents on the soft, formable sandy soil of the sea.
cational framework. Analysis was focusing on prece-
Figure 2 dent studies originated in nature, where design prop-
Sand patterns erties, design intentions and parameters were iden-
formed by water tified. During Morphogenesis abstract generative
streams, models (Figure 1) were developed by incorporating
Merseyside, UK. the parameters of the previous phase. Finally, dur-
ing Metamorphosis, the previous prototypes were
translated into building proposals, including all nec-
essary deliverables, such as floor plans, sections, de-
tails, physical models and visualisations.
The method described here, was applied on a
12 week, design studio unit with a cohort of 63 stu-
Figure 3
dents, tutored by six tutors, part of a total of 240
Surface iterations
students in the entire year cohort. Students had to
by Grasshopper
work collaboratively during the first four weeks (anal-
algorithm, using
ysis and morphogenesis) and develop their individ-
singular or dual
ual projects in the remaining time of eight weeks. The
paths.
final project submissions were first marked by the en-
tire group of studio tutors, second marked and mod-
erated by a team, not involved in tutorials and ex-
amined by external examiners. The entire process
was monitored during two semesters. Analysis of de-
sign outputs, marking statistics, moderator and ex-
ternal examiners reports, which will be presented in

Figure 4
Section of thermal
bath / SPA facilities.
During the phase of morphogenesis, a Grasshopper Figure 5

algorithm was developed in order to generate sur- Various shell types
face deformation prototypes in hills and valleys by in relation to the
simulating the flow of water along vectorised path- reciprocal unit.
ways. The algorithm required the introduction of 'at-
tractor paths' in order to develop various landscape-
like iterations. As a second step, a number of these it-
erations were materialized as physical models using
contour crafting, CNC milling and 3D printing tech-
nologies, which then offered up a range of different
tectonic solutions (Figure 3).
Finally, hill and valley typologies resulting from
the water-flow simulation process served as the spa- Figure 6
tial prototype for a spa facility, composed of hills, Physical model of
caves and valleys, and able to accommodate pools, double curved shell
saunas, steam rooms and relaxation zones (Figure with suspended
4). The building's double-curved roof helps its in- membrane.
tegration with the building site, a green, hilly land-
scape. Toyo Ito's reinforced concrete roof-scape
for the Kakamigahara Crematorium (Turnbull 2012)
served as a structural precedent for this project.
DESIGN EXPERIMENTATION 02: BIRD'S

NEST
This project analysed reciprocal structures in nature, During the morphogenesis phase the team exper-
such as they appear in birds' nests. Structures like imented with different surfaces composed of vari-
these are formed from at least two elements, each ous types of units, starting with physical models and
of which is supported by another. All elements meet continuing with grasshopper algorithms and digital
with each other along the span and not at the ver- modelling. In particular, it examined the relationship
tices. between unit, shape morphology and structural per-
formance, such as plane surfaces, and single and dou-

Figure 7 ble curvature (Figure 5).
Snake skin analysis Following these investigations, reciprocal timber
diagram. shell structures were developed in order to accom-
modate an event/pavilion space. Investigations into
cladding solutions included textile membranes (Fig-
ure 6) and ETFE foils.
DESIGN EXPERIMENTATION 03: SNAKE-

SKIN
This project analysed reptile skin and scales (Figure
7). Taking place within the educational framework
of the Raffles Design Institute in Shanghai, with the
assistance of third-year students, this study involved
translating snakeskin patterns into a scale-like, fold-
able surface, composed of units that were stuck to-
gether. The final surface was found to be flexible
enough to form a large variety of shapes. Hand-
made models were constructed from standardized
strip modules, but the final design was based on a
folded unit created using a parametric Grasshopper
algorithm. Thus, the prototype evolved into a non-
standard solution, made from irregular modules.
Finally, the prototypes were transformed into a
shading/space-dividing wall, with the 1:1 prototype
built from recycled cardboard panels. Each module
was laser-cut and then assembled layer by layer. The
final output proved to be an extremely stiff and stable
structure, which also allowed for spectacular lighting
effects (Figure 8).
Figure 8
DESIGN EXPERIMENTATION 04: BRANCHED
Finalized wall
installation. BLOCKS
This project analysed branching and growth algo-
rithms in trees. During morphogenesis, branching
algorithms appearing in nature were applied to dif-
ferent geometrical units, such as lines and cubic vol-
umes, by using a Grasshopper script (Figure 9). A
set of iterative models (digital and physical) was pro-
duced, exploring structure, organization, massing
and typology.
The final project proposal translates the typo-
logical prototypes produced during the form-finding
(Figure 10) exploration into a volumetric accumula-

tion of cubic spaces, aiming to accommodate a ther- and fabrication.
mal baths and spa facility. The same branching al- Other projects are concerned with using nature
gorithm was used as the primary organizational tool. as a source of spatial/typological innovation. The
The proposed building achieves an innovative spa- 'Hill and Valley' project, for instance, proposes a new
tial/organizational floor plan, and the massing is in- building type inspired by the topology of sand for-
tegrated into the landscape (Figure 11). mation. The entire spa complex evolves as a contin-
uous surface, hosting each of its functions either by
enclosure or through surface deformation into hills Figure 9
and valleys. Yet other examples, like the 'Branched Branching
Blocks' project, operate on an organizational/compo- algorithm studies.
sitional level. In this project a generative, branching
algorithm was applied in order to organize the lay-
out and massing of a building, again leading to un-
expected spatial and typological innovation.
Finally, some projects focus on the relation-
ship between geometry and structural performance
and/or material innovation (e.g. 'Snakeskin', 'Birds
Nest'). Both of these projects use algorithms as a
means of achieving structural and formal coherence,
while focusing on smart material fabrication solu-
tions by using either timber or paperboard- a re-
minder of the pre-computational attempts of Otto
and Candela.
Although biomorphic projects are often criti-
Figure 10
cised for ignoring contextual integration, case stud-
Algorithmic cube
ies like the 'Hill and Valley' project demonstrate a high
cluster.
degree of integration into the natural context. The
same can be said for the 'Branched Cubes' thermal
bath proposal, which manages to achieve contextual
integration with its design and materiality and could
therefore claim proximity to Wright's definition of or-
ganicism as a unification of design and nature. An-
other important observation is that biomorphic de-
DISCUSSION AND CONCLUSIONS sign solutions can operate on many different levels,
The way that nature has influenced design in the ranging from the size of an object or a building com-
above design experimentations falls into several cat- ponent, through to a whole building, or even projects
egories, which in turn give rise to a number of conclu- on an almost urban scale.
sions. Some projects, such as the 'Branched Blocks', The positive effects of including biomorphic de-
use nature mainly as a source of formal inspiration. sign in architectural education are obvious. The un-
Although this approach is often criticized by some, dergraduate design schemes shown here demon-
it has always been a legitimate approach to form- strate a high degree of structural and typological
finding, especially in relation to mathematical rules innovation, combined with well-functioning floor
and equations, and it leads to a chain of innovative plans and sections. The appropriate level of knowl-
solutions related to tectonics, typology, materiality

Figure 11
Finalized project
proposal for a
thermal bath.
edge has been acquired, as verified by external ex- phism and bionics for developing innovative solu-
aminers, enhanced by the acquisition of new skills tions and design-based research. They met all ac-
(e.g. digital design and fabrication). Such case stud- creditation requirements set by RIBA and ARB and
ies prove that undergraduate architectural education did not receive negative comments or remarks by
can move beyond the boundaries of classical curric- the external examiners, thus proved to be appropri-
ula and offer innovative design methodology that is ate for an undergraduate design studio module.
in tune with both the zeitgeist and with the evolving Meanwhile, the strict boundaries between terms
demands of the job market. In addition, simple, even such as biomimicry, biomorphism, zoomorphism,
playful, large-scale exercises like the Snakeskin in- geomorphism, anthropomorphism, organicism and
stallation wall prove that there is both metaphorical bionics are becoming ever blurrier. Emerging com-
and literal space for 1:1 scale constructions in under- putational tools and design techniques, such as gen-
graduate education. Traditional structures and hier- erative, algorithmic and parametric design, in combi-
archies in education, though, do present several bar- nation with digital technologies like CNC fabrication
riers to be overcome; this approach requires site sim- and 3D printing, are embracing nature as a source
plification, and teaching content must be adapted to of inspiration, and will allow constructive new syner-
the appropriate level, addressing antiquated and un- gies between biology and architecture in the years to
necessary health and safety requirements, and pro- come.
viding instruction in computational skills, an area
that is often ignored.
The projects featured in this paper clearly
demonstrate the enormous potential of biomor-

ACKNOWLEDGMENTS
My acknowledgments go to the design team of 'Hill
and Valley', Sen Lin, Haochen Jia, Jun Zhou and Jin-
hui Zhu, the design team of 'Bird's Nest', Maria Kaik,
George McLoughlin, Oluwaseyi Oladimeji, David Old-
ham, Yipei Tan, the design team of 'Sankeskin', Timur
Bissenbayev, Yuzhan Cheng, Qiuchen Yu, Minxuan
Zhu and Ben Prasetya Tanuatmaja and the design
team of 'Branched Blocks', Xiao Gu, Yuedi Liu, Yangt-
ing Yang, Jingley FU and Wenxuan Zhang.
In addition, my grateful acknowledgements go
to the Studio 04 tutors, Richard Dod, Daniel Wilt-
shire, Ürün Killic, Fei Chen and Simos Vamvakidis,
for their enthusiasm and hard work, and to the
workshop team of Stuart Carroll, Michael Baldwin,
Stephen Bretland and Aleksandar Kokai at the Liver-
pool School of Architecture.
REFERENCES
Agkathidis, A 2015 'Generative Design Methods: Imple-
menting computational techniques in undergradu-
ate architectural education', Proceedings of eCAADe
2015, Vienna, pp. 47-55
Aldersey-Williams, H 2003, Zoomorphic: New Animal Ar-
chitecture, Laurence King Publishing, London
Gruber, P 2011, Biomimetics in Architecture: Architecture
of Life and Buildings, Springer, Vienna and New York
Kosinski, D 2001, Henry Moore, Sculpting the Twentieth
Century, Yale University Press, New Heaven
Kuhlmann, D (eds) 1998, ‘Biomorphism in Architecture:
Speculations on Growth and Form, Bauhaus Univer-
sity of Weimar Publishers, Weimar
Nachtigall, W 2005, Biologisches Design, Systematischer
Katalog für bionisches Gestallten, Springer, Heidel-
berg
Turnbull, J 2012, Toyo Ito: Generative Order (Kassler Lec-
tures), Princeton Architectural Press, Princeton

Building Information Modelling - the Quest for Simplicity
Within Complexity
Anetta Kepczynska-Walczak1
1
Institute of Architecture and Urban Planning, Lodz University of Technology
1
anetta@p.lodz.pl
There is a common expectation of technology to better help us manage the

complexity of life and to simplify our daily tasks. However, these developments
also raise a question of whether design technologies encourage complexity at the
expense of simplicity in the design process. Does computation cause complexity?
Or does it enable simplicity? This paper aims to answer these key questions,
posed as the main focus of the eCAADe 2016 Conference, by confronting different
approaches to teaching Building Information Modelling (BIM) in schools of
Architecture. The scope of the paper is based on both the author's knowledge of
recent BIM implementations in the academic curricula and experiments
conducted at Lodz University of Technology. Necessary prerequisites enabling
understanding the complex knowledge are discussed. What is more, the scheme
for the integrated BIM pedagogy is proposed.
Keywords: Building Information Modelling, BIM, semantic model, information

visualization, integrated design
INTRODUCTION to address complex situations in architectural envi-

We often expect a technology to better help us man- ronments. On the one hand the more advanced tech-
age the complexity of life and to simplify our daily nological support may evoke more challenging tasks;
tasks. However, these developments also raise a se- on the other hand the tasks may provoke technolog-
ries of questions of whether design technologies en- ical growth and development. It is observed lately
courage complexity at the expense of simplicity in such intricate methods have been gradually involved
the design process. Does computation cause com- in the processes of building design, construction and
plexity? Or does it enable simplicity? [1] There is maintenance. The most meaningful is Building Infor-
no doubt such defined problematics require mani- mation Modelling.
fold analysis. To answer these questions and workout As Building Information Modelling is high on
a clear statement the paper will focus on Building In- agenda nowadays and influences the workflow in ar-
formation Modelling as a key area in a design process. chitectural design practice and industry, it is timely to
It is observed new digital technologies influence confront and discuss the topic. There are many ini-
design methods and offer exciting opportunities for tiatives in Europe in the subject area, consolidating
architectural design. Recent developments in com- regional BIM clusters, proposing EU funded projects
puter technologies and digital design tools enable us and other activities the author has been involved.

BIM has also become a very widely discussed re- a platform neutral, open file format specification. It
search topic, what can be proved by the presence is registered by ISO, and it is an official International
of the theme at conferences such as CAAD Futures, Standard. It is a commonly used collaboration format
ACADIA, SIGRADI, CAADRIA and eCAADe. A variety in Building Information Modelling based projects. To
of related papers can be given as examples: from the sum up, IFC is an open format for data exchange, par-
focus on BIM in education (Vinšová et al. 2014) and a ticularly intended to describe building and construc-
novel approach to combine BIM and VR (Kieferle and tion industry data.
Woessner 2015), through the application of BIM in To conclude, the mostly agreed definition nowa-
digital reconstructions of heritage objects (Boeykens days is to interpret BIM as a process involving the
et al. 2012), to programming for the enrichment of generation and management of digital representa-
BIM capabilities (Sharif and Gentry 2015). What is tions of physical and functional characteristics of
more, there are many other conferences and sym- places, while building information models (BIMs)
posia focusing on BIM entirely, gathering people mean files which can be exchanged or networked to
from academia and practice with an ample industry support decision-making about a place.
involvement.
A PLACE FOR BIM IN ACADEMIC CURRIC-
WHAT IS BIM NOWADAYS? ULA
Building Information Modelling has become a com- BIM has been receiving an increasing attention in the
mon concept and the term has been already adopted architecture, engineering and construction industry.
to everyday language. To some extent, this common Due to official regulations concerning the compul-
use is often exaggerated and excessively applied to sory use of BIM in practice and being adopted in more
3D modelling or even CAAD drawings. Therefore, it and more countries, there is an urgent necessity to
is timely to recap what BIM actually stands for. adapt the curricula in Architecture studies to the de-
There is a number of definitions of Building In- manding market.
formation Modelling, however, all of them point out Actually, there is an ongoing discussion on the
the same core characteristics. Thus, BIM can be un- best suited place in academic curricula for imple-
derstood as a concept, an approach, a process, a menting BIM concept and methodology. On the one
methodology, a knowledge-base for project infor- hand, in some institutions BIM is taught as early as in
mation, a technology or a set of tools essential to the first semester of Bachelor studies (Vinšová et al.
embed all necessary data about designed object in 2015). On the other hand, there are arguments to let
an integrated way. The richness and complexity of students mature for learning such complex tools, and
the output are strongly related to the scope and as a result, BIM is introduced at fourth of five years of
level of information possible to read or retrieve from studies (Magdy M. Ibrahim 2014).
the semantic model. This semantic model is the The author's personal experience in teaching
key to comprehend the idea of BIM, though it does BIM goes back to the 90s of the twentieth century.
not mean it encompasses the entire definition. One Experiments and didactics were based on the indus-
of the definitions states the following: "Building In- try first BIM software for architects. However, it was
formation Modelling (BIM) is a collaborative way of not defined BIM at that time but was named "Virtual
working, underpinned by the digital technologies Building Modelling in 1:1". By adding the scale 1:1, it
which unlock more efficient methods of designing, was emphasized it would reflect the real size of a de-
creating and maintaining our assets" [2]. Besides, it is signed building.
worth mentioning here, that BIM is not IFC. The lat- There is no doubt the evolution of IT in general,
ter stands for Industry Foundation Classes and it is and to point out more specifically, of CAAD software

in architectural design domain have been progress- courses represent the approach to teaching architec-
ing significantly since that time. That is why the re- ture design studio. What is more, as it is the early
vision questions arise: Does the use of complex de- stage of education (the third semester of bachelor
sign methods offer simplicity to the design process programme) proposed tasks should not be too com-
itself? Is it possible to design complexity with sim- plex. The idea is to provide simple steps to achieve a
ple methods? Or, in other words, to process complex complex result. Nevertheless, such courses require a
tasks with simple tools? set of prerequisites to enable students proceed with
In order to answer the above questions and to the tasks and develop new skills. What was surpris-
discuss a place for BIM in academic curricula, two ing, the BIM courses revealed a huge gap between
cases of BIM courses conducted at Lodz University of the two study programmes. While students of Archi-
Technology have been described, analysed and com- tecture were able to meet the requirements, students
pared. Generally, these two courses are delivered at of Architecture Engineering suffered from the lack
two bachelor study programmes which have been of basic knowledge on building structures. There-
expected to be similar, however, some key differ- fore, at the beginning the latter group was not able
ences appeared. The study programmes discussed in to proceed with preparation of an informative virtual
the paper are as follows: Architecture and Architec- model. So, the classes had to be reconfigured to help
ture Engineering. Both of them are eight semesters students achieve learning outcomes at the end of the
long, and BIM is taught at the same semester, viz. the semester. This situation provoked undertaking an in-
third semester of studies. A scope of the two courses vestigation on differences in curricula. Both study
and the learning outcomes have been planned the programmes are relatively new, so such experience
same, and the main goal is to teach BIM and how may lead to valuable conclusions and improvements
to apply it in the design process. The classes start of the curricula. For the purpose of the investigation
with an introduction to the BIM concept, then a set a comparison of pre-selected courses assumed to be
of exercises is prepared to allow students learn cho- prerequisites has been prepared (see Figure 1). Each
sen software supporting a design process enriched course has been described concisely in the following
by information about a building. Finally, a design paragraphs to picture the problem. The brief is based
task is provided to recap knowledge and to practice on the official description of courses published on-
newly developed skills. The theme is adjusted to line [3] and [4].
the level of third semester students' capacity, so in
a case of Architecture study programme, it is a fam- BIM prerequisites at Architecture study pro-
ily house at a given real location, while students of gramme
Architecture Engineering are asked to design a sum- Semester 1:
mer cottage on a chosen plot. The task is not a con- General Building I. Students are acquainted
cept design only. The basic requirement is, how- with the principles of technical drawing and basics
ever, to decide on building construction and mate- of building engineering. They are also familiarised
rials, to estimate the thickness and the skins of exter- with various construction types as well as typical so-
nal walls, a roof, slabs, fundaments. Apart from ge- lutions for basic building elements (wall, roof, beam,
ometry and structure the environment and climate slab). Students are expected to apply their knowl-
issues are taken into account and as a result, such edge throughout small projects such as a modular
information is embedded in the virtual model of deco-ordination of the plan and a design of external
signed building. There are some limitations planned walls for a single family house. (lectures and tutorials)
a priori. Hence both study programmes are lead- Building Materials. The aim of this course is to
ing to qualifications in architecture, these basic BIM provide students with information on basic building

materials (brick, wood, concrete, stone, glass, etc.), skills crucial for architectural profession - analysis of
their properties and applications. Some processes, given information and site; design representation
which take place in the building materials (heat trans- with drawings and models. Principles of professional
fer, moisture transport, hardening), are explained. ethic (respect, dependability and responsibility) are
The appropriate use (structure, insulation, finishing) also outlined. (lectures and studio)
of various materials is discussed. Potential risks and Semester 2:
dangers as well as methods of their mitigation are General Building II. Further information on
outlined. (lectures and lab) building methods is provided, including principles of
Introduction to Architectural Design. The load bearing; foundation types; principles of thermal
course aims at introducing basics of the designing insulation; stairs design. Students learn how to de-
process and its principles. A particular attention is sign various roof types and a ribbed ceiling slab. Ad-
given to stimulating architectural design thinking, ditionally, they are to present a written report from
creativity and responsiveness. Design exercises are their visit to the Building Industry Exhibition. (lectures
focused on solving simple problems requiring small and tutorials)
scale interventions. Students are expected to achieve Computer Aided Design Studio I. The aim of
Figure 1
Comparision of
prerequisites of two
BIM courses.

the course is to develop spatial cognition and to pro- semester. (lectures and tutorials)
vide an introduction to computer aided design as Timber Structures. The course is envisaged
a basic tool in engineering practice. Learning out- as an introduction to the basic properties of wood
comes include precise digital drafting, basic 3D mod- and wood-based materials. Traditional and contem-
elling and designing in the virtual environment. Stu- porary structural solutions are described. The is-
dents should be able to create 2D architectural draw- sues related to proper maintenance and exploitation
ings. They are also expected to perform a concept of a timber construction are also explained. Stu-
design of a simple architectural form in a chosen soft- dents apply the knowledge throughout designing
ware. (lab) a beam-column structure, a roof structure, and a
Introduction to Architectural Design II. The glued-laminated wood girder. (lectures and tutorials)
course includes three small-scale projects: 1. an ob- Building Installations. The course consists of
ject in open landscape (for example, a rest stop along two parts. The first part aims at acquainting students
a highway); 2. a similar facility in an urban setting (for with the basics of designing hot and cold water sys-
example, an entry to a botanic garden); 3. a more tems; fire-extinguishing water installations; domestic
complex design (for example, a holiday bungalow). and rainwater sewage systems in a building and its
Students are expected to focus on interrelation be- surroundings. The second part focuses on the con-
tween human being, architecture and environment struction and design of simple central heating sys-
(setting). At the end of the course they should be tems, mechanical ventilation systems and a variety of
able to merge functionally uncomplicated architec- heat sources. Both parts include lectures and a de-
tural forms into groups. (lectures and studio) sign studio. Students are expected to apply knowl-
Semester 3: edge to design installation systems in a given multi-
Building Physics. Students are provided with family house (a cold and hot water supply, an internal
the theoretical and practical knowledge of basic con- sewage system, sanitary and rainfall installation con-
cepts and problems concerning heat and mass ex- nected to the municipal sewage systems). Then they
change in a building shell, sound propagation, mois- focus on a design of a central heating or mechanical
ture exchange, energy balance of a habitable build- ventilation in a single-family building. (lectures and
ing, which are necessary for proper architectural de- tutorials)
sign process. Students are also familiarised with solar Integrated Architectural Design. Throughout
energy issues and their application including passive this course students learn the basic principles of ar-
and active systems and their efficiency. At the end of chitectural design of housing. During the lectures
the course they should be able to assess the effect of they are provided with theoretical and practical in-
a building modification and estimate the basic heat formation concerning a design task - a small complex
and energy characteristics of building elements. (lec- of single-family houses (at least three buildings) with
tures and tutorials) the emphasis on the urban context and functional
General Building III. The course extends stu- solutions (proper orientation, proportion and func-
dents' knowledge on technical and structural issues tional relationships, etc.); the appropriate structural
related to architectural design. Topics include foun- system and building materials should be also taken
dation methods and depths; water drainage systems; into consideration. (lectures and studio)
windows, doors, gates types, constructions, methods Computer Aided Design Studio II. The aim of
of fitting and sealing; flooring and floors; wall plas- the course is to provide an introduction to BIM ap-
tering and painting. The Polish Building Code is also plications, to develop practical skills in creating infor-
discussed. A documentation of a single-family house mation models of architectural objects and project
is to be compiled by each student by the end of the documentation based on the models and to learn es-

sential visualization techniques. On successful com- Architectural Design Studio I. The main goal
pletion of the course students should be able to ap- of the course is to familiarise students with a design
ply BIM concept to architectural design tasks. This in- process of residential buildings and complexes. The
cludes preparing a comprehensive documentation of course is based on a design process starting with a
architectural design based on the model information wide range of analysis of existing exterior site condi-
as well as creating sets of detailed data on the archi- tions, through the formulation of architectural forms
tectural objects based on the spatial BIM model. (lab) using adequate materials, to finding optimal archi-
tectural solutions. The design studio is enhanced
BIM prerequisites at Architecture Engineer- with lectures expanding on associated issues, such as
ing study programme the impact of the building on the user, functional lay-
Semester 1: out and interrelations between building and its sur-
Building Materials. The aim of this course is to rounding, residential buildings typology, etc. (lec-
provide students with information on basic building tures and studio)
materials (brick, wood, concrete, stone, glass, bitu- Computer Methods in Architecture I. The aim
minous adhesives, mineral binders, etc.), their prop- of the course is to develop spatial cognition and to
erties and applications. Some processes, which take provide an introduction to computer aided design as
place in the building materials (heat transfer, mois- a basic tool in engineering practice. Learning out-
ture transport, hardening) are explained. The appro- comes include precise digital drafting, basic 3D mod-
priate use (structure, insulation, finishing) of various elling and designing in the virtual environment. Stu-
materials is discussed. Potential risks and dangers as dents should be able to create 2D architectural draw-
well as methods of their mitigation are outlined. (lec- ings. They are also expected to perform a concept
tures and lab) design of a simple architectural form in a chosen soft-
Fundamentals of Civil Engineering. The ware. (lab)
course covers an introduction to the principal build- Semester 3:
ing elements, structural systems and traditional tech- Building Physics and Acoustic. Students are
nologies including ventilation, exhaust and smoke provided with the theoretical and practical knowl-
ducts. The problem of loads is emphasised. Students edge of basic concepts and problems concerning
are expected to make calculations for determining heat and mass exchange in a building shell, sound
the value of loads, including an impact of functional propagation, moisture exchange, energy balance of
arrangement and climatic conditions. This project is a habitable building, which are necessary for proper
to be done for a given rib-and-slab floor. (lectures and architectural design including lighting and acoustics.
tutorials) Methods of numerical analysis of heat flow including
Introduction to Architectural Design and Er- 2D and 3D modelling are outlined. Students are also
gonomics. The course focuses on the basic knowl- familiarised with solar energy issues and their appli-
edge of architectural design ideas and principles. cation including passive and active systems and their
Students should learn the rules of architectural draw- efficiency. At the end of the course they should be
ing, methods of presenting and discussing architec- able to assess the effect of a building modification
tural issues. To achieve basic understanding they are and estimate the basic heat and energy character-
expected to prepare a presentation on a modern/- istics of building elements. Students are expected
contemporary well known architect. The final assess- to conduct calculations concerning heat exchange,
ment is a design of a small building (expo pavilion, energy balance, and reverberation time for selected
summer house, etc.). (tutorials and studio) rooms. (lectures and tutorials)
Semester 2: Concrete Technology. The course aims to famil-

iarize students with the technology of cement con- OBSERVATIONS AND CONCLUSIONS
crete including issues related to properties of con- The following deliberation presents the perspective
crete and the requirements, concrete classification, of teaching computer aided architectural design and
normalization and classes, components of concrete being situated in predefined settings. The main mo-
mixtures, their workability, placing and curing. Meth- tive to undertake the above search was the observa-
ods of concrete testing are also outlined. (lectures and tion that students of two courses of study, viz. Ar-
lab) chitecture and Architecture Engineering, taking part
Building Installations. Students are introduced in BIM courses had differential background knowl-
to the classification-related terminology, designing, edge. It is worth mentioning here, that it was in-
operations, setting up and exploitation of typical san- tended these two courses should differ in the lan-
itation systems and district heating. Tutorials include guage of instruction only. The Architecture course is
assessment of the influence of particular installa- taught in Polish, whereas the Architecture Engineer-
tions on hygiene and sanitation safety, the acoustics- ing course is delivered in English. Nevertheless, there
related comfort and fire hazard, people, buildings are diverse names of particular courses, the overall
and the surrounding. In addition, students are ex- learning outcomes are the same. Both courses con-
pected to design a simple central or water heating sist of eight semesters, both are concluded by final
installation, based on legal requirements and princi- bachelor diploma projects and both promote engi-
ples of technical knowledge. (tutorials and studio) neers in architecture. What is more, the two study
Architectural Design Studio II. During this programmes are relatively new, so such investiga-
course students are acquainted with the architec- tion may reveal the most emergent conflicts, lead to
tural design of small commercial buildings (a hotel, a valuable conclusions, and in consequence, improve-
kindergarten, local bus/railway station, etc.) in an ur- ments of the curricula.
ban context. Lectures provide information not only The examination of the courses, presumed to be
on the issues related to the design task (e.g. ho- prerequisites to teaching BIM, have helped to under-
tel structure & programme; vertical communication stand differences in the learning progress. Through
in building), but also on various concepts behind the analysis of students' earlier learning outcomes
the modern and contemporary architecture, such as: and prerequisites, and moreover, by an observation
"less is more", "we do not create art we solve prob- of diversified knowledge it was possible to denomi-
lems", "less is bore" and many other. (lectures and stu- nate key problems. First of all, as the lack of coher-
dio) ence in students' knowledge emerged very early at
Computer Methods in Architecture II. The aim the "Computer Methods in Architecture II" course, it
of the course is to provide an introduction to BIM ap- was possible to adapt the course to students needs
plications, to develop practical skills in creating infor- while keeping the goal and project tasks unchanged.
mation models of architectural objects and project Thus, it can be considered as an emergent reaction
documentation based on the models and to learn es- to the situation. Secondly, a study of complemen-
sential visualization techniques. On successful com- tary courses was undertaken. The primary impres-
pletion of the course students should be able to ap- sion while reading the courses catalogues was rather
ply BIM concept to architectural design tasks, and in not unexpected. The courses looked alike and dis-
particular to prepare a comprehensive documenta- tinguished tiny differences were not supposed to
tion of architectural design based on the model in- be the reason of the problem. However, after sev-
formation, as well as to create sets of detailed data eral discussions with students and observations of
of the architectural objects based on the spatial BIM their state of knowledge, it became perfectly ob-
model. (lab) vious the prerequisite courses were not the same.

Figure 2
Integrated
multidisciplinary
BIM course.
The courses for Architecture Engineering students it requires a set of prerequisites not possible to ap-
focused more on civil engineering tasks and calcu- prehend in such a short time. Secondly, to under-
lations while students of Architecture learnt more stand BIM basics it is essential to learn building struc-
about building structures with the capacity essen- tures first. Until then we may speak about two di-
tial for architects. Additionally, students of Archi- mensional drafting or three dimensional modelling,
tecture benefited more since the "Computer Aided but it is not equal to building information modelling,
Design Studio II" course became complementary to however, often mistaken even by the teachers. An-
other courses provided at the third semester, viz. In- other conclusion is that BIM courses should be pro-
tegrated Architectural Design, General Building III, vided on several levels of education and should start
Building Installations, Building Physics and Timber as a course dedicated to certain study programme,
Structures. It was the initiative beyond the stiff cat- for example for architecture students, civil engineer-
alogue courses' descriptions. As a result, the de- ing students, environmental engineering students,
sign studio project benefited from the associated etc. The BIM course described earlier would stand
courses and students performed better understand- for the basic, introductory one. Then, an advanced
ing of complex knowledge. What is more, they were course should be proposed enabling working in mul-
able to develop also technical skills and apply them tidisciplinary teams and offering a simulation of the
in their final projects. real problems to be solved in a co-operative way. It
The observations allow to draw some conclu- is worth mentioning here, that among many defini-
sions. First of all, BIM is not suitable for the first tions, finally BIM is about working together.
semester or even the second semester students since To sum up, there is no doubt BIM process is com-

plex and multifaceted, however through carefully de- tectural, structural, environmental and ICT solutions.
signed curriculum it is possible to make such com- To sum up, the aim of this newly proposed course is to
plex knowledge and practical skills simple to acquire. provide a scenario for a simulation of a complex de-
sign process and preparing students to meet the AEC
FURTHER PLANS industry expectations. To achieve such complex goal,
Hence BIM is high on agenda and it is a real need to it will be necessary to involve an innovative teach-
prepare students to meet the growing challenges of ing methods based on Design Thinking and Problem
industry and technology, a new concept of interdis- Based Learning (Kepczynska-Walczak 2016).
ciplinary design process simulation course based on
BIM has been proposed. It is planned to continue the FINAL REMARKS
idea of teaching by "simple steps towards complex The plethora of new possibilities generates chal-
results". lenges in education. They are assumed not to pro-
The idea coincides with a new programme of voke replacements but rather extensions to the old
undergraduate studies "Control Systems Of Intelli- and tested methods. Furthermore, taking into ac-
gent Buildings" launched at the Faculty of Electri- count also the tendencies on the market, it is possible
cal, Electronic, Computer and Control Engineering to draw some conclusions on how the design educa-
at Lodz University of Technology. The curriculum tion should be developed in the near future. There-
has been developed in co-operation with other fac- fore, the proposed scenarios would be worth consid-
ulties, among them the Faculty of Civil Engineering, ering in a broader discussion.
Architecture and Environmental Engineering. The The impact of industry on academia is increasing,
new concept for advanced BIM course would include so it is worth considering how these two may benefit
an integrated design studio merging students from from each other. It refers directly to the concept of re-
both faculties and four courses of study (see Figure 2). search by design (Verbeke 2013; Herneoja et al. 2015)
They would form interdisciplinary teams gathering perfectly reflecting research activities conducted in /
representatives from different fields of studies, prin- with and through practice.
cipal to challenge intelligent buildings design. So, There is no doubt BIM belongs to highly com-
students of architecture, civil engineering, environ- plex concepts, though learning it can be a simple and
mental engineering (building installations) and con- successful process. Besides the factors described ear-
trol systems of intelligent buildings would work to- lier, the success depends on how carefully and explic-
gether to bring a project to fruition: from the initial itly BIM course is planned and delivered. It is worth
concept to the 'in use' phase [5]. Participation of stu- mentioning here the author's approach towards dis-
dents and tutors of versatile backgrounds and knowl- cussing BIM in the context of Complexity & Simplic-
edge would stimulate exchange and, what is more, ity has been formulated from the perspective of com-
simulate a design process as close as possible to the puter aided architectural design education.
reality. The latter would be achieved through the in- The author believes the paper will contribute to
volvement of extramural stakeholders from the AEC better understanding of BIM concept, to teaching
industry. It is necessary to stress, this sector is very methodologies development, to a discussion on up-
active in the adoption of Information and Commu- dating the academic curricula and strengthening the
nication Technologies. Therefore, Integrated Project BIM idea in a design studio context. It is time for
Delivery (IPD) is highly on demand. It is an emerging BIM - complex in its nature - to simply immerse in
process framework bringing together people, sys- academia!
tems, business structures and practices to keep infor-
mation flowing smoothly towards integrated archi-

REFERENCES 20programme&tryb=full-time
Boeykens, S, Himpe, C and Martens, B 2012 'A Case Study [5] http://www.building.co.uk/the-six-phases-of-bim
of Using BIM in Historical Reconstruction: The Vi- /5059652.article
nohrady synagogue in Prague', Digital Physicality -
Proceedings of the 30th eCAADe Conference - Volume
1., Prague, pp. 729-737
Herneoja, A, Pihlajaniemi, H, Österlund, T, Luusua, A and
Markkanen, P 2015 'Remarks on Transdisciplinarity
as Basis for Conducting Research by Design Team-
work in Real World Context through Two Case Stud-
ies of Algorithm Aided Lighting Design', Real Time -
2, Vienna, pp. 61-70
Kepczynska-Walczak, A 2016 'In Search of the Best Peda-
gogy in Teaching Architecture and Urban Planning',
EDULEARN16 Proceedings, Barcelona, pp. 5420-5429
Kieferle, J and Woessner, U 2015 'BIM Interactive - About
combining BIM and Virtual Reality - A Bidirectional
Interaction Method for BIM Models in Different Envi-
ronments', Real Time - Proceedings of the 33rd eCAADe
Conference - Volume 1, Vienna, pp. 69-75
Magdy, MI 2014 'Thinking the BIM Way - Early integra-
tion of Building Information Modelling in education',
Fusion - Proceedings of the 32nd eCAADe Conference -
Volume 2, Newcastle upon Tyne, pp. 427-435
Sharif, S and Gentry, R 2015 'BIM for Masonry: Develop-
ment of BIM Plugins for the Masonry Unit Database',
Real Time - Proceedings of the 33rd eCAADe Confer-
ence - Volume 1, Vienna, pp. 567-576
Verbeke, J 2013, 'This is Research by Design', in Murray, F
(eds) 2013, Design Research in Architecture, Ashgate
Publishing, Farnham, pp. 137-161
Vinšová, I, Achten, H and Matejovská, D 2015 'Integrat-
ing BIM in Education: Lessons Learned', Real Time -
2, Vienna, pp. 127-131
Vinšová, I, Matejovská, D and Achten, H 2014 'The Un-
bearable Lightness of BIM', Fusion - Proceedings of the
32nd eCAADe Conference - Volume 2, Newcastle upon
Tyne, pp. 411-415
[1] http://www.oulu.fi/architecture/2016ecaade/them
e
[2] https://www.gov.uk/government/uploads/system/up
loads/attachment_data/file/34710/12-1327-buil
ding-information-modelling.pdf
[3] http://programy.p.lodz.pl/kierunekSiatka.jsp?l=
en&w=Architecture&p=4827&stopien=first-cycle%
20programme&tryb=full-time
[4] http://programy.p.lodz.pl/kierunekSiatka.jsp?l=
en&w=Architecture&p=4935&stopien=first-cycle%

CAAD EDUCATION | Fabrication
Crafting as Inquiry into Computation
Exploring wire-bending in traditional practice and design education
Vernelle A. A. Noel1
1
The Pennsylvania State University
1
vernelle@vernellenoel.com
In an effort to sustain and revitalize the dying craft of wire-bending, I present the
development of three novel approaches: Digital Crafting, Computational
Crafting, and Crafting Fabrication. Computation and digital technology were
integrated with traditional wire-bending principles to design and fabricate
artifacts. In this work, I present three artifacts resulting from these novel
approaches, and analyze how these methods may be used in design education and
practice. Our findings benefit the practice because it offers opportunities for
teaching computational and traditional skills to older and younger generations
through wire-bending.
Keywords: wire-bending, Bailey-Derek Grammar, craft, computational design,

fabrication
Figure 1 INTRODUCTION
Wire-bender Traditional craft cultures and practices are disap-
performing pearing due to globalization, commercialization, and
wire-bending. techno-centric developments [1]. There has been in-
Photo (C) 2015 creasing concern around the world about the decline
Vernelle A. A. Noel. and potential disappearance of culturally significant,
traditional craft practices in Trinidad and Tobago, Sin-
gapore, Japan, and Indonesia, just to name a few.
Scholars have been looking to craft as a teaching tool,
leading to the introduction of crafting as a methodol-
ogy for teaching electronics, programming, and tool
development (Rosner and Ryokai 2009; Buechley and
Perner-Wilson 2012; Buechley and Eisenberg 2009; Wire-bending is a "specialized art, combining ele-
Jacobs and Buechley 2013; Zoran and Paradiso 2013). ments of structural engineering, architecture, and
This paper focuses on how preservation, education, sculpture" to create two-dimensional (2D) and three
and practice in the traditional craft of wire-bending dimensional (3D) forms [2]. This practice was devel-
can benefit from computational methods. oped in the 1930s in carnival in Trinidad and Tobago.
It became "one of the most highly developed of the
CAAD EDUCATION | Fabrication - Volume 1 - eCAADe 34 | 311

carnival crafts" (Crowley 1956). In this practice, wire that Bailey uses wire like a pen or pencil, materializing
and other thin, flexible strands of material are bent his design ideas in wire without the use of drawings
with hand tools and assembled to create structures or sketches as representations for design or fabrica-
(Figures 1 and 2). Wire-bending however, is on the tion. Derek sketches his design ideas on paper, but
decline due to old artisans retiring or dying, the slow his knowledge on which materials and connections
rate of transmission of knowledge, the younger gen- to use during fabrication of the artifact remains hid-
eration's love of technology, and the craft's labor- den. These wire-benders use their mental repertoire
intensive nature (Noel 2013). This project begins to of connection details, their palette of tools, and ma-
demonstrate ways of reinterpreting a traditional craft terials, to fabricate these wire-bent structures (Figure
to integrate computation and digital technology. It 3). This lack of externalization is a factor in the slow
builds on crafting as a methodology for the devel- rate of transmission of this knowledge.
opment of culturally-relevant computation in design A second reason for the craft's decline is its labor-
education, practice, tools, and methods. It does intensive nature. Historically, most wire-benders in
not attempt to replace a traditional practice with a the Trinidad Carnival have been men. One reason
techno-centric one, but instead endeavors to explore for this may be the manual strength and dexterity
how wire-bending could benefit from computation; needed to bend, cut, and clinch wires and other com-
address waning interest by the younger generation; ponents. Currently a lacuna of women, children,
and investigate the possibilities for design education and the physically infirmed exists in traditional wire-
between craft and computation. bending. A third reason for the craft's decline is the
younger generation's love of technology. A large per-
centage of youths today are familiar with, and enjoy Figure 2
engaging with digital technologies like computers, Large dancing
gaming programs, and smart phones to name a few. sculpture in the
Traditional wire-bending is a low-tech practice with Trinidad Carnival
little to no opportunity at this time for using comput- constructed using
ers or other digital technologies. wire-bending
Through this project I will attempt to find an- principles. Photo
swers to the following questions: How can computa- (C) 2010 Vernelle A.
tion aid in making wire-bending more inclusive, ad- A. Noel. (Noel
dressing the lack of participation by women, children, 2015). Image
and the physically infirmed? How can computation republished with
address the labor-intensive nature of the craft? How permission.
can the younger generation's interest in, and famil-
iarity with digital tools, and technologies be used to
revitalize interest and participation in wire-bending?
My experiences in traditional craft, the cultural de-
In traditional wire-bending, as with many other tra- sign practice of Trinidad Carnival, computational de-
ditional crafts, knowledge is mostly tacit - residing sign, and education motivates this work on craft prac-
in the expert, with little separation between design tices, computation, and technology. To my knowl-
and fabrication. Designers usually have the ability edge, no established pedagogy on wire-bending or
to materialize their ideas without externalization via design computation exists in the Republic of Trinidad
sketches or drawings. This is the case with expert and Tobago (Trinidad and Tobago). This project ex-
wire-benders Albert Bailey, and Stephen Derek. Dur- plores ways of integrating computation, digital de-
ing my fieldwork and interviews in 2013, I observed
312 | eCAADe 34 - CAAD EDUCATION | Fabrication - Volume 1

Figure 3
The traditional
wire-bending
process.
sign, and technology in a culturally-relevant manner means through which designers, sculptors, and oth-
into craft cultures to aid in their sustainment, revital- ers interested in wire-bending can begin to under-
ization, and advancement. stand the craft; and to teach computational princi-
The organization of this paper is as follows. This ples in a culturally-relevant manner (Figure 4).
section gives a brief introduction and background In Muslimin's study of a parametric approach to
to the problem, and the practice of wire-bending. fabrication for traditional ceramics, he tests the use
Section 2, presents related work that merges and of digital modeling and CNC routers to directly fab-
(re)configures traditional models of practice using ricate negative molds, in lieu of traditional mold-
computational approaches. In Section 3, I introduce making processes. He also reconfigures fabrication
three (3) crafting methodologies - Digital Crafting, devices based on digital methods of transformation,
Computational Crafting, and Crafting Fabrication - illustrating the versatility of computation in the form
for making artifacts based on wire-bending princi- of visual embedding and shape grammars (Muslimin
ples. In Section 4, I present the results of these ap- 2013). In his work on weaving, Muslimin analyzes
proaches. Section 5 concludes the paper and out- tacit knowledge embedded in traditional weavers to
lines future work. develop computational design rules, and combine
traditional principles with digital tools. Through this
RELATED WORK investigation into traditional weaving he develops a
In a previous study, Noel developed the Bailey- new assembly system and a weaving grammar, ex-
Derek Grammar, a computational design tool that de- tending the language of architectural weaving (Mus-
scribes the materials, connections, rules, and steps in limin 2010a; 2010b; 2014). In his work on hybrid bas-
wire-bending from Trinidad and Tobago (Noel 2015). ketry, Zoran attempts to merge the digital and the
This design tool was developed to aid expert wire- traditional by using hand weaving techniques to lace
benders in their practice and teaching; to provide a reeds through 3D printed structures he digitally de-
Figure 4
Application of the
Bailey-Derek
Grammar in the
wire-bending
process.

signed and fabricated (Zoran 2013). In another work, category of work. Computational Crafting, I define
he develops a handheld digital milling device that as the employment of low-tech tools and processes
enables carving of digital 3D objects out of phys- from a traditional craft, in the design and fabrication
ical material (Zoran and Paradiso 2013; Zoran and of artifacts. Noel's previous work demonstrating ap-
Paradiso 2012). Schulte et. al. attempts to reclaim plication of the Bailey-Derek Grammar is an example
the forgotten technique of steam bending through of work in this category. Lastly, Crafting Fabrication
digital steam bending, design, and form generation is the use of digital and non-digital tools, processes,
(Schulte et al. 2014). These previous studies relate to and technologies from a traditional craft, in the de-
mine in their application of computational and digi- sign and fabrication of artifacts. It is an attempt at a
tal strategies in traditional crafts, combining the dig- hybrid process that captures the affordances of craft-
ital and the traditional, and attempting to revive his- ing traditionally, computationally, and digitally. Zo-
torical techniques. ran's work on weaving, and Schulte et al. work in
steam bending are examples of this category. The
CRAFTING AND COMPUTATION following experiments explore how the traditional
In this work I shall describe Digital Crafting as the craft of wire-bending might be sustained and revital-
reinterpretation of a traditional craft by using digi- ized through a computational approach, for applica-
tal processes and technologies in making artifacts. tion in education and practice.
This includes employing digital programs for design,
optimization, and automation, and CNC machines Experiment 1: Digital Crafting
such as 3D printers or CNC routers for digital fab- The first experiment attempts to use digital ap-
rication. It does not currently include the assem- proaches and technologies in the (re)conceptualizing
bly process. Design in digital crafting comprises dis- of wire-bending. This experiment was motivated
crete descriptions and steps, such as shapes, num- by my desire to develop a piece of software for the
bers, or words (code). Muslimin's work in ceramics design and fabrication of artifacts using points and
mentioned previously would be an example of this
Figure 5
Steps the author
went about in
developing the
Digital Crafting
Method.

lines. The software (program) would take a digitally (steps 6 and 7). All the components are then assem-
modeled surface as input, and generate 3D geome- bled to create the final artifact.
try (lines and points) needed for its fabrication. Tools
and materials included a computer, Rhino3D soft- Experiment 2: Computational Crafting
ware, RhinoPython, a 3D printer, tape measure, and The second experiment combines the Bailey-Derek
wooden rods. Firstly, I designed 3D connections to Grammar, and traditional wire-bending techniques.
receive rods (wire or wood), then modeled them in Tools and materials included 18-gauge wire, a tape
Rhino 3D. I took note of each step taken to model the measure, pliers for bending and cutting, a felt pen
connections, researched examples of algorithms in for marking, masking tape for adhesion, and a 2D
RhinoPython, then wrote my own algorithms to auto- template of the design. The Computational Crafting
mate the process. To begin developing the algorith- method illustrated in Figure 6, began with drawing
mic script, I had to find out which digital tools (com- a template as a guide for bending the wire (step 1). I
mands) were available to me. I tested each one of the used my hands and tools (step 2), to bend, smoothen,
875 commands available in RhinoPython to under- and manipulate wires so they followed the outline on
stand their functions, variables, inputs, and outputs. the template. As wire was bent along the outline,
Just as crafts persons would familiarize themselves masking tape adhered bent segments to the tem-
with the tools available to them, and understand how plate, so it would keep its shape. After bending all the
they work, I did the same. The final program does the longitudinal and transverse wires, I assembled the ar-
following: tifact. The Bailey-Derek Grammar provided the in-
structions needed to make the connections needed
1. Generates points and lines on a surface ac- for assembly (steps 3 and 4).
cording to the resolution desired by the user;
2. Generates the geometry for connections ac- Experiment 3: Crafting Fabrication
cording to user requirements; The third experiment combines digital and non-
3. Labels the connections; and digital tools, processes, and technologies. Tools and
4. Outputs line (rod) lengths needed for fabrica- materials included 1/16" diameter steel wire, a com-
tion. puter, Rhino 3D software, tape measure, pliers, a felt
tip pen, masking tape, and a CNC wire-bender. The
The Digital Crafting method illustrated in Figure 5, Crafting Fabrication method, illustrated in Figure 7,
takes a 3D model of an artifact as input (step 2). By began by drawing 2D outlines of the design in Rhino
changing variables in the algorithm, users can gener- 3D (step 1). These drawings were exported, then
ate several design options (Step 4). After users have opened in the DiWire interface to work with the CNC
chosen their desired design for fabrication (step 5), wire bender. Points on the line had to be edited be-
the 3D connections are generated, and 3D printed
Figure 6
Steps involved in
the Computational
Crafting Method.

Figure 7
The Crafting
Fabrication
Approach.
fore sending to the CNC wire bender (steps 2A and online. I designed algorithms to automate the gen-
2B). Since the resolution of the curve in Rhino 3D was eration of the connections and output information
higher than that able to be bent by the CNC wire needed for fabrication; used a 3D printer to digitally
bender, some points had to be deleted and others fabricate the connections; measured and cut rods;
moved until the new line was a best fit to the original. and assembled the artifact (See [3]).
Straight wire was inserted into the machine where it In the Computational Crafting approach to wire-
was bent according to the design (step 3). Each piece bending I drew the lines for my design in digital de-
of wire was cut and closed to form a loop with mask- sign software (Rhino 3D) to create a template; bent,
ing tape. All the components were then assembled measured and cut wires with hand tools; and fol-
to form the artifact with the Bailey-Derek Grammar lowed instructions on wire-bending based on the
providing fabrication instructions (steps 4, 5 and 6). Bailey-Derek Grammar [See 4]. In addition, after fabri-
cating and assembling the artifact, I documented the
RESULTS steps taken so the processes and knowledge could be
In this project, I develop three approaches to the tra- shared with others (Figure 9). I employed my knowl-
ditional craft of wire-bending: Digital Crafting, Com- edge of shape grammars and computational design,
putational Crafting, and Crafting Fabrication. to illustrate the rules and steps used to fabricate the
In the Digital Crafting approach, I reinterpreted artifact.
traditional wire-bending. I digitally modeled 3D con- In the Crafting Fabrication approach, I drew lines
nections and forms, wrote algorithms, and digital for constructing the artifact in digital design soft-
fabricated an artifact (Figure 8). I used 3D modeling ware, measured and cut wire; used a CNC wire-
skills to design and model the connections. In order bender to digitally fabricate those lines; and followed
to design these connections, I referred to previous the instructions shown in Figure 9 to assemble the ar-
literature and documentation on wire-bending in tifact. (See [5] and [6]).
Trinidad and Tobago, and searched for other projects

Figure 8
Artifacts resulting
from 3 different
approaches to
crafting in
wire-bending.
Figure 9
Instructions for
fabricating the
artifact.

DISCUSSION AND FUTURE WORK which older experts and the younger generation can
This project set out to investigate how education and share knowledge between each other. Older experts
practice in the traditional craft of wire-bending could can teach novices how, why, and when they make
benefit from computational methods to address sus- certain connections while computationally minded
tainment of the craft, the younger generation's love designers can demonstrate the benefits of external-
of technology, and the labor-intensive nature of the izing this knowledge. Figure 9 can be used to teach
practice. The skills that can be taught using these others how to fabricate an artifact, giving experts
three (3) approaches to wire-bending are shown in more time to deal with invention rather than conven-
Table 1, while the ways in which the practice benefits tion, and increase the rate at which this knowledge is
from this integration is highlighted in Table 2. shared.
The findings suggest that the Digital Crafting ap- Results from the Crafting Fabrication approach
proach to wire-bending can be used to teach 3D to wire-bending show that it can be used to teach
modeling, programming, knowledge sharing, and digital drawing and representation; traditional wire-
digital fabrication. Designers and students can learn bending techniques; and how to use CNC machines
design computation, apply it in the cultural prac- and hand tools. Employing the CNC machine re-
tice of wire-bending, carnival, architecture, and other duces the labor-intensive nature of bending by hand,
possibilities extending beyond its current applica- thereby opening up the practice to those with phys-
tions. Digital crafting can also encourage dialogue ical limitations. This approach is also a mediating ar-
between traditional wire-benders, and the computa- tifact for youths to teach older experts how to draw
tional generation since the expertise and knowledge lines on the computer, and bend them with the CNC
of traditional experts can inform computational tech- desktop machine. Because of the experts' experience
niques to develop the novel. This assembly process with tools and machines, they may be able to create
in Digital Crafting was not very labor-intensive, be- novel ways of using the wire-bending machine in col-
ginning to create a space where women, children, laboration with the younger generation.
and the infirmed can participate. One limitation of Employing these computational tools and meth-
this method includes the limited sensory engage- ods in this culturally relevant manner can be an av-
ment between the crafts-person and the wire, as lit- enue through which users explore design in carnival,
tle to no actual bending and manipulating of wire architecture, and art. These approaches can begin to
occurred. Another limitation is the inability to make develop research and design curricula around craft,
design changes during fabrication, since the compo- computation, and digital technology through wire-
nents are discrete and design changes must be made bending, dancing sculptures, architecture, and carni-
earlier in the process. val. Reddy says, "if the language skills and the habit
Results from the Computational Crafting approach to of engaging in reconstruction are not [...] preserved,
wire-bending illustrate that the method can be used then there will be no culture" (Reddy 1993). The ap-
to teach students traditional techniques, computa- proaches to crafting that incorporate computation, il-
tional thinking, and experience sensorial engage- lustrated above, can benefit preservation, education,
ment with their materials and tools. They can also and practice in the traditional craft of wire-bending
practice dexterity, learning much from the flow of since it describes wire bending with different meth-
materials and tools (Ingold 2013). Students can learn ods: algorithms, drawings, and computations. Hav-
to appreciate the skill, innovation, and history of the ing access to these varied approaches is important for
craft, and those who created it, to inform the future, sustainment and (re)configuration of the craft for de-
and the sustainability of the craft. The computational sign practice and education.
crafting method acts as a mediating artifact through Some questions to be addressed in future work

Table 1
Skills that can be
taught through
these 3 approaches
to crafting in
wire-bending.
Table 2
Affordances of the
different
approaches to
crafting in
wire-bending.

are: How can classifying materials for application in Ingold, T 2013, Making : Anthropology, Archaeology, Art
the Bailey-Derek Grammar extend the grammar, and and Architecture, Taylor and Francis, Hoboken
inform design practice, education, and innovations? Jacobs, J and Buechley, L 2013 'Codeable Objects: Com-
putational Design and Digital Fabrication for Novice
How can classifying tools and methods for design
Programmers', Proceedings of the SIGCHI Conference
and fabrication lead to hybrid systems? How can on Human Factors in Computing Systems, New York,
the structural properties of these connections be em- NY, USA, pp. 1589-1598
bedded visually in designs? Which of the three ap- Muslimin, R 2010a, 'Parametric Fabrication for Tradi-
proaches facilitates deeper cognitive processing in tional Ceramics', Proceedings of SIGRADI 2010, pp.
expert wire-benders, and novices? Future steps in 222-228
Muslimin, R 2010b, 'Learning from Weaving for Digi-
this study shall include: (1) sharing and testing these
tal Fabrication in Architecture', Leonardo, 43(4), pp.
approaches with others to observe what they learn 340-349
from these approaches; (2) conducting design and Muslimin, R 2014, EthnoComputation : on Weaving Gram-
fabrication workshops with practitioners and stu- mars for architectural design, Ph.D. Thesis, Mas-
dents; (3) developing educational construction kits sachusetts Institute of Technology
and systems for children and designers; (4) develop- Noel, VAA 2013, Trinidad Carnival: Improving Design
through Computation and Digital Technology, Ph.D.
ing a primer to teach wire-bending, 3D modeling,
Thesis, Massachusetts Institute of Technology
programming techniques, and digital fabrication to Noel, VAA 2015, 'The Bailey-Derek Grammar: Recording
absolute beginners; and (5) continued leveraging of the Craft of Wire-Bending in the Trinidad Carnival',
computational approaches and digital technologies Leonardo, 48:4, pp. 357-365
to progress the practice. Reddy, M 1993, 'The conduit metaphor: A case of frame
conflict in our language about language', in Ortony,
A (eds) 1993, Metaphor and Thought, Cambridge
ACKNOWLEDGEMENTS University Press, Cambridge, pp. 164-201
I would like to acknowledge Felecia Davis, Albert Rosner, D.K. and Ryokai, K 2009 'Reflections on Craft:
Bailey, and Stephen Derek for their guidance and Probing the Creative Process of Everyday Knitters',
support. This work was funded in part by The Proc. of Creativity & Cognition 2009. (2009), pp. 195-
204
Government of the Republic of Trinidad and To-
Schulte, M, Mankouche, S, Bard, J and Ng, TY 2014,
bago, and The Stuckeman Center for Design Comput- 'Digital Steam Bending: Re-Casting Historical Craft
ing. Grateful acknowledgement is given to CARIRI, Through Digital Techniques', ARCC Conference
NEDCO, Danielle Oprean, Shivaram Punathambekar, Repository, 0(0)
José Pinto Duarte, and Eduardo Castro e Costa for Zoran, A 2013, 'Hybrid Basketry: Interweaving Digi-
their support. Special thanks to my friends at the tal Practice within Contemporary Craft', Leonardo,
46(4), pp. 324-331
Stuckeman Family Building.
Zoran, A and Paradiso, J 2012 'The FreeD: a handheld dig-
ital milling device for craft and fabrication', Proceed-
REFERENCES ings of CHI 2013: Changing Perspectives, p. 3
Buechley, L and Eisenberg, M 2009, 'Fabric PCBs, Elec- Zoran, A and Paradiso, JA 2013 'FreeD: A Freehand Digital
tronic Sequins, and Socket Buttons: Techniques for Sculpting Tool', Proceedings of the SIGCHI Conference
e-Textile Craft', Personal Ubiquitous Comput., 13(2), on Human Factors in Computing Systems, New York,
pp. 133-150 NY, USA, pp. 2613-2616
Buechley, L and Perner-Wilson, H 2012, 'Crafting technol- [1] http://ignca.nic.in/UNESCO_Report.pdf
ogy: Reimagining the processes, materials, and cul- [2] http://www.ncctt.org/50years/index.php/110-mas-
tures of electronics', ACM Transactions on Computer- stuff/17888-lewicito-cito-velasquez
Human Interaction (TOCHI), 19(3), p. 21 [3] https://www.youtube.com/watch?v=JHjIpIeyNSc
Crowley, DJ 1956, 'The Traditional Masques of Carnival', [4] https://www.youtube.com/watch?v=EhrQN5EViFo
Caribbean Quarterly, 4(3/4), pp. 194-223 [5] https://www.youtube.com/watch?v=i9YmJSF3Fpk
[6] https://www.youtube.com/watch?v=sRSNvb-WccI

FlexiMold: Teaching Numeric Control through a Hybrid
Device
Serdar Aşut1 , Winfried Meijer2
1
Yaşar University 2 Delft University of Technology
1
serdar.asut@yasar.edu.tr 2 winfried.meijer@tudelft.nl
This proceeding presents the process and outputs of a collaborative workshop

which was held between Yaşar University in Turkey and Delft University of
Technology in the Netherlands. The aim of the workshop was to observe the
educational potentials of a custom-made formwork device towards teaching
CAD/CAM to architecture students. This flexible formwork, which we call
FlexiMold, is a hybrid device that is used manually by following computerized
numeric information. The students designed an architectural object which has a
complex shape and used this formwork to fabricate it in actual scale. We present
the workshop objectives, process and outcomes in this proceeding.
Keywords: CAAD Education, Human-Numeric Control, Flexible Formwork,

Double-Curved Surfaces
INTRODUCTION which is operated manually by following the data

The aim of this proceeding is to present the experi- sets which are derived from a digital model. The over-
ence and outputs of a collaborative workshop which all aim of the workshop was to explore the benefits
was held between the Faculty of Architecture at Yaşar and drawbacks of such a hybrid approach towards
University (YU) in Turkey and the Faculty of Archi- teaching CAD/CAM to architecture students. We pro-
tecture and Built Environment at Delft University of pose that hybrid media, which enable and encourage
Technology (TUDelft) in the Netherlands. The work- the students to engage with the design and fabrica-
shop was held in April 2015 with the participation of tion information both cognitively and bodily, enable
the bachelor students of the Architecture and Inte- a more efficient learning environment. The work-
rior Architecture and Environmental Design depart- shop had provided us a set-up in which we could em-
ments of YU and the master's students of the Design pirically observe and analyse the validity of this pro-
Informatics chair of TUDelft. posal.
The content of the workshop was focused on In addition to introducing skills and know-how
teaching the use of Computer Aided Design and within CAD/CAM, the workshop aimed at familiar-
Manufacturing (CAD/CAM) techniques for designing izing the students with the collaborative aspects of
and fabricating architectural objects which consist of design and fabrication as well as assisting them to-
double-curved surfaces. To this end, we have de- wards developing insight on the tight relationships
veloped a custom made flexible formwork device between fabrication techniques and design thinking.
which we call FlexiMold. FlexiMold is a hybrid device The workshop setup was structured towards achiev-

ing these objectives. The first phase of the workshop numeric information and manual labour. Our hypo-
was based on distant collaboration in which the stu- thetical statement of Human-Numeric Control (HNC)
dents of both universities worked in parallel; while is rooted on this hybrid situation.
the students of TUDelft were developing the design,
the students of YU were developing the fabrication Figure 1
system and examining its potentials and constraints FlexiMold Flexible
in order to communicate with the design team. Af- Formwork Device.
ter the design proposal is completed, all of the stu-
dents have gathered in Izmir for a five days' hands-
on workshop in order to fabricate the design object.
The workshop was funded by the Scientific Research
Project Fund of Yaşar University, TUDelft Design In-
formatics Chair and Polkima Plastic Solutions.
THE DEVICE
FlexiMold is a flexible formwork device which was de-
veloped at TUDelft Design Informatics master's pro-
gram in the previous semesters at the elective course HUMAN-NUMERIC CONTROL
that is titled Technoledge. It is developed by upgrad- Cross claims that design has its own distinct 'things
ing its precedents such as the Zero Waste Formwork to know, ways of knowing them, and ways of find-
(Oesterle et al. 2012), Flexible Mould for Precast Con- ing out about them' which remain largely tacit (Cross
crete Elements (Raun et al. 2010) and Formwork Ta- 1982). This indicates one of the critical challenges
ble (Spuybroek 2004). faced in teaching CAD to design students, as CAD
FlexiMold is constructed on a 70*70 cm wooden mostly requires to communicate explicit information
base. It has 49 steel rods, each of which can be in- which do not mostly overlap with the implicit realms
dependently moved up and down manually. These of design knowledge. Therefore, we need to develop
rods are placed in the base on a grid of 10*10 cm. educational tools and methods which enable the stu-
The formwork is shaped by adjusting all of the rods, dents to experience the digital workflows of CAD
and then, placing a flexible polyethylene sheet on practices through implicit, tacit and intuitive means.
top of them. The sheet is able to perform -almost- Latour and Yaneva (2008) claims that an archi-
any double-curved surface thanks to the grid pattern tect has to be equipped with diverse tools which are
which was CNC milled on both sides of the sheet (see the aids of imagination and instruments of thinking
Figure 1). tied to the body. This concept is rooted on the phe-
The user receives a data set from the digital nomenological perspective of Merleau-Ponty (1962),
model for shaping the formwork in order to cast each who have defined the concept of embodied percep-
surface. The data set consists of 49 numbers, each tion towards illustrating the instruments which are
of which refers to the position of each single rod. the extensions of one's bodily range and the bod-
The user adjusts all of the rods by following this in- ily synthesis. Within our hypothesis, FlexiMold is an
formation. Once all the rods are set in correct posi- instrument of thinking tied to the body which sup-
tions and the polyethylene sheet is firmly placed on ports imagination and it performs as the extension
the rods, the formwork is ready for casting the mate- of the student's bodily range and synthesis towards
rial. Therefore, we propose that this device is a hybrid enabling an embodied perception of the digital in-
medium, which is operated using both computerized formation.

Norman (1994) introduces the concept of soft tech- ates. Also, it helps them to directly observe the rela-
nology in order to refer to compliant and yielding tions between numeric data and form. We even had
systems that acknowledge the initiative and flexibil- the chance to observe that, they were most of the
ity of the person; whereas hard technology refers to time able to fix the problems on the formwork itself
those systems that put technology first, with inflex- when there is a mistake related with the datasheet
ible, hard and rigid requirements for the human. In which is received from the digital model. And all
this sense, FlexiMold addresses the concept of soft of these are performed through intuitive and simple
technology as it is a compliant and yielding system acts.
which allows initiatives and flexibility. Instead of con- This mode of intuitive interaction with the de-
straining the human with rigid requirements, it inte- vice corresponds with certain fundamental notions
grates with him/her while sustaining his/her intuitive of Human-Computer Interaction (HCI) research even
skills. This is where our statement of HNC is rooted though our study is not directly related with HCI.
on. Unlike a complete Computer Numeric Control Namely, the theory of affordances and the synchro-
(CNC) process with which the human cannot inter- nization of action and perception are key to under-
fere, the human undertakes a core role in an HNC pro- stand the potentials that rise on the interaction be-
cess. The user of the FlexiMold receives data from tween the human and FlexiMold. These two are
the computer, understands and interprets it, then op- key concepts in Implicit Human-Computer Interac-
erates the device manually in order to complete the tion (iHCI) studies, which address the so-called third-
process (see Figure 2). Therefore, it is a hybrid situa- wave in HCI research. Gibsons's (1977) theory of
tion between CNC and manual machining. affordances was introduced into design by Norman
(1990) in order to address the relationship between a
Figure 2 physical object and a person -or for that matter, any
Human-Numeric interacting agent, whether animal or human, or even
Control (Photo machines and robots-. Just like argued by him, Flex-
credit: Yaman Umut iMold affords its user towards determining how it is
Bilir). possibly used, and therefore provides a steep learn-
ing curve. On the other hand, FlexiMold enables the
natural synchronization of the perception and action
spaces, which is argued as being a factor that enables
us to perform complex tasks by Sutphen et al (2000).
Eventually, FlexiMold is an HNC medium which en-
ables direct and organic interaction with numeric in-
formation rather than through consecutive lines of
The bodily and cognitive interaction of the human text or, referring to Ishii and Ullmer (1997), painted
with the system in this hybrid situation provides po- bits. It introduces the qualities of traditional and typi-
tentials for reframing the fabrication process towards cal craftsmanship into our relationships with numeric
capturing the tacit dimensions of design knowledge. information.
By this means, it can lead to several experiments
even where fabrication becomes a creative perfor- WORKSHOP SETUP
mance. In our particular case, the educational po- The workshop was applied as a part of regular
tentials of this performance were the most signifi- semester courses in both universities. The partici-
cant ones. First of all, it enables the students to fully pants from TUDelft were the 32 students which were
comprehend how a CNC machining process oper- enrolled in the Technoledge course which is offered

at the first year of the Design Informatics master's tailing the proposal and preparing it for the fabrica-
program. This elective course which runs for 10 tion for 4 weeks. Eventually, they developed a de-
weeks focuses on the design and fabrication of a non- sign in the form of a mobius stripe which consists
standard architectural object. The course requires of 59 unique panels that has double-curved surfaces
the students to work in teams by using CAD applica- (see Figure 3). The object was approximately 2.30
tions. Pre-rationalization of the design is one of the in height and 4 meters in diameter. The panels had
core learning objectives of the course. The partici- hexagonal shapes. The size of each hexagon was lim-
pants from YU were the 17 students who enrolled in ited in a boundary of 60*60 cm due to the dimensions
the CAD/CAM and Rapid Prototyping course which of the FlexiMold.
is offered by the Faculty of Architecture for the third
Figure 3
and fourth year bachelor students of the Architec-
The design model.
ture and the Interior Architecture and Environmen-
tal Design departments. The objective of this elective
course is to introduce the students the uses of CAD/-
CAM techniques in spatial applications.
After completing the design, all 49 students have
gathered in Izmir for 5 days for the hands-on fabrica-
tion phase. The workshop was coordinated by Ser-
dar Aşut and Winfried Meijer and it was applied with
the supports of Peter Eigenraam and Thijs Welman
from TUDelft, Bilge Göktoğan from YU and Mark Gi-
raud from Polkima Plastic Solutions. The infrastruc- The course of the YU students started only after the
ture which was necessary for the fabrication was pro- TUDelft students completed the design proposals
vided by yumak (Yaşar University MakerLab). and selected the one to be fabricated due to the
schedule differences between the two universities.
DESIGN AND PRE-RATIONALIZATION At this stage, the YU students were asked to produce
The students of TUDelft were given the task to design 3 FlexiMolds, understand how to use these devices
an architectural object which has a complex shape, as flexible formwork, test its opportunities and limi-
can self-stand and provides spatial qualities while tations for the fabrication and guide the design team
considering the qualities and constraints of Flexi- accordingly. The students worked with the form-
Mold and fiberglass. The design brief challenged the works for 4 weeks in order to determine the most fea-
students to model a complex shape by using para- sible techniques for outputting the datasets from the
metric tools. Also, they were asked to consider the parametric model and shaping the formwork using
qualities of fiberglass and the limitations of the form- them. They communicated these decisions with the
work (such as its dimensions or its capabilities to design team and prepared manuals to be used on the
bend) starting with the early phases of the design site during the fabrication.
process and pre-rationalize the design decisions ac- The students of TUDelft have developed the
cordingly. At the first phase, the students worked parametric model using Grasshopper and Rhino 3D.
as 8 teams for 2 weeks and presented a draft pro- The model served as an assistant to be used for gen-
posal. One of these proposals was selected to be de- erating variations of the design and the necessary in-
tailed further. The selection was based on the aes- formation for the fabrication. Eventually, the model
thetic qualities of the object and its producibility in was able to output charts which contain the numeric
this workshop. Then, the students worked on de- data sets and CAD drawings with 2D and 3D drawings

of the panels. correspond to the positions of each adjustable rod
of the formwork. Also, it produces flanges which en-
FABRICATION close these panels and output drawings which repre-
The students of both universities gathered in Izmir in sent the shape of these flanges.
order to fabricate the design object at the campus of The panel surfaces were made by using fiber-
YU. They formed three mixed teams each of which glass. The reason for using fiberglass was that it can
worked on a separate FlexiMold. So, each team was perform well with curved surfaces, it is lightweight
responsible for producing approximately 20 panels and that it is fairly easy to learn how to cast it. The
and finishing the assembly in five days. flanges were produced by laser cutting mdf boards
which has 3 mm thickness. The flanges, which are
Figure 4 cut and assembled in the form of frames, define the
A sample chart boundary of each panel. Also, they provide surfaces
containing the for connecting each panel with its neighboring pan-
numeric data set. els. Moreover, they perform as a structural skele-
ton for the whole object when assembled. These
mdf flanges provided proper stability to the structure
while they were able to perform slight bends when
necessary during the assembly.
Figure 5
Students
positioning the rods
on the formwork.
The panel fabrication process was simply done

in three sequential steps such as; adjusting the form-
work, casting the fiberglass and finishing the panel.
In order to adjust the formwork, the students used
the charts with numeric data sets for positioning the
49 rods manually (see Figure 5). After all of the rods
The information regarding the fabrication process were positioned precisely, they firmly placed the flex-
was derived from the parametric model which was ible polyethylene sheet on top of them. Then, they
completed using Grasshopper before the fabrication positioned the flange frame considering the refer-
week. This model first analyses the surface geome- ence points. After that, they started casting 3 lay-
try of the whole object, then divides it into unique ers of fiberglass mat inside the frame and applied the
hexagonal panels. It analyses the shape of each panel resin on it. After the panel was cured and removed
surface and generates output in the form of charts from the formwork, a group of students started to
(see Figure 4) which contain numeric data sets that work on panel finishing (such as sanding, smoothing

Figure 6
The steps for
preparing a panel
on FlexiMold.
and polishing), while another group starts to adjust boundary of the formwork surface. Hence, it was pos-
the formwork for a new panel. By this means, the sible to calculate the positions of all 49 rods. Then,
students were able to produce one panel in down to a corner of the panel was matched with a reference
45 minutes (see Figure 6). Finally, they assembled all point on the formwork surface so that the panel fits
of the panels by using bolts on the flanges and built on the surface with correct location and orientation
the whole structure. Despite the complex shape of (see Figure 7).
the object and its panels, only moderate force was
needed to make all the panels fit together precisely. CONCLUSION
The workshop was concluded with a successfully
Figure 7
completed product (see Figure 8). On the other hand,
The panel being
its process was more important for us than the ob-
cured on the
ject itself. After this first collaborative teaching expe-
formwork.
rience which we practiced together, it is fair to claim
that the workshop process within the learning ob-
jectives was way more efficient than we expected it
to be. We were able to observe that certain learn-
ing outcomes were achieved during and at the end
of the workshop process. Also, it was able to provide
us input towards improving our strategies in teach-
ing CAD/CAM.
One of the most important difficulties which was The broadest thing is that design is highly related
faced during the fabrication was caused by the mis- with making; therefore, such hands-on experiences
match between the shape of the hexagonal panels in which the students are invited to build architec-
and the square shape of the formwork. Each panel tural objects in actual scale need to take place more
had to fit on a different group of rods. However, it common in architecture education. They encourage
was necessary to calculate the position of all 49 rods personal involvement for the students -which is a
in order to achieve the curve continuities on the sur- key aspect in learning. This experience has provided
face. This difficulty was overcome by the paramet- them opportunities to comprehend the strong rela-
ric model. For each panel, the model created ghost tionships between design thinking and the fabrica-
extensions to the surfaces so that the boundary of tion and material systems.
its projection onto the formwork was as same as the Another learning outcome is that the students
had the chance to understand the complex organiza-

tion of the labour which is necessary for both design- hybrid device in the form of an HNC system. Even
ing and building. Even our scale was much smaller the students with the least knowledge on CAD/CAM
than an actual building design and construction, we were able to understand the fundamentals of nu-
tried to introduce as much complexity as possible re- meric control easily by the help of this hybrid sys-
garding the design and fabrication management. We tem. After a quick warm-up process, all of the stu-
asked the students to form different groups to prac- dents were able to develop an understanding on the
tice different tasks and to communicate with each relationships between data and form seamlessly. Be-
other all along the process. We also tried to re-mix cause, such a hybrid device minimizes the need for
these groups for each time as much as we could so explicit instructions as being an instrument of think-
that each of them had the chance to work with dif- ing tied to the body. It enables an embodied percep-
ferent people on each phase of the project. Even- tion of the fabrication information by introducing in-
tually, they practised close and distant collaboration tuitiveness and bodily interactions.
to solve a variety of problems with different people Besides being a good tool for learning, this hy-
each of the time. brid fabrication device performed as an efficient
The particular learning outcome of this work- communication interface between the students.
shop was that the students had the chance to fully Considering that the group was fairly diverse, it was
comprehend how a CNC machining process works necessary to encourage the participation of all in
by integrating their own selves bodily and cogni- each phase of the process. In our case it simpli-
tively into the process. This is achieved through a fied the communication between the students who
Figure 8
The built object.

come from different countries, have different educa- REFERENCES
tional backgrounds and focus (from structural engi- Cross, N 1982, 'Designerly Ways of Knowing', Design stud-
neering to interior design), have different skills (from ies, 3(4), pp. 221-227
Grasshopper experts to the ones who have met it for Gibson, JJ 1977, 'The Theory of Affordances', in Shaw, R
and Bransford, J (eds) 1977, Perceiving, Acting, and
the first time in this workshop) are studying in differ-
Knowing: Toward an Ecological Psychology, Lawrence
ent levels (bachelors and masters') and have different Erlbaum, pp. 67-82
language skills; and enabled them to easily collabo- Ishii, H and Ullmer, B 1997 'Tangible Bits: Towards Seam-
rate on complex tasks. less Interfaces between People, Bits and Atoms', Pro-
We have applied two more follow-up workshops ceedings of the ACM SIGCHI Conference on Human
in the following semesters with new students. Each Factors in Computing Systems, pp. 234-241
Latour, B and Yaneva, A 2008, 'Give me a Gun and I will
of these workshops focused on developing a certain
Make All Buildings Move: An ANT’s View of Archi-
aspect of our first experience. For example, the main tecture', in Geiser, R (eds) 2008, Explorations in Archi-
issue which we have tried to elaborate in the second tecture: Teaching, Design, Research, Birkhäuser Archi-
workshop was the diversity of the students' profiles tecture, pp. 80-89
and the organization of design and fabrication man- Merleau-Ponty, M 1962, Phenomenology of Perception,
agement between them. The third workshop, which Routledge
Norman, D 1990, The Design of Everyday Things, Basic
was applied by not flexible but stable formworks, fo-
Books
cused on developing the distant collaboration phase. Norman, D 1994, Things That Make Us Smart: Defending
We are planning to present these studies in future Human Attributes In The Age Of The Machine, Basic
proceedings. Books
Oesterle, S, Vansteenkiste, A and Mirjan, A 2012 'Zero
Waste Free-Form Formwork', Proceedings of the Sec-
ACKNOWLEDGEMENTS ond International Conference on Flexible Formwork,
We would like to thank Yaşar University for fund- pp. 258-267
ing the project through Scientific Research Project Raun, C, Kristensen, MK and Kirkegaard, PH 2010 'Flex-
Fund; Prof. Dr. İ. Sevil SARIYILDIZ for supervising ible Mould for Precast Concrete Elements', Proceed-
and enabling the collaboration and Mark Giraud from ings of the International Association for Shell and Spa-
tial Structures (IASS) Symposium, pp. 36-51
Polkima for providing us the workshop materials and
Spuybroek, L 2004, NOX: Machining Architecture, Thames
the hands on know-how. & Hudson
Sutphen, S, E, Sharlin, B, Watson and Frazer, J 2000 'Re-
viving a Tangible Interface Affording 3D Spatial In-
teraction', Proceedings of the 11th Western Canadian
Computer Graphics Symposium, pp. 155-166

Fabricating Semi Predictable Surfaces
A workshop series on digitally fabricating freeform surfaces with
aggregates
Joachim B. Kieferle1 , George Katodrytis2

1
Hochschule RheinMain 2 American University of Sharjah
1
joachim.kieferle@hs-rm.de 2 gkatodrytis@aus.edu
Working with CNC routers and robots is widely used in architecture and
fabrication. Our paper describes ways to use these tools more intuitively yet
accurately and without a profound knowledge. We have developed a workshop
format, in which even inexperienced participants are able to quickly start
working with these tools by shaping the non-rigid material sand. Various
production methods and tools are incorporated such as "manual", "gestural",
"CNC" and "robotic" to create various 3D forms which are captured by methods
like 3D scanning, vacuum forming or glueing.
Keywords: Education, Digital Fabrication, Sand surface, Formwork
INTRODUCTION intuitive, imprecise and unpredictable material like

Digital fabrication with CNC routers and robots over sand, we have generated surprising results and we
the last years has become a common method to have encouraged the participants to experiment fur-
manufacture components for architecture. Nearly ther. The approach is not meant to replace the ex-
all modes of digital production such as milling, saw- isting production methods but to extend the possi-
ing, laser or hot wire cutting operate with both pre- bilities of design and digital fabrication in this field.
dictable and precise results - within production tol- The process involves robotic movement as pattern
erances. However all these methods require a high making, sand-casting and mold making to produce
level of knowledge in modeling and control of the custom-made but also mass-produced panels. The
machines (Kieferle et al. 2008). sand particles friction and behavior as well as robotic
Our research has taken the form of a series of arm movement parameters (depth, width, speed and
international and interdisciplinary workshops that angle) can produce unpredictable and exciting re-
adopts alternative approaches which allow inexpe- sults. The input diagram of movement and output
rienced users to use advanced tools like CNCs and pattern of sand can hardly be completely controlled,
robots. Since the intention was not to teach students which allows for "creative" discrepancies.
working with CNC and robots on a high skill level but
to teach them the methodology beyond fabrication, GOALS
tools as a kind of open interface (Brell-Cokcan and Even though there is abundance of sand on earth and
Braumann 2013). By combining precise tools with an sand casting is an ancient form of construction, there

are only few examples of mold making using sand. By shows manual, CNC vacuum and robot fabricated ex-
using sand which collected from the desert, together amples.
with high-tech machinery like routers and robots, our In the first, manual method, participants were re-
goals were to: quired to manually form sand surfaces, experiment-
ing with heaps and carving. This exercise was in-
• establish a relationship between form and
tended to offer a qualitative experience in sand be-
production method
havior, discover approximate slope angles and exper-
• enable users with a low knowledge level of
iment with different shaping tools. The results clearly
digital fabrication to produce free form sur-
showed, that due to the imprecision of manual free
faces
form shaping only limited slopes and angles could
• develop a low cost method for a free form con-
be achieved except when drilling small holes. Due to
crete formwork of precast concrete volumes
the direct control and feedback while shaping with
and panels
hands the overlapping of repeated shapes was not
achieved. This manual method of hand drawing on
WORKSHOP SETUP the sand was inspiring as the first conceptual "sketch"
Our workshop series consisted of a 3-day intense and study for the following more complex digital fab-
work of 15-20 participants of various technical knowl- rication methods.
edge backgrounds. Participants were asked to re- Figure 1
search with a 'hands on' approach and find out how Three typical results
different techniques influenced the shaping of sand sand and vacuum
surfaces. They did so by using boxes of 40×60×12 cm formed snapshot of
filled with sand. The limiting factor of the box size manual (top), CNC
was the weight of the sand but this encouraged more vacuum (middle)
creative patterns. and robot (bottom)
production.
MANUAL, CNC AND ROBOTIC PRODUC-
TION
Three production methods were used for the fabrica-
tion process:
• manual
• 3-axis CNC with attached vacuum nozzle
• 6-axis robot with a simple tool head
As participants with different digital skill levels partic-
ipated in the workshops, the setting was designed in
a way that everybody could work with the machines
and easily transfer their data. As only sand was used
with simple tools, a vacuum nozzle attached to the
CNC and a wood head attached to the robot arm, the
limitations for the robot arm movement was defined In the second, 3-axis CNC method, a more mechan-
by the depth and speed of movement. Typical results ical input was developed. A gray scale image with
of the sand surfaces and cast in the vacuum formed the same size as the sand surface was used to control
plastic are shown in Figure 1, from top to bottom it

the vertical (z direction) movement of the 3- axis CNC The overlaid zigzag pattern with a spacing of 5 mm
head. Brightness and darkness values of the image controlled X and Y movement. Instead of a router,
were mapped to Z-values of the tool path and thus a a vacuum nozzle was attached to the 3-axis CNC to
vertical movement between plus 1 and minus 4 cm. suck out the sand at the position of the nozzle. This
subtractive method showed results quite similar to
Figure 2 standard CNC milling except for the falling sand par-
CNC with attached ticles in areas of steep slopes. The results were quite
vacuum nozzle. predictable (Figure 2).
In the third, 6-axis robot method (Figure 3), the
main control was a 3D spline representing the tool
head position. This spline could be drawn in any 3D
CAD software. The intention was not to generate the
path in an algorithmic way but to allow students to
freely "draw" shapes to overcome automated repe-
tition (Cheng and Hegre 2009). There are very so-
phisticated projects showing how robots can shape
algorithmic paterns with sand [1], however to make
it as easy as possible for students with hardly any ex-
Figure 3 perience controling robots, the setup had to be very
Robot carving sand simple. This is why we only asked them to draw a 3D
patterns. spline controling the movement of the tool tip. How-
ever this already required a planning in how often
and how deep the tip might pass by certain positions.
The oblique angle of the tool and thus of the robot
head was controlled by a line from the tool head to
a static reference point, about 50-70 cm above the
sand box surface. This production method showed
various unexpected results because of the overlaying
of sand mass, when the head crossed previously pat-
terned heaps of sand. Unlike in controlled produc-
tion processes like milling, the overlapping move-
ment of the attached head at the robot shuffled the
sand into the previously carved indentation influenc-
ing both the new and the existing carving. Differ-
ent angles of the tool produced different results. The
depth of the tool head in the sand as well as the num-
ber of repeated paths substantially influenced the
generated sand forms. Deeper indentations could
only be achieved by moving the robot arm head sev-
eral times in the sand, continuously carving deeper
into the surface.

MAPPING THE SURFACE FORMS Figure 4
In order to map, record and freeze the sand forms Applying foam to
these surfaces were digitally scanned and physical capture surface.
printed to other materials. Due to the instability of
sand forms a limited number of options was possible.
We experimented with:
• 3D laser scanning
• casting with foam
• vacuum forming
• gluing the sand together with varnish and
other adhesives.
Figure 5
Vacuum forming
3D laser scanning worked well as expected and cre- process of sand
ated precise point clouds of the sand surfaces and the boxes.
plastic surface casts. Experiments with various foams
(e.g. expanding foam and insulation foam) did not
produce any useful results (Figure 4). Already when
applying the foam to the sand surface, the geometry
of the sand particles was altered by the expansion of
the foam. Minor details were flattened instantly.
In contrast, vacuum forming (Figure 5) over the
unstable sand forms quite unexpectedly showed to
be very effective and relatively precise. For the vac- Figure 6
uum forming, the sand box was placed on the vac- Design studio
uum machine table. Sheets of plastic with a thickness project using sand
of 1.5-2mm were heated in the machines oven un- form (student Talin
til they became soft and malleable. Then the sheets Hazer).
were dropped over the sandbox, the vacuum table
moved up and instantly the air was sucked out so
that the plastic covered and copied the whole top
and sides of the sand boxes. After 60-90 seconds
the plastic cooled down and became stable again.
Despite the relatively high forces and movement of
the vacuum, the sand forms were hardly flattened or Tests in gluing the sand at the surfaces together with
pressed and the original sand surface variations were varnish spray showed nice, very detailed but also very
maintained within reasonable tolerances. So for fur- fragile results and the outcome were only useful in
ther production steps like precast concrete elements, small model scale. As a follow up of the workshop, a
this method seemed to be most promising as a form- studio project was done by student Talin Hazer (Fig-
work for medium curved shapes without fine details. ure 6). She ran tests with injecting glue into the sand
And if the one-way plastic could be replaced by e.g. using the robot arm. However the use of scripted
an elastic, reusable material like rubber, hardly any robot movement to draw and spray adhesives and
waste for form work would be produced. glue on the sand surface to make it rigid created

Figure 7 some intriguing 3D enclosures.
Clay panels with
attached and APPLYING THE METHOD TO OTHER MATE-
integrated sensors. RIAL
Based on the experience with modeling forms into
sand surfaces, which then had to be cast to other ma-
terials, we ran tests with the viscous material charac-
teristic and plasticity of clay. The intention was to di-
rectly apply simple tools attached to the robot head
to shape a surface. We used the method of a spline
tool head path carving into clay (Figure 7).
In the Middle East clay is a traditional material
historically used for both its plasticity and thermal
prope ties. We decided to test this material. In our
Figure 8 setup we did not only want to create a shaped sur-
Temperature sensor face in clay but to measure and understand the phys-
stacks in clay box ical properties inside the thickness of the clay panel
installed already, in relation to its surface. This experiment was made
humidity sensor by both architecture and electrical engineering stu-
and outside dents. We fabricated "intelligent clay" panels with
multi-sensor not yet embedded electronics to collect and process various
positioned. data.
Three boxes, same size as the sand boxes, were
prepared. Sensors were positioned inside as well as
on the surface of each box. We were therefore able
to measure the properties inside the clay at various
points as well as the external conditions. To under-
Figure 9 stand the flow of temperature, two stacks of three
Typical graph of temperature sensors at various depth positions were
monitored located inside the clay thickness (Figure 8). Inside the
temperature (here 3 box further humidity sensors were positioned, and at
days curve). the outside of the box a triple sensor measuring tem-
perature, humidity as well as the brightness of light/-
sun was installed. Finally all three boxes were filled
with clay and the surface was leveled.
In order to have a neutral reference surface, one
panel was left untreated; the two others were ma-
chined with various surface patterns. Then the pan-
els were set up and the sensors sent the data every
3 minutes wireless to a raspberry pie mini computer
which was set up indoors. Every hour the new data
for presentation was hosted in Germany. For the vi-
(and if needed previous data too) was copied over the
sualization of the data, the raw data was saved in a
web from the United Arab Emirates to Germany, as
CSV dump file. Then automatically converted to an
the server for the evaluation and the web interface

rrd file and visualized on the project website with the ACKNOWLEDGMENTS
rrdtool [2]. The installation was monitored over sev- We would like like to thank all participating students
eral weeks (Figure 9). from AUS as well as HSRM for their contributions and
As expected the results showed the delay of the work in the workshops. This includes nearly 40 stu-
heat flow through clay: the deep positioned sensor in dents of architecture and electrical engineering. We
the clay panel measured higher temperature fluctua- would also like to thank Prof. Dr. Georg Fries of the
tions. A minor change of the surface pattern variation Electrical Engineering Departmnet at HSRM for his
was observed. Another parameter that should have active and valuable participation in this project.
been taken into account was the shrinking character- Special thanks to the AUS CAAD Labs Director
istic of drying clay. John Michael Dosier as well as his team Jennifer
Theokary and Firass Al Joundi. The workshops would
CONCLUSIONS AND FUTURE WORK not have been possible nor that successful without
Due to its complexity, well-trained and experienced their most valuable input as well as their day and
participants mostly use digital fabrication pedagog- night support.
ically and it takes quite considerable time to pre-
pare and achieve satisfactory results. With our ap- REFERENCES
proach digitally inexperienced researchers, profes- Bechthold, Martin 2007 'Teaching technology: CAD/-
sionals and students can achieve good results within CAM, parametric design and interactivity', Predict-
a span of approximately 3 days, understand the po- ing the Future [25th eCAADe Conference Proceedings
/ ISBN 978-0-9541183-6-5] Frankfurt am Main (Ger-
tentials of digital fabrication and work with complex
many) 26-29 September, pp. 767-775
scripts and machines. By combining manual meth- Brell-Cokcan, Sigrid and Braumann, Johannes 2013 'In-
ods with a well chosen digital production framework dustrial Robots for Design Education: Robots as
that provided the students with a steep curve of suc- Open Interfaces beyond Fabrication', Global Design
cess, it helps them to reach a high level quite quickly and Local Materialization[Proceedings of the 15th In-
since especially in digital design and fabrication "... ternational Conference on Computer Aided Architec-
tural Design Futures / ISBN 978-3-642-38973-3] Shang-
it has been fruitful to build up understanding and
hai, China, July 3-5,, pp. 109-117
skills incrementally" (Bechtold 2007). So the feed- Cheng, Nancy Yen-wen and Hegre, Erik 2009 'Serendip-
back from all participants was absolutely positive. ity and Discovery in a Machine Age: Craft and a CNC
The results of this educational exercise are now the Router', ACADIA 09: reForm( ) - Building a Better To-
basis for further research and application to practice. morrow [Proceedings of the 29th Annual Conference
The feedback from manufacturers about implement- of the Association for Computer Aided Design in Ar-
chitecture (ACADIA) ISBN 978-0-9842705-0-7] Chicago
ing this approach into their production looks promis-
(Illinois) 22-25 October,, pp. 284-286
ing and encouraging. The qualitative experiments Kieferle, Joachim B., Katz, Neil and Thaleck, Kruno 2008
would now require a quantitative approach. The pro- 'From Shape to Fabrication', Architecture in Computro
duced forms and point clouds will have to be ana- [26th eCAADe Conference Proceedings / ISBN 978-0-
lyzed and also compared with the results gained from 9541183-7-2] Antwerpen (Belgium) 17-20 September,
a production with other aggregates. Furthermore we pp. 537-544
[1] http://www.dfab.arch.ethz.ch/web/e/lehre/208.ht
will research into algorithms that will help to predict
ml
the behavior of sand when pushed aside so that the [2] http://www.rrdtool.org
final outcome can be better predicted.

Nip and Tuck: A Simple Approach to Fabricate Double-
Curved Surfaces with 2D Cutting
Negar Kalantar1 , Alireza Borhani2 , Ergun Akleman3
1,2
Department of Architecture, Texas A&M University 3 Department of Visualiza-
tion, Texas A&M University
1,2
www.thetranslab.com 3 www.viz.tamu.edu/faculty/ergun/
1,2,3
{kalantar|borhani|ergun.akleman}@tamu.edu
In this paper, we introduce the Nip and Tuck Method, which provides a general
approach to construct complicated shapes without using high-level software
and/or without solving complex mathematical problems. Our framework is based
on discrete version of Gauss-Bonnet theorem, which states that the sum of vertex
angle defect in a given piecewise planar manifold or manifold with boundary
mesh surface is independent of the number of vertices, faces and edges. Based on
this property, architects and designers can simply introduce negative and positive
curvatures in the places they want to obtain desired shapes. We presented Nip
and Tuck Architecture to freshman students in beginning level design studios to
design arches with modular elements along with other methods. Several groups
of students, that chose to use Nip and Tuck approach to obtain individual
modules, were able to design and construct unusual small-scale arches.
Keywords: Nip and Tuck , Double-Curved Surfaces, Surface Active Arches,

Self-Supporting Plywood Structures, Fabrication with Planner Materials,
Freshman Design Studio
INTRODUCTION AND MOTIVATION chitecture continues to be the laser cutter. Conse-

In this paper, we address two challenges that design- quently, for architecture students, moving from 2D to
ers and more specifically, beginning designers deal 3D methods of fabrication often leads to higher man-
with in the context of geometric complexity and find- ufacturing costs and a need to access outside fabrica-
ing form. Despite the widespread development and tion facilities.
adoption of three-dimensional (3D) modeling soft- The other challenge for beginning design stu-
ware that allows for the virtual design of complicated dents and, particularly, foundation architecture (first
shapes, the physical construction of such shapes re- year) learners is how to find forms suitable for their
mains an unsolved problem; currently, it requires ad- designs and determine the criteria necessary for the
vanced machinery capable of milling custom 3D fea- selection of such forms. There is, therefore, a need
tures or 3D printing. However, the most readily avail- to develop form finding approaches to help design
able digital fabrication tool in most schools of ar- students become "form finders" rather than "form

Figure 1
A Self -supporting
structure designed
by Nipping
technique.
Foundation design
students at Texas
A&M University:
Nicholas Houser
and Ezequiel
Campos.
givers." In this paper, we introduce such a form- tise (Sun and Fiume 1996, Mitani and Suzuki 2004).
finding approach that we have termed Nip and Tuck Presently, only a few architectural firms take full ad-
Method. vantage of the currently available graphics and mod-
Nip and Tuck Method is based on construction eling technology to construct more complex forms.
with developable surfaces. These types of surfaces However, in the long term it is reasonable to expect
are made out of thin sheets of flat materials that are that the design of complicated shapes will become a
rolled rather than stretched [1] ; they are widely used common practice, even in small architectural firms. It
for the physical construction of complicated shapes. is, therefore, important to introduce students to such
These flat sheets are excellent for teaching the design design and construction problems at the freshman
of complicated architectural forms because they can level.
easily be cut using laser cutters, one of the most com- One of the most significant challenges to do-
mon digital fabrication tools available to students of ing so, however, is that freshman students do not
architecture. Moreover, using widely available flat have the software expertise and math/physics knowl-
materials such as plywood panels can significantly re- edge necessary to properly engage with these types
duce manufacturing costs. of problems. In this paper, we will present the Nip
The design and construction of large-scale and Tuck method, a process that allows for the intro-
shapes with developable surfaces requires archi- duction of complicated shapes into freshman design
tectural, engineering, and computer science exper- studios. This Nip and Tuck Method permits the con-

trolled curvature of planar panels without the need Our method is based on the introduction of max-
for extensive software expertise and math/physics ima and saddle points in desired locations. Maxima
knowledge. points are obtained by nipping, and saddle points are
Freshmen can design simple planar panels from attained by tucking from an initial piece of paper. To
thin sheets, and cut them using laser cutters; they nip, we remove a piece; to tuck, we add a piece (Fig-
can then turn these planar panels into doubly- ure 2). In this research, for arch designs, we allowed
curved panels and use them to construct compli- students only to nip in order to guarantee convex
cated shapes. One of the most important impacts arches.
of our approach is that it allows students to inter-
Figure 2
nalize geometry through hands-on experiments. An-
Introduction of
other advantage is that this approach will signifi-
maxima and saddle
cantly lower construction costs because the planar
points in desired
panels can be cut from inexpensive materials such as
locations. Initial
plywood.
paper (left), Maxima
As part of our ongoing research into form-
(middle) and Saddle
finding structures, this practical design approach is
(right).
introduced in this paper in an effort to addresses the
following objectives:
• The design and fabrication of complicated Our approach is based on a discrete version of the
shapes using planner materials; Gauss-Bonnet theorem, which states that the sum
• The curvature control of planner panels with- of the vertex angle deflections in a given piecewise
out the need for extensive software expertise, planar manifold (or manifold with boundary) mesh
in order to make non-structural surfaces into surface is independent of the number of vertices,
self-supporting structures; and faces, and edges (Akleman and Chen 2006). Since pa-
• The introduction of strategies for bending per is a developable material, it can be considered a
rigid materials such as plywood, and fixing piecewise linear surface that consists of long strips of
them into their final forms. planes; therefore, our formula can be applied to de-
velopable surfaces such as curved paper.
Figure 3 THEORETICAL PRELIMINARIES

With nipping and In this paper, we introduce an extremely simple ap-
tucking, it is proach to design and construct double curved sur-
possible to use faces with hands-on manipulations of developable
paper to easily materials. Our approach does not require any spe-
obtain a wide cific mathematical or computational knowledge. De-
variety of shapes. signers by using the intuition coming from Gaussian As mentioned, our method comes from the concept
curvature can create designs with nipping and tuck- of discrete Gaussian curvature, which is also called
ing using just paper, glue and stapler (exactly like angle defect. The angle defect, µi , of a vertex i, µi , is
tailors designing clothes with nipping and tucking defined based on the corner angles:where θi,j is the
fabrics). These paper models provide simple proof internal angle of the corner, j, of the vertex i and Ni
of concepts and larger versions of the designs can is the number of corners.
eventually be constructed using thin materials such
as sheet metal and plywood.

of Architecture at Texas A&M University were asked
∑
Ni
to use plywood and design self-standing arches [3].
µi = Θ i ,j (1) The Surface Active Arch design project had two
j=0
stages. In the first, students explored surfaces and
A discrete version of the Gauss-Bonnet theorem geometrically active surfaces by using paper-like ma-
shows that the sum of the vertex angle defects of all terials. Paper is an appropriate material for exploring
of the vertices of a planar mesh is equal to 2(2 − 2g) developable surfaces because it does not stretch and
where g is the genus of the surface (i.e., the total num- is easy to cut and form (Figure 4).
ber of holes and handles), and to (2 − 2g) where In the second stage, students were asked to re-
g is the Euler Characteristic of the surface. This re- make their concepts out of 1/8" thick plywood, and
sult agrees with the Gauss-Bonnet theorem (Chavel explore active uses of the form through the relation-
1994), which states that the integral of the Gaussian ships among the various geometrical and material
curvature over a closed and smooth surface is equal behaviors. Students' final models were iteratively de-
to 2 times the Euler characteristic of the surface [2]. veloped from small study models, eventually evolv-
Vertex angle defect is a very simple concept that ing into various joint and detail models created along
can provide a good intuition sensitivity to control sur- a number of different scales.
face curvature. Note that the value of µi is related to Using this process, we were able to observe the
the Gaussian curvature. effectiveness of the Nip and Tuck process in finding
forms. Using Nip and Tuck approach, students not
1. If µi > 0, then the vertex, i, is either convex only learned how to find forms, but also learned to
or concave. the effect the materials in obtaining final designs. In
2. If µi = 0, then the vertex, i, is locally planar; this particular case, students learned to differentiate
(i.e., it can be unfolded to a plane without any between two types of materials, namely paper and
gap or self-intersection). wood. The most significant observations made by
3. If µi > 0, then the vertex, i, is a saddle point. students working with two types of materials were
that changing the materials changed the properties
In other words, to create a convex or concave region of final forms. In other words, these forms were
with paper, we can simply take out some angle by not driven simply by geometry, but also by material
nipping (or pinching). To create saddle regions we properties.
can add angles by tucking. Using this intuition it is
easy to play with paper to design developable sur-
faces. IMPLEMENTATION DETAILS
Since we know that the total angle will always In the design studio, we considered three strategies
stay constant, we can shape papers simply using glue for the design and construction of the surfaces of self-
and staples. Using this insight, it is possible to shape supporting arch structures: (1) Folding, (2) Bending
planner materials intuitively and construct a variety and weaving, and (3) Nipping (Figure 5). Unlike pa-
of forms (Figure 3). per, plywood is not easy to bend or fold; therefore, for
bending the plywood, we used three types of meth-
ods (1) Making relief cuts, (2) Cutting a series of kerfs,
METHODOLOGY and (3) Soaking plywood and reforming it (Figure 6
To demonstrate the efficiency and simplicity of Nip and Figure 7).
and Tuck Method, we used the method for designing In all of these techniques, volume is created
Surface Active Arches in a freshman studio. For ex- somewhere between the parts. The convex forms
ploring the notion of three-dimensional structures, that result from applying these strategies can be used
students in freshman design studios in Department

Figure 4
A journey from
small paper models
to large plywood
arches.
Figure 5
Bending and
Folding were two
other strategies
that students used
to design
self-supporting
arches. Left project:
Madison Wilson
and Alexis Flores,
Right project:
Austin Madrigale,
Laura Zipps and
Madison Larmeu.

not only to cover space, but also as load-bearing Figure 6
building elements. Students used all of these strate- Left: different relief
gies to increase the stiffness of a thin layer of ply- cuts, Right project:
wood, thus converting the wood into surface-active Michael Hergert
structures. What is interesting about the Nip and and Sugey Zavala.
Tuck strategy is that the material and its distortion are
the determining factors in the form-generation pro-
cess (Buri and Weinand 2011).
Six groups of students chose to use Nip and Tuck
Method to design their arches. Since arch struc-
tures can only have positive curvature, for design-
ing arches, only nipping approach is sufficient. When Figure 7
nipping, students had to pay attention to the direc- Cutting a series of
tion of the inner layers of the plywood. By cutting kerfs and soaking
out some sections and taking out certain angles, stu- plywood to reform
dents were able to make convex or concave designs. it.
However, if the angle was large, the nipping process
needed to be conducted in progressive stages. Al-
though through soaking and steaming the plywood
surface became flexible enough to be squeezed into
the final angle, reaching that final angle was a grad-
ual progression. After getting the surface into the fi-
nal form, it needed to be fixed into position. Fixing
together two parts of a cut was achieved through var-
ious different methods such as screws and nuts, zip
ties, and rivets. To push the connections further and
experiment with the material, there was also the op-
tion of using methods such as intertwining or cut-
and-tenon (Figure 8).
CASE STUDIES
Here, to exemplify the practical possibilities of nip-
ping, we have included several self-standing arch
projects built from thin plywood modules. Below,
we discuss how the different modules can be made
and how they might be connected together to form Figure 8
arches. In these examples, the main focus is on how Different methods
to create complex, double-curved modules out of de- of fixing parts
velopable surfaces, as well as how to find methods together.
that simplify the process of forming and bending ply-
wood.

On the surface, the constructed arches were treated interconnected. As the number of triangular cuts in-
as 2D domain structures. The design process was creased, the degree of bending and twisting at each
conducted in three stages. In the first stage, stu- cut area was enhanced accordingly. Longer cuts re-
dent proposed their desired single- or double-layer quired less torsional force. In other words, the nar-
arch structure. Once the arch was roughly designed, rower the piece, the more twistable the entire ply-
the central challenge of creating the structural mod- wood surface. In nipping, the ability of a sheet of
ules presented itself. During the second phase, the plywood to absorb the impact of torsional forces is
proposed arches offered a convenient geometric ba- limited by the size of the panel, its core thickness
sis for using nipping to design the different curved and density, and the length, width, and proportion
modules that would eventually become the arch's of the triangular cuts. Different groups of students at-
building blocks. In the third phase, the stability of tempted to determine and apply the optimum com-
the proposed arch was revised, based on the three- bination of these parameters in order to manipulate
dimensional geometrical arrangement of the mod- the bending stiffness of the material. A wide variety
ules. of 3D shapes could be achieved once the students
When a plywood surface is bent in one or two di- regularly or irregularly cut the different parts of their
rections, its stiffness and load capacity increase. For sheets.
this project, the depth of the bent surface became
Figure 9
a structural variable (Figure 9). The nipping method
Different models
enabled student to bend their piece of plywood in
were made to find
both directions, in order to achieve a positive Gaus-
the right curvature
sian curvature. In this case, the local behavior of the
and module.
plywood was changed by vertex deflection. To make
the basic module for the arch, three types of curved
surfaces were considered:
1. Neutral single curve (cylindrical);

2. Negative double curve (hyperboloid); and
3. Positive double curve (spherical).
Students made the double-curved modules by phys- Students cut 2D pieces while leveraging 3D models
ically constructing developable surfaces. Since nega- as masters, guaranteeing the accuracy and quality of
tive double-curved surfaces are difficult to construct their designs. Flattening and then laying the digital
from plywood, the students mainly investigated pos- models of their 3D surfaces onto their 2D materials
itive curved surfaces in their arch modules. helped students estimate the best connection holes
Although a number of different ways to make and angle cuts for the edges.
positive double curvatures were presented, the stu- Continuous arch structures were achieved by
dents typically began with a rectangular sheet. Based overlapping the surfaces of two neighboring mod-
on the vector-drawing files produced through the ules. Most of the arches did not follow a variable cur-
Rhinoceros platform, the students removed certain vature. Consequently, the convergence angles of the
angles as triangular pieces, and then joined the edges were, in some ways, alike.
neighboring edges of the angles together to create By fabricating three-dimensional structural sur-
convex regions. By turning the flat surface of their faces with strict two-dimensional cutting, students
plywood into a series of smaller cut pieces, the uncut were able to quickly grasp the possibilities and limits
surfaces were easily bent or twisted and the planes

Figure 10
Different arches
designed by
nipping methods.
Left project: Lorena
Espinosa and Julia
Grasley, Right
project: Natalie
Akins and Thomas
Holmes.
Figure 11
Different arches
designed by
nipping methods.
Left project: Chesli
Nicole, Kathleen
Berger and Loryn
Taylor, Middle
project: Anna Cook
and Cynthia Castro,
Right project: Elise
Werner and Manuel
Duran.

of their proposed designs (Figure 10 and figure 11). try, but also that the fabricated geometries stemmed
The speed, efficiency, and low cost of prototyping al- from the materials, tectonics, and assembly logic.
lowed the students to improve and fine tune their Therefore, exploring the use of nipping and tucking
structural modules in order to accommodate specific with other materials will even more tightly couple fu-
requirements; this opportunity for revision was avail- ture design explorations with materials-based proto-
able until a very late stage of the design and fabrica- typing.
tion process. Under extreme time constraints and without a
thorough knowledge of digital modeling software,
CONCLUSION AND DISCUSSION freshman students were challenged to explore the
The Nip and Tuck Method offers great freedom for geometric complexities of their designs and trans-
structural curved forms, allowing them to be fabri- form their flat sheets of material into intricate three-
cated from two-dimensional sheet material and laser dimensional convex surfaces; the goal was to cre-
or 3-axis cutting machines. In addition, these sim- ate self-standing structures with improved mechan-
ple methods offer enormous benefits in terms of fab- ical properties that were also light in weight. The
rication, modularity, and customization. To date, arch project sequence emphasized the mutual inter-
little published research has been presented that actions among physical prototyping and simple dig-
describes the structural and mechanical properties ital modeling, weaving together precision, jointure,
of the configuration formed through the nipping detail, connection, and assembly. Armed with the
method. In the future, designers and scholars should empirical knowledge provided by the project, stu-
investigate the effects of its geometry on the stability dents enjoyed a wide playing field of possibilities for
and performance measures of convex and concave designing complex forms that otherwise would only
surfaces, and discuss the advantages to and limita- have been possible through the use of advanced dig-
tions of using nipping to form flat materials. Across ital modeling. Exploring the nipping technique and
a wide array of methods that can be used to trans- its utility in addressing developable surfaces within
form a flat sheet of material into a three-dimensional the bounds imposed by certain material properties
form (such as bending and folding), it seems that nip- helped students to generate complex surfaces and
ping is uniquely suited to forming plywood shapes; deliver robust 3D design capabilities.
it is one of the most efficient shape and structure- Exploring developable surfaces within the con-
forming processes. Combining nipping with other text of the design studio provided a framework for
methods enables the generation of different struc- developing a new in-house open source software
tural forms with a myriad of desirable characteristics. called UNFOLDING to address the constructability
Beyond allowing complex surfaces to be con- and structural geometry of complex forms. This soft-
structed out of regular laser cutters, the nipping ware demonstrates the economic efficacy of using
method also offers an additional benefit; it amplifies minimal amounts of isotropic materials that can only
the stiffness of a surface without adding extra ma- be bent in one direction to construct sculptural and
terial. By exploiting the deformation of the materi- architectural forms of different scales. After being de-
als, the nipping approach generates bending-active signed, the software can be used to unfold a desired
structures and causes more stiffness. form so that it can then be laser cut. Giving future
In this study, students explored nipping ap- designers hands-on experience with modeling and
proach as a form-finding technique by using ply- laser cutting will familiarize them with simple meth-
wood; however, it was possible to apply this method ods of fabrication, thus overcoming some of the cur-
to other materials such as metal or plastic sheets. The rent limitations with exiting CAD software and fabri-
experiences described here illustrate that the final cation techniques.
arch structures were not only informed by geome-

REFERENCES
Akleman, E and Chen, J 2006 'Practical polygonal mesh
modeling with discrete Gaussian-Bonnet theorem',
Proceedings of Geometry, Modeling and Processing,
Pittsburg
Buri, H and Weinand, Y 2011, 'The tectonics of timber ar-
chitecture in the digital age', in Kaufmann, H (eds)
2011, Building with Timber Paths into the Future, Pres-
tel Verlag, Munich, Germany, pp. 56-63
Chavel, I 1994, Riemannian Geometry: A Modern Introduc-
tion, Cambridge University Press, Cambridge
Kalantar, N and Borhani, A 2015, 'Flexible Textile Struc-
tures: An Agency for Informing Form and Matter',
International Journal of Interior Architecture + Spatial
Design, ii, pp. 50-55
Mitani, J and Suzuki, H 2004, 'Making paper craft toys
from meshes using strip-based approximate unfold-
ing. ACM Trans. Graph', ACM Transactions on Graph-
ics (TOG), 23, pp. 259-263
Sun, M and Fiume, E 1996 'A technique for constructing
developable surfaces', Proceedings of the conference
on Graphics interface, Ontario, Canada, pp. 176-185
[1] http://mathworld.wolfram.com/DevelopableSurface
.html
[2] http://mathworld.wolfram.com/Gauss-BonnetFormul
a.html
[3] https://www.youtube.com/watch?v=49SR0qsKeeA

Digital Materialization: Additive and Robotical
Manufacturing with Clay and Silicone
Cristina Nan1 , Charlie Patterson2 , Remo Pedreschi3
1,2,3
University of Edinburgh
1,3
{cristina.nan|R.Pedreschi}@ed.ac.uk
2
cpatters@exseed.ed.ac.uk
Through the use of algorithmic design methods and an ever growing variety of
digital fabrication tools the complexity of process in the architectural discipline
seems to be increasing. As this statement might apply to a variety of different
areas of computational design and process management, this perceived growing
complexity does not have to be viewed as unnecessary complication of design
processes, if palpable and justifiable benefits occur. This paper intends to analyse
and investigate the potential arising from digital tools of fabrication, specifically
robots and 3D printers, and from open source platforms on exploring and
managing complexity while enabling both simplicity of process and simplicity of
implementation through emerging open source cultures. Building on this
assumptions, this paper explores the professional possibilities generated the
implementation of robotics as part of the academic curriculum. The theoretical
concept of Machinecraft will be introduced and showcased on two research
project, both focussing on advanced digital tools, additive manufacturing and
machine engineering. Please write your abstract here by clicking this paragraph.
Keywords: Additive Manufacturing, 3D Printing, Robotics, Digital Fabrication,

Open Source, Architectural Education
INTRODUCTION ond investigates the resulting methodological impli-

This paper explores the possibility of developing an cations on two applied research projects, one devel-
awareness and understanding of the new and com- oped at the Edinburgh School of Architecture and
bined role robotics and open source platforms oc- Landscape Architecture at the University of Edin-
cupy in the architectural process and in interdisci- burgh and the second at the Institute for Advanced
plinary academic teaching, from the stage of defining Architecture of Catalonia in Barcelona. The attempt
the project's framework to the stage of digital fabrica- to redefine the status of the machine in general,
tion. This issue will be addressed on two levels of dis- and specifically of the robot, seeks to illustrate the
course. While the first level discusses the theoretical- robot as an active design agent, capable of influenc-
philosophical framework behind the architectural in- ing and redefining established processes in architec-
tegration of machines, specifically robots, the sec- ture.Simultaneously the analysis focuses on develop-

ing an understanding of the role of computer aided tect. Through the means of open source platforms
design and digital fabrication in architectural edu- these technical interventions and manipulations are
cation and to showcase a set of approaches of how considerably simplified for the user and offer a di-
to make use of open source platforms and ways in rect way of handling complexity through simplified
which these new territories can be implemented in means. In the second case of highly complex sys-
the academic context. As a conclusion, this paper will tems, the adaptation of machine configuration will
investigate on the relevance of computation, digital be achieved by collaborating with mechanical en-
fabrication and open source platforms for the archi- gineers or the machine developers and giving indi-
tectural discipline and its professional identity and cations to how and why the mechanistic operating
how these enable future architects and researchers mode should be changed according to project spe-
to access other disciplinary fields with the outcome cific needs. This last step can be only done, if a prelim-
of advancing and adding complexity to the own dis- inary examination and understanding of the opera-
cipline. It is of high importance to analyse and in- tional system has been done, so that the implications
vestigate the potential arising from digital tools and which result from conducted changes consort with
open source platforms on both exploring and man- the desired outcome. (Nan 2015.) The concept of ma-
aging complexity, while enabling both simplicity of chinecraft is rooted in the assumption, that just as ar-
process and simplicity of implementation through chitects gradually engage in the customisation of de-
the earlier mentioned open source cultures. sign software through scripting, in the development
of customised materials or material systems (Menges
THEORETICAL FRAMEWORK 2008) in order to increase the optimisation and ef-
In order to redefine the importance and relevance fectiveness of both design and making, the same
of the machine, including self-developed or manipu- should be valid for their engagement with customis-
lated machines as 3D printers or robots, for the whole ing themselves machines.The access to and manipu-
of the design and fabrication processes in the disci- lation of mechanical systems such as robots, CNC ma-
pline of architecture, the author presents the theo- chines, 3D Printers and many others, from a software
retical concept of machinecraft. The concept of ma- and hardware point of view, is today radically sim-
chinecraft intends to combine the two operational plified through the access to open source platforms,
ways which are being represented by its word com- such as Processing, Arduino or even Grasshopper, to
position. Machinecraft describes the process of de- just name a few. These type of machine systems were
veloping a fabrication strategy which satisfies the re- previously considered as highly complex and compli-
quirements of design, material and machine parame- cated so that a direct manipulation in terms of their
ters. It relies on a closed informational circuit in which software-hardware through the architect was highly
design, material and machine information constantly improbable. Throughout the Digital Turn and the ad-
complement and influence each other. Machinecraft vancement of the open source cultures these realities
addresses the ability to extend the area of control have changed. Because of the emerging open source
over the machine, just like a craftsman controls his culture the architectural design but also increasingly
tools, this meaning to be able to manipulate the tech- the fabrication processes and strategies can be seen
nical, mechanical configuration of a machine, in or- as network-driven activities, based on an inherent
der to attune it to material and design requirements. collaborative practice. (Ratti 2015.) The generated
This manipulation can be done in two ways. If in pos- effect consists of making complexity accessible and
session of advanced programming and engineering transforming it into simplicity, in the sense that net-
skills, the change of the machine configuration can works of information or mechanical systems, previ-
be done actively by the user, in this case the archi- ously highly discipline specific, become accessible to

architects and free to manipulate. This leads to over- development of a self built ceramic and silicone 3D
coming design and machine limitations, thus allow- printer and ultimately generated prototypes of cus-
ing the architect new degrees of disciplinary free- tomised medical braces and gloves. Additive man-
dom. This new type of interaction between the archi- ufacturing and textiles can be seen as the main line
tect and the machine can be simultaneously seen as of research which continuously goes through all of
a strong catalyser for future cross disciplinary inter- the further described research phases. The educa-
actions. Open source offers the possibility to man- tional value of integrating advanced digital fabrica-
age and handle complexity in new ways by permit- tion, specifically additive manufacturing, as part of an
ting and enabling active manipulation of systems. academic interdisciplinary curriculum is being show-
cased.
RESEARCH STUDIES Lines of investigation.
The previously described theoretical approaches and Fabric-Cast Concrete and Silicone 3D Printing. Ini-
constructs will be exemplified and demonstrated on tially work was undertaken looking at textile form-
two applied case studies, both focussing on additive work for concrete structures as a viable alternative to
manufacturing and advanced digital tools, but which conventional formwork, commonly fabricated out of
show different scales of implementation. wood. The use of textile formwork represents a ma-
jor advantage when dealing with complex geome-
Figure 1
tries, for instance based on single or double curvature
Fabric Cast
as it does not require any additional and material-
Concrete Model.
intensive scaffolding or moulds. As an output of this
first stage, as series of fabric-cast concrete columns
resulted. This process offered the starting point for
understanding textile material behaviour as a tool for
form control and generation. As a next step a series of
experiments followed which investigated the impact
of 3D printing and patterning directly on textile form-
work and fabric-cast concrete. The material interac-
tion between concrete and 3D printed patterns out
of silicon was analysed through a series of material it-
erations. Although the silicone alters how the fabric
stretches in response to the wet concrete, no major
Research Study 01: From Fabric-Cast Con- material difficulties were encountered in terms of the
crete to Silicone 3D Printing casting process and or removal of the silicone pat-
Context. The following presented research project tern design. Through the interaction between con-
was developed during the MSc Material Practice crete and silicone, an untypical hybrid material sys-
2014/15 programme at the University of Edinburgh. tem, a new type of unexpected material aesthetic
The research showcases an investigation into digi- was generated. Being a soft, flexible and water re-
tal material experimentation and simultaneously fea- pellent material, the silicone allowed for precise de-
tures the relationship between aspects of the digi- tail to be taken by the surface. Unlike conventional
tal craft and tool development for digital fabrication. moulds and formwork, the silicone can be easily re-
During the first research phase, experiments were un- moved, the ease and low cost of each print meaning
dertaken with fabric-cast concrete and 3D printing that each formwork is essentially disposable. Rather
with silicone. These initial investigations led to the than the costly and time consuming method of cre-

ating castings, or stitching objects into the fabric, the ics between design and fabrication process.A cen-
silicone alters both the structure of the form as well tral theme that developed throughout this research
as detailed surface finish. was the value of the low cost printer that evolves
to the specific requirements of a material. The ad-
Figure 2
vantage of manipulating a hacked open source 3D
Silicon 3D Printing
printer lies within the possibility of customising ex-
in combination
truder and nozzle for the use of a well defined range
with Fabric-Cast
of diverse materials, in this case clay, porcelain and
Concrete.
silicone, offering thus a greater flexibility in contrast
to commercial printers and so is open to elements
of disruptive innovation. There is currently no com-
mercial printer that would allow simultaneous exper-
imentation with this material range, or provide plat-
forms for different substrates.The experiments with
ceramic 3D printing focused on understanding the
parameters of correlation between the material and
Ceramic and Silicone 3D Printing. These initial ex- the machine. A wider range of materials for extrusion
perimentations with textile formwork and 3D printed has been explored, from clay to terracotta and porce-
patterns on the fabric evolved into an investigation lain, showcasing different extrusion behaviours. In
of additive manufacturing techniques firstly with clay order to obtain an optimal extrusion result, the fol-
and as a second step with silicon on textiles. For lowing parameters had to be considered: extrusion
these purposes a self-built, low cost 3D printer was rate, extrusion speed, material viscosity, nozzle type
developed. As starting point and technical instruc- and material depositing strategy. The focus of this
tion resource for the construction of the 3D printer, work lies in understanding how to manipulate and
the open source kits for a 3D Delta Printer for ce- control the ceramic 3D printer in order to then further
ramics provided by the researchers Jonathan Keep advance to developing a material depositing strategy
and Dries Verbruggen were used. The developed low for medical hand braces with locally deposited sili-
cost Delta Printer is suited to printing taller objects cone through 3D printing.
in clay due to the stationary circular base and tall
build volume, compared to a Cartesian Printer. It is Figure 3
only through the 'free revealing' and open source de- Delta Printer for
velopment that this type of cross-disciplinary project Clay and Cartesian
is accessible and manageable for architectural stu- Printer for Silicone.
dents. Despite beginning the project with minimal
experience of 3D printing from the student side, as
a 'user innovator' the internet allows 'access to the
rich libraries of modifiable innovations and innova-
tion components that have been placed into the
public domain' (Von Hippel 2005, 121). This new
type of emerging 'user-innovator' is characterised by Medical gloves and compression braces. A key aim
a movement between roles, in this between archi- that developed during the further line of investiga-
tectural designer to product engineer, which inher- tion was to find relevant applications for the low
ently enables a revised approach towards the dynam- cost method of printing silicone onto textiles. This

Figure 4 evolved into the creation of customised compression
3D printed pattern braces for medical purpose, by making use of low
design on flexible cost materials and accessible additive manufactur-
textile. ing technology. Initial printing tests were executed
with latex and silicon. The latter was adopted for
further experiments due to its homogeneity, which
corresponds to industrial standards, the more flexi-
ble extrusion process, easy availability and low cost
of material.The main design principles for the com-
pression braces and gloves relies on investigating
the relationship between patterning and stretching.
Using pattern design to influence the compression
and stretching behaviour of textiles can be of high
Figure 5 relevance for the medical development in order to
3D printing Silicone enhance ergonomic features of support braces or
Pattern Design on to customise protheses. The research development
Substrate. was done through multiple digital and material it-
erations, by simultaneously correlating material be-
haviour, pattern geometry, fabrication process and
machine. Through the parallel development of the
material, digital and mechanical customised flexible
prototypes were developed. In order to obtain work-
ing prototypes a precise calibration of specially ma-
terial and machine parameters was necessary, thus
enabling the designer or architect to obtain an in-
creased control over both design and fabrication pro-
cess.
Software and Hardware. For this experimental se-
ries an existing plastic FDM printer was hacked and
technically upgraded. A new RAMPs board was in-
stalled as well as a new printhead developed to suit
the silicone. The main effort was put into control-
ling and manipulating the extrusion rate and flow of
the silicon, based on previous experience with ce-
ramic 3D printing. This involved developing an ex-
truder to mechanically dispense the silicone straight
from the cartridge, having previously explored but
realised limitations of printing with air pressure and
solenoid valve allowing precise control of extrusion
between movements. Rather than developing a sep-
arate reservoir of silicone, the extruder was designed
around a standard 300ml sealant cartridge. The ex-
truder developed within the Engineering FabLab

consisted of a mechanical screw, that forced the sili- developed. The theorist Richard Sennet describes
cone through a PTFE tube to a nozzle mounted onto 'tacit knowledge' as the implicit experiential under-
the printhead platform.The digital workflow con- standing of a process or material, '...knowledge that
sisted of designing the differentiated patterning, so has become so self-evident and habitual that it seems
custom print paths, to enable the creation of silicone just natural' (Sennet 2008, 183). From an educa-
objects impossible with standard slicing software. tional point of view the tacit knowledge represents
The modelling was based in Rhino and Grasshopper an emerging set of skills which are typical for crafts-
was used to vary the density of the pathways over manship and can be only developed and accessed
a cylindrical form, originally scan of an arm, and so through a hands on approach. This applies not only
influence and manipulate the areas of compression. to the traditional context of craft, but also to the dig-
This output was then turned into a single path, to ital context. By being directly involved in the me-
enable consistent, clean printing. As the mechani- chanical assembly of a 3D printer an experience is
cal extruder of the printer did not have the capabil- being generated which is of significant educational
ity to consistently retract between movements, it was value and confirms the concept of machinecraft: the
necessary for each layer of the object to be printed architect actively engaging in the technological de-
in a single path. This printing strategy had a sig- velopment of the chosen digital fabrication tools in
nificant impact on the process of designing the fi- order to adapt them to specific project demands and
nal prototypes. Further experimentation was done to thus overcome machine limitations. This type of
with an XBox Kinect 360 scanner, and free software consequential emerging knowledge through direct
to scan elements of the body in order to increase the process involvement can be also defined as 'sticky'
ergonomics of the braces. As a final step Silkworm, a or 'local' information. The American economist Eric
plugin for Grasshopper, was used for generating the von Hippel considers that this 'information used in
G-code. The workflow of scanning, manipulation and technical problem solving is costly to acquire, trans-
printing led to a series of combined digital simula- fer, and use in a new location' (Von Hippel 1994, 429
tions and physical experiments, in order to obtain a ). This is the information immediately available to the
higher design refinement. user-innovator, that is used to further development
of the product or process to suit immediate needs.
This generates a valuable educational insight as the Figure 6
interdependence between design, material, process Customised Hand
strategy and tool becomes apparent to the student. Brace with 3D
printed Silicone
Research Study 02: Minibuilders Pattern.
Context and Concept. Constructing to the previous
case study which focuses on small-scale prototyp-
ing and on the integration of new digital cultures as
part of a revised academic curriculum, the following
project creates a change of scale, as it engages with
on-site fabrication strategies. The case study at issue,
Conclusion. Through engaging with textile form-
bearing the name Minibuilders, was developed as a
work to building a 3D printer and developing 3D
robotic research conducted as a group project at the
printed pattern designs for textiles, a profound un-
Institute for Advanced Architecture of Catalonia in
derstanding was gained of related software, hard-
Barcelona. Minibuilders is a project which deals with
ware and the material behaviour. During this exper-
architectural robotics and additive manufacturing for
imental series a tacit knowledge of the process was

big-scale structures. The case study illustrates the de- tures to the functioning of 3D printers. The drafted
velopment of a robotic fabrication strategy based on strategy relies on dividing the on-site construction
the use of a multiple robotic system which uses a ma- processes into three phases, according to functional
terial distribution system similar to the current 3D- necessities. The three phases are consecutive and
printing, but adapted to on-site fabrication.The most each correlates with the use of a different robot. The
relevant aspect of this research lies within the fact first robot, the foundation robot, to come into opera-
that instead of choosing to work with industrial robot tion is responsible for raising up the first ten to fifteen
arms for the construction of architectural elements layers which form the foundation of the future struc-
such as walls or columns, a reversed approach was ture. Subsequently, after the foundation is finished,
taken. Using robotic arms or portal 3D printers for the second robot continues depositing the following
construction is based on the idea of using oversized layers and finalises the design. Whereas the founda-
tools, so heavy machinery, in order to build equally tion robot is capable of moving on the ground, the
big structures. Operating with such tools implicates second robot, named grip robot, needs to be manu-
a series of considerable disadvantages for on-site fab- ally positioned on top of the finished foundation lay-
rication: ers. According to the task it needs to fulfil, the grip
robot is designed as a type of climber robot. After
• Inflexibility due to the physical size of the being placed in its position, the grip robot contin-
robot. ues with the successive deposition of the layers. The
• A limited area of reach depending on the grip robot is the robot which completes the form and
length of the robot arm or the size of the 3D which is responsible for the main construction task.
printer. As both the foundation and the grip robot deposit
• Reduced mobility, as industrial robot arms horizontal layers, any resulting structure, indepen-
move on trails and are not able to move inde- dent of its shape, will exhibit a restricted structural
pendently across space. stability. Naturally, the cause of this lies in the ab-
• Building limitation through the size of the sence of vertical reinforcement. In order to counter-
printer itself. In case of portal printers, the act this effect and to offer an increased structural sta-
structure to be erected has to always be bility, vertical layers along the horizontal ones must
smaller than the machine itself, just as in the be added. Concluding, the third phase seeks to ad-
case of a commercial 3D printer. dress this problem and deals with the construction
• Heavyweight for on site construction, as the of the earlier mentioned reinforcement layers. This
additional load of the robotic arm has to be stage is based on the utilisation of the vacuum robot
considered during the planning and the or- which is designed for ensuring a vertical motion. As
ganisation of the construction phase. indicated by its name, the robot creates a vacuum be-
tween himself and the surface of the structure in or-
The Robots. Taking these criteria into consideration, der to be able to advance vertically. The robot moves
the here developed strategy for on-site printing is along the structure, by following predefined paths.
predicated on the development of a series of mo- These pathways originate from a previous, detailed
bile robots, which can act independently from one structural analysis of the design. Grasshopper and
another and thus fulfil separate functional demands. the plug-in Karamba were used in order to gener-
The three developed robots with their built-in tech- ate this information. After the completion of this last
nology represent a hybrid between robotics and 3D layer, design and construction process can be consid-
printing: while the mechanic specifications corre- ered as finalised.
sponds to the ones of robots, the integrated mate-
rial deposition system correlates in its procedural fea-

Material System. The used material for the extrusion Future Investigations. The described printing and
is a composite material based on marble powder and construction strategy represents a proof of concept
a two-components resin system. The two compo- to showcase that small robots may be more appropri-
nent resin system was added as to accelerate the cur- ate to be used o print large-scale structures on site,
ing time of the extruded layers in order to facilitate rather than industrial robot arms, which are more
a smooth and stable movement of the robots on top suited for fabrication processes along an assembly
of the deposited material layers. A series of parallel line. The refinement of the robots in terms of their
physical and digital experiments was done so that the moving mechanisms, the development of a sustain-
right mixing ratio to obtain an optimal viscosity and able material system, the detailing of the extrusion
texture of the material for the extrusion could be de- nozzle and their further specialisation regarding the
termined. The development of the material, the ex- tasks they fulfil may represent main investigation ob-
trusion nozzle and the extruder happened in parallel, jectives for future research. This project showcases
as they represent correlated and interdependent pa- the importance of architects being directly involved
rameters. Material, machine and building data thus in the development of actual fabrication strategies, in
form a reciprocal informative circuit. This type of ap- terms of machine development and process restruc-
proach generates a new level of complexity regard- turing and organisation.
ing the architectural input and the levels of inter-
vention, control and manipulation through the archi- CONCLUSIONS
tect.Considering that the here showcased study was The implications of the architect as a robotic in-
handled as a proof of concept, only a limited amount ventor and the incorporation of machine, fabrica-
of exploration was invested in the development of tion and material parameters in the architectural pro-
the composite material itself in terms of its fitness for cess are evidenced. Through the integration of self-
the construction industry. Considering the addition developed robotic devices as autonomous design
of the two-component resin system the material can- agents a new understanding of digital fabrication,
not be considered as fully sustainable or ecological. the construction site as a robotic environment and
Figure 7
Minibuilders, from
left to right: grip
robot, foundation
robot, vacuum
robot.

Figure 8
Minibuilders while
printing
protostructure.
the crafts nature of the architectural discipline may cesses, adding an element of manageable complex-
arise.Through the means of open source platforms ity, so that the parallel development of hardware and
and programming new areas of competence, such software can sustain an intelligent network develop-
as product engineering, material development and ment of design, material, fabrication and machine.
mechanical engineering, but also influence can be
accessed by designers and architects alike. The im- Acknowledgements
portance of mastering these digital tools goes be- This paper and the presented theoretical part re-
yond just expanding a digital skill-set, but it demon- lies on the author's PhD thesis, completed under
strates the disciplinary relevance of advancing in par- the guidance of Professor Dott. Arch. Paolo Fusi,
allel design and fabrication strategies, which are in- conducted at the HafenCity University Hamburg and
trinsically connected to one another. The emerg- funded by the DFG, the German Research Foun-
ing digital tools of design and fabrication, specially dation.Minibuilders, the here described research
the generation of new desktop fabrication tools such project, was developed during the Open Thesis Fab-
as 3D printers, laser cutters or 3D scanners in com- rication program (OTF) 2013/2014 at the Institute for
bination with the developing open source culture Advance Architecture of Catalonia. The OTF program
have the potential to revolutionise and democratise is constructed on the idea of offering a suitable aca-
the discipline and thus initiate a new industrial rev- demic framework and guidance for its participants to
olution. (Anderson 2012) Digital fabrication prac- conduct individual or group research according to a
tices have the potential to expand the control of the self-imposed research agenda. Minibuilders was con-
maker, in this case architects and designers, over the ducted by a group consisting of six researchers, un-
design process and to overcome machine limitations. der the guidance and with the help of multiple staff
They can be seen as extensions of the traditional pro- members of the mentoring institute. Further infor-
Figure 9
Minibuilders during
on-site printing.

mation on the individual members of the research
team, the extended academic support, sponsorship
and collaborators can be viewed on the official web-
page dedicated to this project.
REFERENCES
Anderson, C. 2012, Makers : The New Industrial Revolu-
tion, Random House
Von Hippel, E. 1994, '"Sticky Information" and the Lo-
cus of Problem Solving: Implications for Innovation',
Management Science 40, 4, pp. 429-439
Von Hippel, E. 2005, Democratizing Innovation, MIT Press
Menges, A. 2008, 'Integral Formation and Materialisa-
tion: Computational Form and Material Gestalt', in
Kolarevic, B. and Klinger, K. (eds) 2008, Manufactur-
ing Material Effects: Rethinking Design and Making in
Architecture, Routledge, New York, p. pp. 195 – 210
Nan, C. 2015 'Architectural Robotics', Computer-Aided Ar-
chitectural Design Futures. The Next City - New Tech-
nologies and the Future of the Built Environment, 16th
International Conference, CAAD Futures, Sao Paolo
Ratti, C. and Claudel, M. 2015, Open Source architecture,
Thames and Hudson
Sennett, R. 2008, The Craftsman, Yale University Press.

CAAD EDUCATION | Creative
Practices
Computation in Early Design Education as Investment in
Attitudes
Desantila Hysa1 , Mine Özkar2
1
Epoka University 2 Istanbul Technical University
1
drrumbullaku@epoka.edu.al 2 ozkar@itu.edu.tr
While education programs are generally defined in terms of learning outcomes

known as knowledge, skills and attitudes, it is not always obvious how attitudes
are to be gained. This study focuses on the formation of attitudes of
accountability and sharing of knowledge in computational approaches to basic
design education. We posit that computational thinking, even without computers,
is supportive of both the reflective practice and the learning of such values in
design education. We report on comparatively observed collaborative design
processes of first year architecture students who externalized their thought
processes through visual rules. While a reflection-in-action stage helps in
learning the know-how, a second reflection on reflection-in-action deepens the
understanding and initiates habits for how to think and act within and beyond the
design domain, leading to attitude formation.
Keywords: design computation, attitudes, visual rules, learning outcomes,

foundation design studios
INTRODUCTION: COMPUTATION AND edged in beginning architectural design studios.

MORAL COMPETENCES By definition, learning outcomes are a compos-
Knowledge, skills and attitudes are often presented ite of knowledge, skills and attitudes that an individ-
as a triad of competences within the context of learn- ual will attain as consequence to his or her successful
ing outcomes in higher education (Adams 2004). engagement in a particular set of higher education
Learning outcomes, for individual modules of educa- experiences (Adams 2006). Nevertheless, while most
tion such as a course or a program constitute state- institutions seek to articulate the learning outcomes
ments of what a learner is expected to know, under- in terms of acquired knowledge and developed skills,
stand and be able to demonstrate as competences at they do not explicate how attitudes are expected to
the end of a period of learning. For instance, learning be gained perhaps due to the difficulty in assessing
outcomes for a first year design studio are defined in attitudes in short term education cycles.
terms of the basic design skills and knowledge that Attitudes are notoriously indistinct in computa-
the student is expected to acquire. Recently, com- tional design education. Often deemed a technol-
putational design skills, mostly the idea of external- ogy, computational design is usually associated with
izing processes with (ideally visual) rules, but also the skills and abstract knowledge that absolves the stu-
use of digital tools, have slowly come to be acknowl- dent of the social responsibilities of design thinking.
CAAD EDUCATION | Creative Practices - Volume 1 - eCAADe 34 | 357

We posit that computation enables the designer to that design thinking should be included as a funda-
externalize the process in service of a social point of mental competence to be acquired in higher edu-
interest and thus instill desired attitudes, i.e. moral cation. Even if this suggestion is not commonly fol-
competences such as sharing knowledge and ac- lowed through, Buchanan (2008) has been advocat-
countability. ing the teaching of design attitudes such as respon-
Studies that demonstrate how attitudes are con- sibility for good organisation and decision-making in
veyed to learners in higher education is scarce. Skills disciplines other than design.
generally refer to the learned capacity to do some- Similarly but within the context of the design stu-
thing. Instead, attitudinal or affective outcomes usu- dio, we suggest that the knowledge and skills ac-
ally involve the change or development of particular quired while dealing with form in design education
values such as empathy, ethical behaviour or respect may extend to attitudes and organisational ability.
for others (Ewell 2001). Such values, noted as "the Stiny (1980) points out that visual thinking is instru-
hidden curriculum" (Dutton 1991) of a learning set- mental in understanding the spatial relations that
ting, may arise not only from the social context but constitute designs. While rule-based design meth-
the curriculum as well. In this respect, Adams (2006) ods are generally associated with parametric design,
suggests that learning outcomes should not be con- rule-based approaches can support visual and per-
sidered as statements only. Instead, they represent ceptual properties of shapes. There have been pro-
a particular methodological approach for curriculum posals in integrating such approaches to design edu-
design. cation in general (Knight 1999) yet there is not much
As such, being able to perform a certain set of research done on the attitudes supported by such
domain-related skills is not enough, unless the stu- pedagogical approaches.
dent acquires a general habit of making use of those We aim to show that computational design
skills inside and outside the domain (Resnick 1999) thinking supports the learning of positive values in
and with a broader picture in mind. Attitudes are design education. The hypothesis is merely based on
generally defined as a collection of habits that drive the fact that it enables an externalisation of design
personal behaviors as well as one's patterns of think- thinking, sharing, and hence accountability. Our mo-
ing. In this paper, the scope is limited by account- tivation is to provide yet another reason for introduc-
ability and collective practices that we propose as di- ing computing early on in design education among
rect results of computational thinking in design edu- with other fundamental skills and knowledge. Try-
cation. ing to grasp the relational system underlying a de-
sign through visual rules helps in developing rea-
Visual Rules in Foundation Design Studio: soning, choices, and judgement. A designer's shape
Reflection in and on Action rules are not only tools to control the relations be-
In foundation design studios, the student is usually tween shapes but they also facilitate the reasoning
engaged with various abstract tasks to define rela- process towards an increased understanding and the
tions between forms as the given formal vocabulary improvement of personal attitudes.
facilitates the reasoning process. Commonly there
is interest in utilising tools such as comparative dis- A DESIGN EXPERIMENT IN A FOUNDA-
play, conversation, and systematic visual representa-
TIONAL DESIGN STUDIO
tions, to support the externalisation of the reason-
Our study involves comparatively observing collabo-
ing. These approaches encourage the students to re-
rative design processes in volunteering first year un-
flect on their own reasoning in the decision-making
dergraduate architecture students, over the course of
context. Duly, Meijers et al. (2005) has suggested
a month in various types of group projects. In the
358 | eCAADe 34 - CAAD EDUCATION | Creative Practices - Volume 1

study, students were encouraged to externalize their while defining relationships in different orders.
thought processes through rules for reflection. While The sequence of the assignments was as follows:
a first stage of reflection helps in learning the know-
how or design skills, namely reflection in action, a 1. Folding a blank sheet of paper in five consec-
second stage of reflection is necessary for developing utive steps.
habits of the mind and of knowing how to think and 2. Unfolding the paper, emphasizing the folding
act: reflection on reflection in action. In two tier tasks, lines and extracting emerging shapes.
students first completed a design assignment, then 3. Making a two-dimensional composition with
they were introduced to notion of visual rules while seven of the elements derived from Stage 2;
assessment. Students were then asked to interpret each repeated three times. Overlapping is al-
others' work through visual rules and apply them. lowed; materials are black and grey on white
This protocol was repeated in a sequence of as- paper (35x50 cm). Students are introduced
signments so that students might develop regular to visual rules while collective assessment, by
opportunities to critically evaluate themselves and identifying possible rules within existing com-
their peers. The assignments progress by adding a positions.
new variable in each stage. Towards the end of the 4. Making a two-dimensional composition with
process, they were expected to reflect on the collec- three of the elements used in Stage 3; each re-
tive and individual process comparatively, in terms peated seven times. Overlapping is allowed,
of what rules they used and how changed their own materials are black and grey on white paper
rules on the way. This demonstrated that rule-based (35x50 cm). Then, exchanging the design
approaches to design help not only in dealing with work with the fellow student. Identifying with
form relations, but they become instrumental for the visual rules, possible repeated patterns and
external interpretation of actions. When treated as relations present in the design organization.
a whole, rule-based approaches help in developing (Student 1 on Student 2 and Student 2 on Stu-
strategies of knowing how to act and think within dent 1)
and beyond the design domain, leading to attitude 5. Re-making a two-dimensional composition
formation. with three elements used by the fellow stu-
The design experiment was conducted with fif- dent, each repeated seven times. Students
teen volunteering first year undergraduate architec- are asked to follow and apply the identified
ture students at Epoka University in Albania. Stu- rules in the design of their peer. Scaling and
dents had been exposed, for at least one term, to the overlapping is allowed; materials are black
basics of design. The design tasks were structured and grey on white paper (35x50 cm). Ex-
on two-dimensional formal organizations achieved changing again the design work with the pre-
through a cut-and-paste technique of various shapes vious fellow student. Identifying with visual
within a given boundary. As part of the design rules, possible repeated patterns and rela-
knowledge conveyed in the foundational design stu- tions present in the design organization: Stu-
dio, students were asked to achieve unified wholes dent 1 on (Student 2 on Student 1) and Stu-
by paying attention to the relationships established dent 2 on (Student 1 on Student 2)
between primitive elements and at the same time to 6. Making a two-dimensional composition with
preserve the identity of those elements. Overlaps of the three elements used in Stage 4, by follow-
elements were allowed. Moreover, students were en- ing the rules identified in Stage 5. Apart from
couraged to consider the shapes emerging from the scaling and overlapping, one of the elements
overlapping and the background as a design input is assigned a red color, while the rest is black
and grey on white paper (35x50 cm).

In the normal run of a foundational design course, as- folding pattern. Below are illustrations for some of
signments are of short duration and proceed in a fast the students' works corresponding to this step. Stu-
pace by adding a variable each time. In this proposal, dents were not yet introduced to how visual rules
we followed a slower process that also moved back work within a design organisation. Although there
and forth. Students were asked to interpret shape re- were some attempts to consider the background and
lations with visual rules, and to re-make the assign- the overlapping as design input by Student S1 and
ments by reflecting on rules. The aim of interpret- Student S3, as illustrated in Figure 2., the organisa-
ing the design works with rules is to show students tions generally lacked order (Figure 2, Student S4)
that rules are good tools to formalize their design and non-repetitive shape-relations were prevailing
thinking process and to freeze in time the possible (Student S2).
decisions taken. Such an interpretation helps stu-
dents exchange their ideas and learn from one an- Figure 1
other, even without the presence of an authoritative Different folding
figure, such as an instructor. In the long run, this may patterns produced
support students to become self-critical. The prob- by students during
lem with rules stands in their abundance. Students the study. Their first
had difficulties in deciding on which rules to identify move was the same:
and apply. Generally students were encouraged to the shortest side
identify rules standing for the general layout orga- was folded onto the
nization of the canvas (35x50 cm), part-to-part and longest side.
part-to-whole relationships.
The Process: Reflection-in-Action and Re-

flection on Reflection-in-Action
The series of the design assignments started with a
paper folding exercise (Figure 1). Each student was
given a blank sheet of paper in A4 format. Without
using any drawing or measurement equipment, stu-
dents were asked to fold the paper in five consecu-
tive steps. The haptic feedback is not taken into ac- At the end of Stage 3, students were introduced to
count. Paper folding served as a starting point to in- the notion of visual rules in support of the panel dis-
troduce students to the notions of shape embedding cussions that usually take place for assessment. They
as they start with a whole in their hands. Students were encouraged to view and represent their own
proceeded by folding what emerged from the pre- process through formal devices such as rules. View-
vious step in a repetitive fashion. While the guiding ing the whole process as computable, makes the stu-
principle of "the whole gets divided into its parts" was dents be in control of their design decisions. This is
the same for all of the students, the outcomes were also possible due to the nature of two-dimensional
varying. design works taking place in foundation design stu-
Students were asked to make a unified whole dios, where compositions are achieved through ge-
with seven shapes extracted from the folded paper ometric transformations of primitive shapes and set-
pattern and to repeat each shape three times. The tling relations between the different transformations.
sub-shapes vary for each student. The key is in the Keeping in mind the potentials of visual rules, stu-
making use of the relational system present in each dents were asked to re-make the assignment of Stage

Figure 2
Student works of
Stage 3, before
being exposed to
visual rules. (left to
right: Student S1,
Student S2, Student
S3, Student S4).
Figure 3
Design works of the
same students,
done in Stage 4,
after being exposed
to visual rules.
Figure 4
The interpretation
of the Student S3
work done in Stage
4, with a small
change of the
design as it could
have ended,
referring to the
followed rules. A
shape rule relating
the extracted
sub-shapes of Stage
2.
3, as prescribed in Stage 4. The works of the same pre- rules of the design process of Student S3 in Figure 3,
vious students are illustrated in Figure 3, after being done retrospectively. Ignoring the different weights
introduced to visual rules. There is a clear improve- assigned to the shapes in composition, the student
ment from Stage 3 to Stage 4, in terms of the spatial work is redrawn as shown in Figure 4 and Figure 5.
relations defined. This view is relevant for more than The following stages of the study were set in sup-
half of the group participating in the workshop. port of the argument that rules help the external-
The comparison between the two groups of illus- isation of the design thinking process, hence shar-
trations points towards more responsible acts taken ing and accountability. The development of such
by the students from Stage 3 to Stage 4, as a result attitudes necessitates the inclusion of collaborative
of making use of visual rules in their personal de- learning similar to role playing. The design process
sign processes. Below follows an interpretation with was reversed at this point. Students were first asked

Figure 5
The possible visual
process to obtain
Student S3 work:
(left to right, top
row) A shape rule
showing the
repetition of the
unit; the same rule
applied recursively;
(left to right,
bottom row) New
shapes are
perceived at
another level, as a
result of the
alignment of the
shape boundaries.
The new element
translates on the
horizontal axis and
reflects according
to the hypotenuse;
to identify some rules standing for the visually re- tion of the layout, parts are arranged in alignment Shape rule where
peating patterns and relations, within the composi- with each-other as shown in Figure 7. Student S6 the canvas height
tions done in Stage 4 by their fellow student. While acknowledges this spatial relation in her Rule 2 and becomes a
this reflection stage helps in learning the know-how reapplies it in her work by fitting the overall design determinant; a
of design, namely reflection in action, a second re- within the boundaries. As a result, in the second in- shape rule for
flection stage we think is necessary to deepen under- terpretation, the student handles the figure-ground assigning different
standing and developing habits of knowing how to relationship along the boundary better compared to tones to the parts,
think and act: reflection on reflection in action. After the first. by relating
identifying the visual rules, students, by working in Furthermore, Rule 3 and Rule 4 point to shape contrastively the
pairs, were asked to follow and apply the rules again relations standing for the overlapping between two figure and the
in a new organisation. This process was repeated shapes and the emergence of the background as an- ground.
twice. other shape. Student S6 wrote: "Something very pos-
Making students see their design works as visual itive in Student S5 work is that the overlapping part is
computations helps them not only observe and in- one of the main shapes itself. But the problem is that
terpret but also share their design decisions. In turn, she is not consistent in the repetition of Rule 3." Stu-
students become competent in critically judging dif- dent S6 commented on another rule as well: "I have
ferent design decisions rather than following them noticed that there is an approach where the back-
blindly. In Figure 6, additional to interpreting the ground participates itself to form shapes that already
composition with rules, Student S6 does not hesitate exist. But this approach was lost in many other cases.
to comment on possible "deficiencies" in the design Sometimes some strip lines are formed and the eyes
of her fellow Student S5. Starting with the organisa- get lost and confused while reading the shapes."

Figure 6
Interpretation of
Student S5 design
by her fellow
Student S6, through
visual rules.
Student S6 rules
point to shape
transformation
(Rule 1), general
layout (Rule 2),
shape relations and
emergence of new
shapes from the
overlapping and
the background
(Rule 3 and Rule 4).

Figure 7
(from left to right)
The design work of
Student S5, as done
in Stage 4; the
interpretation of
Student S6 on
Student S5 work;
the same work as in
the middle, but
different tones are
assigned.
Figure 8
(from left to right)
Design work done
in Stage 4 by
Student S1; its
re-interpretation
done by her fellow
Student S2 in Stage
5; design work
done by Student S1
in Stage 6.
These statements show once more the potentials of DISCUSSION AND CONCLUDING REMARKS
including visual rules in the studio discussions. Stu- This study turns a spotlight on attitudes that ideally
dent S6 not only interpreted design actions with ref- complement the acquired knowledge and skills, but
erences to visual rules, but was also able to pro- are often left tacit. Contrary to the general view of
duce arguments for why some relations were good seeing attitude formation as a relevant competence
and why some were not. Although there is room for master graduates, the aim here is to make atti-
for improvement, there is a clear evidence that the tudes part of the learning process at the very begin-
interpretation of the second student on the first is ning, along with the gaining of basic knowledge and
more consistent with the original intentions. We re- skills. Our focus is on the formation of attitudes that
port that ten out of fifteen students developed more directly relate to the sharing of knowledge and the
aware and consistent interpretations for the work of accountability of the design decisions.
their fellow students. The students not only inter- Accountability is defined as the responsibility for per-
preted what they saw, but also reflected on the other formance, or the obligation to report to others, to
students' reflections while both extracting rules and explain, to justify the taken actions (Huisman and
reapplying them. Currie 2004). Our basic premise is that making stu-
dents see their designs as visual-spatial computa-
tions helps the students observe what they are doing,
how they are doing and more importantly, that they
reach an understanding of why they are doing things

the way they do. Moreover, rules that capture the role of attitudes in student learning in higher educa-
visual stages of a transformation externalise knowl- tion. In this study we showed how students might
edge to be shared and to be questioned at various learn to be accountable to others by way of a com-
instances of design and production. Previous studies putational approach to the foundation design studio
have acknowledged the role of formal tools like visual course. The general objective is to make students
rules in externalising the design thinking process in learn to practice attitudes in other settings as well.
support of the "reflective practice". Previous studies have acknowledged the difficulty
Our study included observing fifteen first year of assessing attitudes as they are a time-dependent
architecture students with the aim of assessing atti- competency. A further development of this study
tudes. The students used visual rules as a tool to re- might be an integration of visual rules as part of a stu-
flect upon their personal and their fellow students' dio pedagogy for the whole academic year and to ob-
design processes while defining or identifying differ- serve the habits of thinking and acting that students
ent shape relations. We acknowledge that their pro- may develop over longer periods of their education.
cesses would have been better studied over a longer
period of time and in future studies, more system- Acknowledgements
atic methods should be tried for assessing attitudes. We sincerely thank the volunteering participants
All of the students were able to define rules and re- who are students at the Department of Architecture,
late parts at one level though some of them did not Epoka University: Ana Zhibaj, Arjola Pilua, Brisilda
succeed to form unified wholes as shown in Figure Cani, Dorel Habili, Ilda Sadrija, Ira Balukja, Jon Bruci,
8, regarding the work of Student S2 on Student S1. Kejsi Sula, Lediana Sula, Manjola Logli, Polikseni
The Student S2 himself is reported saying that he felt Bano, Redjan Hysa, Valeria Nikolli, Xhenisa Dushaj,
trapped while reapplying the rules as if he was doing Xhulia Xhakanaj.
the same composition again. This might be related
with the abundance of rules identified in a composi- REFERENCES
tion and their specifications. Adams, S. 2004, 'Using Learning Outcomes: A consider-
Nevertheless, by working in pairs, students tried ation of the Nature, Role, Application and Implica-
to be accountable to one another. For instance, one tions for European Education of Employing Learning
of the students claimed in the post-experiment dis- Outcomes at the Local, National and International
cussion that "although we were working on each Levels', United Kingdom Bologna Seminar, Edinburgh
Adams, S. 2006, 'An Introduction to Learning Outcomes',
other's design works, I felt the whole process very in Froment, E., Kohler, J., Purser, L. and Wilson, L.
personalised. Now I may easily recognise my fellow (eds) 2006, EUA Bologna Handbook: Making Bologna
student's work amid many others." This means that in Work, Raabe Verlags, Berlin
interaction with each other, students start to under- Buchanan, R. 2008, 'Introduction: Design and Organiza-
stand their personal differences as well as others' de- tional Change', Design Issues, Vol. 24, No.1, pp. 2-9
sign reasoning process. By visualising their steps as Dutton, T.A. 1987, 'Design Studio Pedagogy', Journal of
Architectural Education, 41(1), pp. 16-25
an articulation of reflection, students learn to share Ewell, P. 2001, 'Accreditation and Student Learning Out-
their learning with others and understand whether comes: A Proposed Point of Departure', CHEA Occa-
they are doing things differently or similarly. This sional Paper
may be developed further towards the role of rules Huisman, J. and Currie, J. 2004, 'Accountability in Higher
in developing personal styles and multiplicity in de- Education: Bridge over Troubled Water?', Higher Ed-
sign as shown briefly in the paper folding exercise in ucation, 48, pp. 529-551
Knight, T. 1999, 'Shape Grammars in Education and Prac-
Figure 1. tice: History and Prospects', International Journal of
This research is part of an ongoing study on the Design Computing, 2, pp. 2-13

Meijers, A., van Overveld, C. and Perrenet, J. 2005, Criteria
for Academic Bachelor's and Master's Curricula, TU/e,
Eindhoven
Resnick, L 1999, 'Making America Smarter', Education
Week Century Series, 18(40), pp. 38-40
Schön, D. 1984, The Reflective Practitioner: How Profes-
sionals think in Action, Basic Books
Stiny, G. 1980, 'Kindergarten Grammars: Designing with
Froebel’s Building Gifts', Environment and Planning B,
7, pp. 409-462

Objects-to-sense-with
Computational Tools for Embodied Spatial Learning
Athina Papadopoulou1
1
Massachusetts Institute of Technology
1
athpap@mit.edu
This paper proposes objects-to-sense-with as tools that provide sensory-based

learning of space and expand spatial knowledge beyond merely the formal and
visual aspects currently dominant in design discourse. To reintroduce
sensory-based learning methods in architecture education, this paper first revisits
the sensory pedagogies formulated in the early 20th century, reviews precedents
in the arts that utilize body-centered sensing technologies, and reframes previous
discussions on the pedagogical role of technologies as tools for thinking. Finally,
this paper describes the development of a wearable tool with embedded sensors
created by the author and describes how the results are visualized. The developed
tool, which is used to record sensory data in-situ by the user and allows for a
body-centered representation of space, serves as an example of an
object-to-sense-with that can be used to achieve a sensory-based and
body-centered understanding of architecture.
Keywords: architecture education, design tools, sensor technologies, embodied

perception
INTRODUCTION sign pedagogies into an active, situated exploration

Although physical spaces are being perceived and of physical spaces that incorporates all the senses?
experienced through all the senses (Millar 2008), the This paper proposes the use of computation
design of spaces has been traditionally taught in ar- tools as "objects-to-sense-with" in the built envi-
chitectural studios, usually focusing on formal visual ronment which allow us to acquire embodied spa-
qualities at the expense of the other senses, and sep- tial knowledge. These tools, which take advan-
arating spatial learning from the actual physical envi- tage of today's self-tracking and sensing technolo-
ronment. A great number of studies in the field of en- gies, suggest methods inspired by Montessori's and
vironmental psychology have demonstrated the di- Moholy-Nagy's sensory training exercises. Alluding
rect psychological response of the sensory aspects of to Seymour Paper's "objects-to-think-with", objects-
space to those who occupy it (Gifford 2001). These to-sense-with suggest an embodied pedagogical de-
studies however have only had a limited influence in sign approach that regards computational tools not
the formal education of the architect. What learn- as the educational end goal but as powerful tools for
ing tools and methods can we use today to turn de- learning.

Precedents in education, arts and technologies that surroundings through which an appreciation of the
serve as a background for the proposed objects are sensory qualities of the material environment can be
first reviewed, then a description of a wearable com- achieved (Montessori 1912).
putational tool developed by the author is provided A few decades after the formulation of the edu-
as an example of an object-to-sense-with, and finally, cation of the senses by Montessori, László Moholy-
possible scenarios of such tools in educational set- Nagy formulated sensory training methods for archi-
tings are being discussed. tects and designers while teaching in the Bauhaus
school. According to Moholy-Nagy, the first-year ed-
EDUCATION OF THE SENSES ucation at the Bauhaus school aimed at broadening
The focus on sensory training used by physicians the sensory experiences of students and enriching
such as Jean-Marc Gaspar Itard and Edouard Seguin their emotional values. Moholy-Nagy's sensory train-
in special education in the middle of the 19th and ing exercises, offering knowledge beyond textbooks
beginning of the 20th century, provided the back- and descriptions, helped in providing an experien-
ground for Maria Montessori's radical sensory peda- tial understanding of textures and tectonics (Moholy-
gogies (Winzer 1993; Montessori 1912). The aim of Nagy 2005).
the education of the senses as formulated by Montes- Like Maria Montessori, László Moholy-Nagy was
sori in the Montessori Method was "the refinement concerned about providing the students with an un-
of the differential perception of stimuli by means derstanding of the various different sensory qualities
of repeated exercises" (Montessori 1912, 173). In of material objects. Through sensory training exer-
the Montessori Method, Maria Montessori describes cises, students were able both to build a repertoire
techniques for the training of each of the senses, each of different sensations and identify the subtle dif-
of which involved a special set of didactic material, ferences within the same sensation. The students'
from sets of wooden blocks with different textures for projects varied from tactile tables to a "luna-park for
training the sense of touch, to sets of bells for training the fingers" to a "smell-o-meter." Such projects al-
the sense of sound. lowed students to study the senses by making tools
According to Montessori, the isolation of the to study and visualize them. Although the sensory
sense under study was an important parameter of assignments allowed students to measure the dif-
sensory education. To train the sense of touch, chil- ferent sensations, the goal was not a scientific mea-
dren would be asked to discriminate between tex- surement and taxonomy of those but rather an ex-
tures while blindfolded; to train the sense of sound, periential learning though creative experimentation
children would be asked to discriminate between (Moholy-Nagy 2005).
sounds in the dark (Montessori 1912). In order to
engage the students in a creative participatory sen- SELF-TRACKING FOR SELF-DIRECTED
sory exploration of the environment, these sensory LEARNING
exercises would usually take the form of games. Sensing technologies that keep track of data related
By extending children's perceptual understanding to their health, performance and mood allow peo-
through the sensory training exercises, Montessori ple to self-monitor their behavior and empower them
aimed at equipping her students with refined aes- to take control over their actions through expanded
thetic judgment. Montessori's radical pedagogies self-awareness. In the Quantified Self movement,
were structured around auto-education methods, fo- dedicated self-trackers collect, measure and analyze
cused on learning through self-discovery and self- data about their behavior through phone applica-
instruction. Her sensory training methods assume tions, self-made tools and other tracking devices
an auto-didactic exploration of one's own physical aimed at an empowered self. According to Gary

Woolf, co-founder of the Quantified Self movement, FROM OBJECTS-TO-THINK-WITH TO
tracking data about one's own body and behavior is a OBJECTS-TO-SENSE-WITH
way to take control of the data in a data-driven world In his book Mindstorms, Seymour Papert (1993) used
and turn digital technologies into self-evaluation and the term objects-to-think-with to refer to objects
self-discovery tools [2]. The automated digital track- and technologies that act as tools for thinking, and
ing process allows for surprises in the learning pro- particularly to stress the importance that computers
cess as it expands self-awareness beyond physical can have in self-directed education. Papert broadly
observation, revealing aspects about one's own self defines "objects-to-think-with" as "objects in which
that would otherwise be unnoticed [3]. there is an intersection of cultural presence, embed-
In a similar manner that proponents of the Quan- ded knowledge, and the possibility for personal iden-
tified Self movement claim to gain a better aware- tification" (Papert 1993, 11). In Mindstorms, Papert
ness of their own bodies and behaviors, we might suggests that just as the gears he played with in
argue that tracking technologies could help us be- childhood became his personal model for mathe-
come more aware of our sensory responses to the matics, computers could universally become objects-
built environment through the mapping of our sen- to-think-with for children in learning new concepts.
sory actions in space. In fact, artists have already Papert introduced the Turtle in the LOGO program-
taken advantage of such technologies to promote a ming language in order to make the learning process
body-centered awareness of the built environment. in computers, as well as the learning process about
In his "Bio-Mapping" project, Christian Nold created computers, more tangible and concrete.
a body-centered cartography of the city by superim- The LOGO Turtle moves on the screen follow-
posing people's paths and emotions on city maps. ing the user's typed instructions and gradually forms
The bio-maps were made possible by asking partic- geometrical shapes by creating a line along her
ipants to walk in the city while using a simple self- path. Children can identify with the Turtle, imagin-
tracking device consisting of a biometric sensor and ing themselves moving while she moves. As Papert
a GPS (Nold 2009). Another body-centered spatial argues, "Drawing a circle in a turtle geometry is body
project created through self-tracking was "Running syntonic in that the circle is firmly related to children's
Stich," by Jen Southern and Jen Hamilton. Running sense of and knowledge about their own bodies" (Pa-
Stich was exhibited as a 5m x 5m projection of the pert 1993, 63). In this way knowledge becomes ac-
multiple urban trajectories traced by different partic- tive, situated, and self-directed. Children manage
ipants who used their GPS phone applications [1]. to acquire knowledge in a concrete and personal-
Such approaches point toward alternative ge- ized manner, taking control of their learning. Papert
ographies, mappings, and representation of the uses Piaget's distinction between concrete and for-
physical environment based on the emotions and the mal thinking, emphasizing the importance of con-
senses. If sensory education in the beginning of the cretizing abstract concepts. Whereas concrete think-
20th century was made possible by special sets of di- ing is embodied and occurs in an early developmen-
dactic materials and experimental studio prototypes, tal stage, formal thinking is an abstract form of think-
this paper suggests that by taking advantage of to- ing and occurs later in the development of the child.
day's technologies we can propose novel tools and Although Papert mostly refers to children's learn-
methods for the enrichment of the senses that not ing methods, both children and adults can bene-
only render students aware of the material qualities fit from body-centered, situated and active learning
of objects but also take the education of the senses strategies. A great amount of literature in embod-
out of the classroom, allowing for direct, experiential ied cognition (Johnson 1990) and enactive cogni-
learning of the built environment in-situ. tion (Noe 2006) argues for body-centered knowledge

Figure 1
Wearable tool with
embedded sensors
developed by the
author as an
example of an
Object-to-sense-
with.
and embodied perception. Moreover, research in the term suggests a pedagogical role for computa-
the cognition of space demonstrates that memory of tional tools in architecture and other the spatial dis-
spaces is mainly body-centric: people tend to recall ciplines focused on the senses and the physical ex-
spaces they have experienced relative to their bodily ploration of the built environment.
actions and movements (Papadopoulou 2015).
Alluding to Papert's objects-to-think-with, the THE DEVELOPMENT OF AN OBJECT-TO-
term objects-to-sense-with coined here refers to the
SENSE-WITH
use of objects that promote sensory-based learn-
When visiting architectural spaces of interest, archi-
ing by allowing knowledge to emerge through our
tects generally use photos, videos and quick hand
bodily actions in the physical environment in a self-
sketches to capture and document the form and ex-
directed manner. Tools for architectural represen-
perience of space. Even though perspective draw-
tation and documentation of space are traditionally
ings, images and videos express an experiential point
focused on the built environment and material ob-
of view of the building, they are usually based on
jects per se, excluding the body and the senses from
visual experience, having limited reference to ele-
the depiction of space. Using objects-to-sense-with
ments of the tactile, aural, olfactory and kinesthetic
to perceive and analyze space can lead to sensory-
experience of space. Thus, the motivation behind the
based architectural maps and drawings through the
object-to-sense-with developed by the author was
depiction of the interactions between the body, the
to suggest a tool that can offer multimodal body-
built environment and the objects within.
centric documentation of spatial experiences. This
As in the case of the LOGO Turtle, objects-to-
multimodal documentation of spatial experience can
sense-with can be body syntonic drawing tools that
allow architects and students to engage in a spatial
depict space based on one's own paths, trajectories
discourse beyond mere formal aspects. The depic-
and bodily actions. Objects-to-sense-with are not
tion and documentation of our bodily interactions
tied with specific technologies or aesthetics. Rather,
with space can allow a discussion on experiential out-

comes of specific architectural solutions and com- icant variance in the functions and indoor/outdoor
parisons between different architectural spaces on uses, the temperature values would remain more or
a sensory basis. Thus, built spaces could be ana- less the same. Thus, there would be no need for real-
lyzed based on possible trajectories of movement or time, continuous tracking of the temperature data as
on tactile or even olfactory responses. Although the there would be no significant variation in the values.
same objects-to-sense-with could be used in various
Figure 2 spaces by different people, each outcome would be
Example of unique since sensory-based documentation is always
body-centered temporal, ephemeral and personal. Therefore, using
visualization of objects-to-sense-with does not imply making objec-
spatial interactions tive measurements. Rather, it allows us to document
using the sensory experiential traces, suggesting new possibilities of in-
data recorded teractions and motivating learning through personal
through the Object- exploration and discovery. Through the mapping
to-sense-with of body-centric data, objects-to-sense-with can offer
demonstrating the ways to render visible and tangible the unseen qual-
rules used for the ities of space and thus bring into the studios and the
visualization. design discourse elements beyond the mere material
aspects of built space.
The prototype developed by the author (Figure
1, Figure 3) is only an example of a possible object-to-
sense-with. In an educational setting, students could
build their own objects-to-sense-with and adjust the
architecture of the tool to the sensory interactions
they would like to focus on. As was the case in the
sensory training methods used by Montessori and
Moholy-Nagy, different sets of didactic material and A simple two-dimensional sensory notation system
different sensory training and mapping tools could was developed that could also be drawn manually,
be used for each sense or senses under study. In or- allowing for comparisons between the automated,
der to provide evidence that such a tool can be de- software-generated sketch and the one drawn by the
veloped in an educational setting that does not re- user. The prototype, which was designed as a wear-
quire advanced hardware and software knowledge, able tool with embedded sensors, collects data re-
simplicity in the technical implementation was an im- garding the position x,y of the user in space, as well
portant parameter in the development of the tool. as the sound and the tactile interactions of the user
In order to record and represent the sensory data at each x,y position. Since the tool is continuously
in a simple manner, a set of sensors related to vi- recording the data, when the data are inserted into
sual, tactile, auditory, kinesthetic sensory interac- the visualization software, the movement path of the
tions were used and a simple set of rules to visualize person in space is gradually recreated as an animated
the data (Figure 2). Sensory information that required graphical representation, together with a representa-
complex implementation, such as information re- tion of the auditory and tactile interactions happen-
lated olfactory sense, was not addressed. Moreover, ing along the path.
environmental qualities such as temperature were By depicting each location of the person as a
not directly captured. Unless the space had a signif- point in space, the trajectory of the person in space
is finally visualized as a line. The time the person

spent at each x,y location is visualized by varying the The values from the magnetometer and ac-
diameter of the points that compose the line of the celerometer combined were used to calculate the
path. The tactile interactions are visualized as diago- movement path of the user in space by providing in-
nal lines at each location, and the sound levels as cir- formation regarding the orientation of the user. The
cles with varying diameters, having as a center each magnetometer, which calculates the angle between
of the x, y locations that compose the final path (Fig- North and the device's y axis, was used to provide
ure 2, Figure 4, Figure 5). Two cameras, one added measurements regarding the position of the user.
to the wearable prototype, and one added as a head The accelerometer was used to distinguish between
equipment, complement the tracking setup to offer pause and movement and to compensate for the
additional information for processing regarding the magnetometer's tilt. The real-time clock was used to
visual and tactile interactions. synchronize all different data when processing them.
All measurements were directly stored in an SD
Implementation card using a microcontroller SD card shield and were
The prototype, which was 3d-printed in plastic, was imported afterwards as a text file into Processing soft-
designed as a wearable tool that can be adjusted ware. In the algorithm developed in Processing by
to the user's wrist. The microcontroller (Arduino the author to visualize the user's path, a 2D unit
Uno R3) as well as the sensors were part of the de- vector is being created whose location is continu-
sign of the tool and were embedded in the pro- ously being updated according to the stored mea-
totype. The prototype also enclosed a camera to surements of the magnetometer and accelerometer.
record the materials the user interacted with through Keeping the traces of the past 2D unit vectors, and
touch. The following sensors were connected to adding the newly instantiated vectors that represent
the microcontroller: An electret microphone, an in- the more recent locations in time, a line is gradually
frared proximity sensor (VCNL4000), a triple axis mag- being traced on the screen that represents the path
netometer and accelerometer (LSM303DLMTR), and of movement of the user of the tool exploring the
a real-time clock. The microphone, proximity sen- space.
sor and clock were embedded in the wrist-wearable Using the acceleration values, a threshold is de-
tool, whereas the triple axis magnetometer and ac- fined to distinguish between movement and pause
celerometer were attached to a chest-strap that the in the user's path. When pause is detected, a circle
users wear in addition to the wrist-wearable tool (Fig- is drawn having as center the current location of the
ure 3). unit vector, whose diameter varies according to the
The electret microphone was used to collect duration of the pause. To visualize the auditory infor-
information regarding the auditory interactions by mation, for each new unit vector that is being instan-
measuring the sound levels at each of the user's lo- tiated, signaling a new position of the user the space,
cations throughout the user's path in space. The in- the sound level values captured at the given posi-
frared proximity sensor was used to collect information and time are mapped into the diameter of a cir-
tion regarding the tactile interactions. The proxim- cle having as center the location of the vector. Thus,
ity sensor was attached to a ring on the user's finger the larger the circles, the louder the sounds that were
and was connected to the wrist-wearable tool. Since captured at the specific location and time in space.
the infrared proximity sensor provides information To visualize the tactile interaction in Processing,
regarding the distance of an object through the re- a distance threshold is defined to determine whether
flected infrared light, and the sensor is located at the the touch condition is true or false depending on
user's finger, we can conclude that a small distance the proximity of the objects detected through the IR
value signals a tactile interaction with an object. sensor. When a tactile interaction occurs, then a di-

agonal line, whose length can be adjusted to corre- visual attention. In both cases, the program takes
spond to the proximity of the object, is drawn on the a snapshot when the conditions that signify salient
screen starting from the location of the 2D unit vec- features are met. All features can be combined in
tor. The orientation of the line represents the direc- a single interface embedding information regarding
tion of the action of touch in space, as it corresponds movement, sound, touch and vision.
to the heading of the unit vector. In future development, the algorithm can be im-
To filter information from the camera, a color proved by taking into account the speed of move-
matching algorithm was developed, which, through ment. In the current state of development, only
color recognition of the video frames, discerns the pause and movement are represented in the gen-
objects that were part of the longest in time tactile erated sketch, assuming that the person is moving
interactions with the user. The same color-matching at constant speed. By using the acceleration values,
mechanism was applied to video recordings that the speed of movement can be calculated as well,
were captured by the camera close to the visual field rendering more accurate the mapping of the move-
of the user to detect the objects that were stared at ment path in space. Moreover, the use of a gyroscope
the longest, and possibly most attracted the user's could help distinguish between changes in the user's
Figure 3
The designed
prototype as an
Object-to-sense-
with and the
sensors used to
document sensory
interactions in
space.

Figure 4
Example of
body-centered
visualization of
spatial interactions,
using the sensory
data recorded
through the Object-
to-sense-with,
demonstrating a
high level of tactile
interactions, and
pause in multiple
locations during
spatial exploration.
orientation due to rotations of the body in the same applicable in large setting or outdoor environments,
location and changes of orientation due to walking since students would need to freely navigate. There-
in space. fore, it is suggested here that objects-to-sense-with
can offer an alternative to traditional sensory isola-
Scenarios of use tion methods by allowing the users to filter their sen-
An important aspect in the education of the senses sory interactions in space through the tool.
as proposed by Montessori and Moholy-Nagy is the One way to filter the sensory information is
ability to focus on one of the senses by filtering out to control how and which sensory data are being
other sensory interactions. In the tactile and auditory recorded. For example, users could adjust the hard-
training exercises they propose, students are usually ware and software to output information regarding
blindfolded so that visual information does not inter- only touch, sound or movement, combinations of
fere with the training of the other senses. Although these, or all of these together. This flexibility would
being deprived of other senses so as to focus on a allow architecture students to study a built space
specific sense might be an appropriate method for through the lens of a specific sense. Moreover, in
small exercises in a studio setting, it would not be the interface where the data are being visualized stu-

dents can alternate between different channels of chitecture students to communicate ideas in the es-
sensory information in a similar manner that they can tablished media used so far in the studio and archi-
add and subtract material information in a design in- tecture practice which is drawings and sketches. If
terface by switching on and off layers of grouped ob- an educational setting allows for more experimen-
jects. Superimposing the multimodal body-centered tal approaches, then one could imagine multisensory
maps onto a building's plans could provide insights representations of space. For example, in the case of
on the experiential qualities of the space and body- the developed tool, apart from a visualization of the
centered ways to reflect on future designs. sound levels, the recordings of the actual sounds can
A wearable tool for documenting our sensory in- be embedded in the interface, allowing interactions
teractions with the material environment can be a in multiple modalities, and rendering more complete
powerful learning tool, as it can take design educa- and accessible the provided information.
tion out of the classroom for in-situ spatial explo-
Figure 5 rations. A common procedure in architecture stu-
Example of dios is to review and analyze precedents as well as
body-centered to visit and analyze the area of intervention. Instead
visualization of of merely reviewing textbooks and documenting
spatial interactions, the site using videos, pictures and sketches, which
using the sensory mainly rely on visual means, objects-to-sense-with
data recorded would allow instead a multi-sensory situated learn-
through the Object- ing method of studying both built spaces and their
to-sense-with, surrounded physical environment, offering a body-
demonstrating a centered understanding and evaluation of architec-
high level of sonic ture.
interactions. The object-to-sense-with tool allows users to fo-
cus their attention in specific modalities while explor-
ing space, and thus provides a direct way of filter-
ing sensory information. A physical exploration of
spaces, that can be part of studio-based or other type
CONTRIBUTIONS
of educational setting, can be focused on a specific
To provide a historical background for the develop-
sense in order to evaluate spatial qualities and expe-
ment of alternative teaching methods in architec-
riences related to that sense. For example, if focused
ture that go beyond the formal and visual aspects of
only on auditory qualities, the tool would motivate
space and embrace all the senses, this paper offered
students not only to listen carefully to the environ-
a brief review of the sensory pedagogies formulated
mental sounds but also to generate sounds in order
by Maria Montessori in the Montessori Method and
to test the acoustic qualities of both the geometry of
introduced into design education by László Moholy-
the space and the material used.
Nagy at the Bauhaus school. Reviewing the sen-
Visualizing the data collected from multiple sen-
sory pedagogies of the early 20th century allowed us
sory modalities allows body-centered drawings to be
to rethink the methods used in current educational
compared with conventional drawings of space as
models in the design studio and formulate methods
one can overlay or put side by side drawings of ma-
for a multisensory learning of space.
terial boundaries of spaces and drawings of traces
Dedicated self-trackers utilize tracking technolo-
of bodily experiences of the same spaces. In other
gies to gain a better awareness of their bodily needs
words, representing the data in a visual can allow ar-
and behaviors, and artists utilize GPS and other

tracking devices to map urban space from a body- ACKNOWLEDGEMENTS
centered perspective. Briefly reviewing precedents The author would like to thank Professor Terry Knight,
in the arts and the aims of current self-tracking ten- Professor Takehiko Nagakura, and Professor Edith
dencies, the author proposed the incorporation of Ackermann for the discussions and advice on this
such approaches in design education to suggest work.
tools for sensory pedagogies, reframing the peda-
gogical methods proposed in the early 20th century. REFERENCES
Seymour Papert used the term objects-to-think- Cohen, R 1985, The Development of Spatial Cognition,
with to refer to technological tools that act as tools Psychology Press.
for thinking, enabling students to acquire concrete Gifford, R 2001, Environmental psychology: principles and
knowledge through a body syntonic, self-directed practice, Optimal Books, Colville
learning. Building upon Papert's terminology, while Johnson, M 1990, The Body in the Mind: The Bodily Basis
of Meaning, Imagination, and Reason, University of
stressing the sensory aspect of embodied knowl-
Chicago Press
edge and the use of sensing technologies, the author Millar, S 2008, Space and sense, Psycology Press/Taylor &
coined the term objects-to-sense-with to propose a Francis, New York
pedagogical role for sensory-tracking tools and of- Moholy-Nagy, L 2005, The New Vision: Fundamentals
fer a response to the need of sensory-based, body- of Bauhaus Design, Painting, Sculpture, and Architec-
centered learning in the architecture. ture, Dover Publications, New York
Montessori, M 1919, The Montessori Method, Frederic A
The author developed a wearable tool as an ex-
Strokes Company
ample of an object-to-sense-with, demonstrated the Noe, A 2006, Action in Perception, A Bradford Book, Cam-
results and discussed scenarios of use and possibil- bridge, Mass
ities for further development. The tool functions Nold, C 2009, Emotional Cartography: Technologies of the
as a generator of simple body-centric diagrams of Self, http://emotionalcartography.net
space that can be used to compare and contrast dif- Papadopoulou, A 2015, 'The Architect's Perspective on
the Tour and Map Perspective', Cognitive Processing,
ferent sensory interactions in the same space or the
16, pp. 331-336
sensory interactions of the same person in different Papert, S 1993, Mindstorms: Children, Computers, and
spaces. Through gathering real-time data regard- Powerful Ideas, Basic Books, New York
ing the users' location in space, their tactile interac- Winzer, M 1993, The History of Special Education: From
tions and the sound levels relative to their position in Isolation to Integration, Gallaudet University Press,
space, and by visualizing the data in Processing, users Washington, D.C
[1] http://journal.media-culture.org.au/index.php/m
are able to document their real-time body-space in-
cjournal/article/view/192
teractions in the physical spaces they explore. [2] http://www.nytimes.com/2010/05/02/magazine/02se
Objects-to-sense-with can be used in architec- lf-measurement-t.html?pagewanted=all&_r=0
ture to offer in-situ exploration of the built environ- [3] http://www.webcitation.org/66TEY49wv
ment that focuses on the senses. Different possi-
bilities can be explored for representing space in
multisensory and multimodal means through the
recorded data, and different scenarios can be em-
ployed to engage students in sensory-based learning
of space by allowing them to make and use their own
objects-to-sense-with as tools for self-directed, body-
centered learning.

'Starting with Nothing' and 'Ending up with Something'
Musical Improvisation and Parametric Spatial Design Improvisation
Jeremy Ham1 , Marc Aurel Schnabel2 , Lawrence Harvey3 ,

Daniel Prohasky4
1,3,4
RMIT University 2 Victoria University of Wellington
1
http://www.jjham.com 2 http://aurel.pro
3
http://www.sial.rmit.edu.au
1
jeremy@surfcoastarchitecture.com.au
2
marcaurel.schnabel@vuw.ac.nz
3,4
{ lawrence.harvey|daniel.prohasky}@rmit.edu.au
Musical improvisation is a complex field of the study of creativity wherein the

musician - according to jazz pianist Herbie Hancock - improvises by 'starting
with nothing' and 'ending up with something'. We explore a 'Musico-Spatial
Design' creative practice design research project that uses parametric spatial
design tools to research musical improvisation, then takes knowledge gained from
this process to illuminate aspects of improvisation within parametric spatial
design processes. Our investigation of musical and parametric spatial design
improvisation provides insights into how both novices and experts engage in
improvisation and how they don't really 'start with nothing' but bring into the
design conversation a varying range of referents to inform designerly intent.
Keywords: Parametric Spatial Design, Musical Improvisation, Scripting ,

Creative process
INTRODUCTION with split second accuracy and without the need for
At the opening concert for the 2015 Melbourne In- notation. As researchers operating at the intersection
ternational Jazz Festival, jazz pianist Herbie Hancock of music and spatial design, this statement serves as
introduced his improvised performance with Chick the starting point for a conversation on the practice
Corea by telling the audience how they would 'start of improvisation within both domains.
with nothing and hopefully end up with something'. In our research we examine a digital drumming
Hancock and Corea are renowned improvisers with a generative improvisation exercise and explore how
capacity to generate incredibly complex and beauti- musicians bring into improvisations a complex reper-
ful music in the moment, with years of performance toire of patterns and phrases that have been learnt,
and practice informing their capacity to make design- copied and adapted over a period of playing expe-
erly decisions on timing, timbre, attack and duration rience. We expand the examination of musical im-

provisation to initiate a discussion on the concept computational approaches to facilitate the produc-
of parametric spatial improvisation. This 'Musico- tion of 'architecture' from music through Music In-
Spatial Design' positioning provides a lens through strument Digital Interface (MIDI). This includes the
which to enable further understandings of architec- work of Ferschin et al. (2001), Levy (2003), Fowler
tural and spatial creative design processes. The in- (2011), Krawczyk (2012) and Tomara et al. (2011).
tention of this paper is to generate a conversation on Extended innovative creative investigations by No-
how musicians and spatial designers create through vak (2007) explores the translation of 'frozen music
improvisation and to serve as the basis for further re- into liquid architecture', that emancipates architec-
search. We question whether musicians and spatial ture from matter and music from sound. The work
designers 'start with nothing' with the aim of 'ending of Swiss architect Jan Henrik Hansen (2015) furthers
up with something'. these explorations in a practice context across a wide
Parallels between music and architecture (we use range of 'musical sculptures' through to building fa-
the term spatial design to enable a wider discus- cades over a number of years, providing a compre-
sion) have been widely examined throughout his- hensive body of translations from one medium to the
tory, and can be conceptualised on three levels: other across scales and media.
through process- the means of inventing and real-
ising, product- the outcome of these processes and
Figure 1
experience- the human interaction with the out-
Bells Beach House
come. Goodman (1974) provides an account of the
by SurfCoast
comparisons between music composition and archi-
Architecture.
tecture as 'allographic arts' that require notation in
the form of plan or score in order to be realised. We
address this by examining improvisation without no-
tation (music) and improvisation through notation
(parametric design).
Elizabeth Martin (1994) in 'Architecture as a
Translation of Music' describes the 'y-condition' as
'the middle position of music and architecture when
translating one to another finding an organic union
between the two'. Musico-Spatial Design connec-
tions have manifested in many ways over the years:
most famously with the collaboration between Ian-
nis Xenakis, Le Corbusier and Edgar Varese for the
1958 Brussels World Fair. This built creative collabo-
ration integrated electroacoustic music and architec-
tural space in the form of a 'poeme electronique' (Xe-
nakis and Kanach 2008). Martin's 'y-condition is ex-
pressed by Steven Holl (1994) in the Stretto House- a
built interpretation of Bela Bartok's Music for Strings
Percussion and Celeste. These built works provide
two contrasting ways in which music and architec-
ture can be combined in built form.
The advent of design computing has enabled

The authors' own investigations between the disci- Jazz, Flamenco, Indian Tabla music, Hip Hop, Rock
plines of music and architecture include the devel- and Blues (Bailey 1993). Composition differs from
opment of a model of virtual architecture generated improvisation in that the outcome is pre-conceived
from the music of J. S. Bach (Christensen and Schn- (usually through the language of notation) prior to
abel 2008) and the development of an integrated the performance; however given the many modali-
design-build design studio over several years involv- ties of musical performance (including digital modal-
ing design students composing music, designing a ities including live coding), the lines between com-
'Music Room', building these at real scale and retain- position and improvisation are often blurred.
ing work on an on-line Virtual Gallery (Ham 2003, Composition and improvisation form a living sys-
2005). tem whereby chaos and order coexist, with chaos
The first authors' own house is an outcome of evolving from order through disruptive improvisa-
a design process that references musical elements tions, resulting in a reordering of the complex sys-
such as polyrhythmic material pattern overlays and tem of musical dialogue. This does not neatly fit into
metaphorical expressions of instrumentation and the 'levels of improvisation' as proposed by Benson
golden section geometries whilst operating within (2003) but operates within the complex system of
the pragmatic constraints of planning and building Cobussen's Field of Musical Improvisation (Cobussen
regulations (Figure 1). Given the significant work un- 2010) founded indeterminism, imperfection and un-
dertaken in architectural and spatial manifestations predictability Kleidonas (2012).
of music, there exists a gap in knowledge in the area Brown (2006) connects music and architecture
of developing understandings of similarities in the through the study of jazz and improvised music to
creative processes of both domains. This area of 'pro- contribute 'insight regarding the human creative po-
cess' forms the basis of this paper. tential that can be enabled in the interplay of dy-
namic systems: filled with risk, diverse durations,
MUSICAL IMPROVISATION AND DESIGN fleeting boundaries and varying qualities of media'.
He describes improvisation as, 'borrowing from the
PROCESS
language of the biological and the computational,
Our Musico-Spatial Design theoretical positioning
best understood as an emergence- one of human
uses parametric spatial design tools to examine musi-
thought and action: that is facilitated by a range of
cal improvisation and then takes knowledge gained
variables, properties and structures'. The degree, and
from this process to illuminate aspects of paramet-
success of compositional and improvisational output
ric spatial design processes. Focussing on the initial
is a function of the ability of the creative practitioner
stages of creativity we are interested in how both mu-
to engage in the creative parameters within which
sicians and spatial designers 'start with nothing' and
they work as defined by the opportunities and con-
'end up with something'. The 'nothing' is the drum-
straints of the interface between idea and artefact.
mer sitting on the drum stool, sticks in hand: the
Improvisation requires the instantaneous musical re-
parametric designer on computer with empty screen
sponse to an internal or external stimulus within the
(perhaps with pencil in hand). The 'something' is an
range of musical choices of frequency, time and dy-
improvised drum solo, or the first iteration and out-
namics Malinowski (2011) as determined by the in-
come of a parametric script respectively.
strument.
Both musicians and spatial designers improvise
as part of their creative practice. Improvisation is
generally regarded as a process of 'in the moment' ex-
temporisation of musical creativity which is evident
in a range of cultures and musical styles including

GENERATING COMPLEXITY THROUGH improvisation over a set template will be significantly
SOLO DRUMMING IMPROVISATION different. The creative approach of architects such as
We base this research on the creative practice of dig- Mies van der Rohe, Frank Gehry and Zaha Hadid to a
ital drumming within a merged Musico-Spatial De- set task will also be markedly different. These differ-
sign creative practice PhD by project work at the ences are founded on philosophical and stylistic dif-
Spatial Information Architecture Laboratory (SIAL) at ferences evolved through creative practice over time.
RMIT University. Project work involved the perfor- Drum improvisation in a very physical practice
mance by the first author (an experienced musician that requires percussive striking with hands (sticks)
and architect) of a series of 80 drum solos on a Roland and feet on the interface of the drum kit. The foun-
TD20 digital drum kit templated to 100 beats per dation of drumming is the physical production of of-
minute for 100 beats (Figure 2). Other improvisa- ten complex overlays of polyrhythms over a temporal
tions were performed across the contexts of Beat and structure. Using hands and feet, the drummer phys-
Fill (playing drum beats) and playing improvisations ically percusses on a resonant surface to produce
over a set guitar track. This generative exercise was a sound. Virtuoso drummers such as Terry Bozzio
designed to test the scope, structure, shape, form have an advanced capacity to produce 4-way com-
and details of improvisational capacity, using MIDI plex polyrhythmic overlays from all limbs- essentially
format to enable computational analysis. Drumming playing four things at once. Although this may op-
is perhaps an under-recognised creative art form that erate well as a mathematical exercise, the key to the
is gaining traction in research through, for example, master drummer is building skill levels to serve a mu-
renowned King Crimson and Yes drummer Bill Bru- sical intention or purpose. Referring to the opening
ford's PhD on the creative aspects of the western statement, the expert drummer, in performing im-
drum kit player (Bruford 2015). Drummers, like archi- provised drum solos does not really 'start with noth-
tects and spatial designers, each evolve unique ap- ing' but brings into each improvisation a combina-
proaches to design solutions. The creative approach tion of referent patterns and phrases (The 'go-to' riffs
of drummers like Buddy Rich, Jo Jones, Bill Bruford, invented, learnt, borrowed and adapted that consti-
Shiela Escovedo, Terry Bozzio and Thomas Prigden to tute style) which are informed by a range of factors,
Figure 2
Model of Drum solo
in Plan, Elevation
and Isometric.

environmental conditions, thoughts, feelings and in- analysis that includes the integration of many forms
tentions that tacitly inform the drummer's response of spatialized representations with tagged descrip-
Improvisation is thus founded on Schön (1983) con- tors, culminating in the use of 3D printed represen-
cept of 'tacit knowing in action' through the recall tations of referent patterns and phrases (Figures 4
and implementation 'in the moment' of a repertoire and 5). This drummer's own representation of im-
of 'referent' (Pressing 1987) patterns and phrases that provised drumming allows for the analysis of micro-
have evolved through played experience. timing, playing behind and ahead of the beat and
In an effort to define this repertoire of 'referent' the variations in velocity that defines individual style.
drum patterns and phrases, the collection of drum The wider research project involves using this para-
solo outputs were curated, with a representative set metric framework to analyse how different drummer
of 100 context-specific referent patterns and phrases improvise over a set musical piece. Drummers thus
extracted in MIDI, then analysed in relation to length, have their own signature elements of style that, when
style, beat type, quality of play, complexity and a brought into the spatial domain, can be compared.
'referent rating'. Using Rhinoceros3D Version 5 with
Figure 3 Grasshopper (GH) the 'ImprovScope' script was built
Roland TD20 Digital that enables spatialization of referent drum patterns
Drum kit as and phrases in plan, section, elevation and 3D (Figure
Instrument and 3). MIDI data was translated into .csv format, record-
Interface. ing tempo, 'drum note', 'velocity', 'note on' and 'note
off' over time to 1/1000ths of a beat accuracy. These
data were read into GH, separated and sorted using
standard GH components to result in the generation
of a series of points in space for each drum note with
velocity and note length data attached.
The principal of this translation is the use of spa-
tial design software to unlock the complexities of the
process drum based improvisation and not to repro-
Figure 4 duce concepts of 'frozen music'. We propose that by
GH Spatialization & translating these improvisations into the spatial do-
Tagging of Drum main, a theoretical 'musico-perspectival hinge' is un-
Pattern based on locked (referring to Pelletier and Pérez-Gómez (1997)
Symbolic perspectival hinge) thus enabling spatially-dominant
Representation of people (e.g. architects) insights into understanding
Drum Kit. the 'infinite art of improvisation' (Berliner 2009). The
GH script overcomes problems inherent in traditional
Figure 5 notation and founds representations on a spiral sym-
Lofted bolic representation of the drum kit as viewed from a
Representation of players perspective. The Snare dum forms the centre,
Improvised Drum with bass drum to the bottom, with arrays of Hi-Hats,
Events in Time. and Cymbals on the outer spiral and Tom Toms radi-
ating around the inner spiral. Velocity of hits are rep-
resented, in this spatialization, by the radius of mesh
spheres. The GH script allows multiple modes of

The eighty drum solo improvisations form a gener- many conservative design practices recycling a well-
ative basis for the creation of a complex data set of established set of design solutions from project to
multi-layered polyrhythms through the interface of project that are known to work.
the digital drum kit. When overlaid onto eachother The core intention of many parametric design
in MIDI, patterns emerge regarding clustering of processes is to push boundaries - as one can see
drum notes, combinations of drum patterns includ- easily from the CumInCAD Index, which is full of
ing flams, closed and open rolls, combinations and complex, innovative and difficult projects brought
the other elements that constitute drumming style into fruition through parametric design processes.
(Figure 6). Each of these improvisations 'started with For example the Sagrada Familia project is one that
nothing' and 'ended up with something' in a musi- has driven innovation through parametric design
cal sense, however each solo was in some way pre- by the need to solve complex geometric problems
conceived through muscle memory and complex spatially and materially across a geographically dis-
cognitive processes and neural transmission where persed team (Burry 1996). In this paper, we focus
'auditory stimulus activity reaches the cerebral cor- on the initial starting point of design, as this starting
tex 8-9 millisecond after stimulation' (Pressing 1987). point is where comparisons with improvisation can
The musical intention may well 'start with (close to) be the closest.
nothing' however evolves as a necessity during live Parametric spatial improvisation, unlike musical
play. This is because the urgency of live play re- improvisation, is facilitated through notation in the
quires immediate action and reaction with little or no form of scripting. The research described above out-
time for reflection-on-action to inform reflection-in- lines an alternative method of notating improvisa-
action. We acknowledge the limitations of this re- tions so that the notation follows the improvisation.
search, however see significant potential in adding to 'Script' is derived from the term for written dialogue
the present framework to include body mapping and in the performing arts, where actors are given direc-
the analysis of brain activity in improvisational activ- tions to speak, perform or interpret. Subsequently,
ity using Emotiv EEG Brainware-apparatus in future 'scripting' is a creative process that describes the
project work. artistic intent of the designer. Scripts can define a
set of rules that perform a set of defined operations
PARAMETRIC SPATIAL DESIGN IMPROVI- on a set of parameters. Software applications can be
programmed and adjusted by scripts allowing for ex-
SATION
ample repetitive tasks to be automated or to gener-
We preface this conversation by stating some of the
ate solutions that fit to a range of parameters (Bilo-
key differences between musical improvisation and
ria 2005). Thus, the lines between composition and
spatial design improvisation. Fundamentally, impro-
improvisation, notation and 'live play' are blurred in
vised music is performed (played) in the moment
parametric design.
meaning the time for critical-path decision making
Thus notation forms the foundation of paramet-
is many times quicker than spatial design. From the
ric design, just as it forms the foundation of musical
outset, free improvisation in spatial design appears
composition. A descriptor of geometry or a virtual
to be more difficult than music. This domain is cen-
geometric elements - each of these have mathemati-
tred around the resolution of a design brief founded
cal relations with a global coordinate system and/or a
on an often-complex array of spatial, financial, struc-
mathematic function which describes the geometry -
tural, servicing, temporal and resourcing constraints.
just as a note is located within a stave (global coordi-
Real risks for spatial designers are posed where the
nate system). The rules between geometric elements
outcome of spatial design decisions may cause loss-
that define new elements of relational geometry mir-
of time, money or even life. This has resulted in

ror the relation between one note/drum beat with re- setting up scripts to allow the freedom of improvisa-
spect to the next - and so on. These 'rules' or musical tion in the production of multiple design outcomes
relations that link notes/elements of geometry is the with a minimum of effort. One moves from simplicity
'medium' or 'skeleton' for musico-spatial parametric of operation to complexity in processing, the other
design - hence - parametric design starts with some- from complexity of parameterisation to simplicity of
thing - that something is notation. generation of multiple spatial alternatives. Although
Instead of using only compositional methods for parametric improvisation in architecture does not
designing, designers utilise scripts to form their own lead directly to a resolved design, yet the interdisci-
generative properties and base for their design im- plinary methodology can be accepted as a valuable
provisation. Developing and amending scripts - de- approach for infusing innovative ideas into overall
signers can quickly improvise and control their ideas design process. The computational processes offer
by amending the parameters or rules to fit their de- the ability to empower designers to delve into other
sign intent (Celani 2008). Thus, there exists a role fields. Allowing for a unique interpretation, represen-
reversal between music and parametric spatial de- tation through the improvisation process that gener-
sign improvisation. The improvising drummer can ates the resulting artifices (Figure 7) (van Velthooven
improvise freely to produce the initial musical output and Schnabel 2016).
quickly and efficiently, only later to be slowed down Thus, significant 'embodied effort' (Ham 2004) is
in mixing, sampling and processing of sound through required to deliver initial representations and to re-
the Digital Audio Workstation (DAW) (Figure 6). solve the complex, wicked and ill-defined problems
The parametric designer however must work (Rittel and Webber 1973) inherent in spatial design.
through the constraints of analysing parameters and This has parallels with the research on representa-
Figure 6
Processing
Interface: Reaper
DAW with 80 Drum
Solo's Overlaid in
MIDI Piano Roll
Format.

Figure 7
Improvisations
using a variety of
scripts to generate
hyperlocalized
architectural forms
(van Velthooven
and Schnabel 2016).
tional media by Bermudez and King (2000) wherein may well 'start with nothing' but a blank screen and
they found that 'analogue representations are far an empty pallet of referents.
more fluid and appropriate than digital media for ini- The development of a repository of parametric
tial and fast development of ideas, the stimulation referents acts to enable efficiencies in parametric de-
of the imagination, free inquiry, the intentional and sign processes. These referents, in the form of el-
random cross-reference of diverse sources', whereas ements of scripts, patches and code can be saved
'digital media are stronger for design development as and recalled digitally, then shared online thus cre-
they demand higher levels of geometrical definition ating hybridized multi-user referents (much in the
and abstraction'. For digital natives and expert users, same way as a band has its trademark 'licks'). Expert
mediating digital and physical media is natural, how- scripters, like expert drummers, can recall, cut, copy
ever the issue arises as to how novice users can best these with speed, accuracy and precision. The key is,
engage in improvisation. For them, the pencil and to refer again to Bermudez and King (2000), to de-
sketch pad may be the ideal spatial design improvi- velop a hybrid media process wherein each medium
sation tools- not the computer. Maybe, for novice facilitates the free flow of design. In both music and
spatial design improvisers, Tom Kvan was right: the spatial design, expert users find this natural and easy
computer is the problem (Kvan 2004) as novice users to the point of their skill and knowledge limitations,

whereas novices struggle. the digital drum kit as a means of generating impro-
A fundamental element of improvisation is the visational output that is curated and translated into
ability to 'play'. The problem is, 'as parametric models spatial form using parametric design processes. In-
exceed levels of complexity that go from toy systems sights into the 'infinite art of improvisation' brings
to the real world practice problems, they become forward the issue of how parametric design tools
increasingly inflexible' Pena de Leon (2014). When such as GH enhance or obstruct the ability of spatial
'starting with nothing', this inflexibility (due to com- designers to improvise. The exploration of the gestalt
plexity) provides barriers for non-expert users to en- within both environments can enhance the under-
gage in the potentials of parametric design tools and standing of issues relevant to the field and lead to
represents a hurdle even for expert designers. Play- meaningful and new results (Schnabel 2005).
ing with toys is thus different to playing with complex We find that expert musicians and spatial design-
systems and play by novices is different to play by ex- ers don't 'start with nothing', however they start with
perts. Parametric improvisation subsequently does something in order to create something else. Expert
not solve all issues connected with the process of ar- musicians like Herbie Hancock and Chick Corea, and
chitectural design and may even cause other issues. also expert parametric designers start their impro-
It allows however, an alignment of cause and effect visation drawing on a rich vein of knowledge cen-
and a reflection of the design intent, its processes and tred on their domain that has been learnt, copied and
to some extent - its outcome (Schnabel 2012). adapted over their creative practice. The key to para-
metric design, perhaps, is to borrow from musical im-
CONCLUSIONS provisation and find ways to best enable the free flow
The research we have presented explores the idea of ideas through play.
of improvisation within a Musico-Spatial Design per-
spective wherein one domain forms a lens through REFERENCES
which to examine the other. Founding this paper Bailey, D 1993, Improvisation: Its Nature and Practice in
on a statement by a virtuoso jazz improviser (Herbie Music, Prentice-Hall, Englewood Cliffs, New Jersey
Hancock), we find that both drummers and spatial Benson, BE 2003, The Improvisation of Musical Dialogue:
a Phenomenology of Music, Cambridge University
designers do not really 'start with nothing' when they
Press, Cambridge
improvise, but bring into the process a repertoire of Berliner, PF 2009, Thinking in Jazz: The Infinite Art of Im-
patterns and phrases that have been learnt, copied provisation, University of Chicago Press, Chicago
and adapted over the period of creative practice. The Bermudez, J and King, K 2000, 'Media Interaction and
founding quote by Herbie Hancock was followed by Design Process: Establishing a Knowledge Base', Au-
a two-hour concert whereby he and Chick Corea im- tomation in Construction, 9(1), pp. 37-56
Biloria, N, Oosterhuis, K and Aalbers, C 2005 'Design In-
provised freely and fluently over rehearsed and un-
formatics. Smart Architecture: Integration of Digital
rehearsed material in a highly complex dynamic in- and Building Technologies', Smart Architecture: Inte-
terplay using a wide range of trademark 'referent' pi- gration of Digital and Building Technologies, Proceed-
ano riffs as the foundation. This is the domain of the ings of the 2005 Annual Conference of the Association
expert, the virtuoso, the 'practiced digital hand' (Mc- for Computer Aided Design In Architecture, Savannah,
Cullough 1998) that most creatives aspire to. Reach- Georgia, pp. 226-235
Brown, D 2006, Noise Orders: Jazz, Improvisation and Ar-
ing the stage of musical and spatial design virtuos-
chitecture, University of Minnesota Press, Minneapo-
ity takes years of practice, repetition, learning, copy- lis
ing and adapting and evolving an individual sense of Bruford, W 2015, Making it Work: Creative Music Perfor-
style. mance and the Western Kit Drummer, Ph.D. Thesis,
A creative practice project is described that uses University of Surrey

Burry, M 1996, 'Parametric design and the Sagrada Fa- Levy, AJ 2003, 'Real and Virtual Spaces Generated By Mu-
milia', Architectural Research Quarterly, 1(04), pp. 70- sic', International Journal of Architectural Computing
81 (IJAC), 1(3), pp. 375-391
Celani, MGC 2008, 'Teaching CAD Programming to Archi- Malinowski, S 2011 'Music Animation Machine', Music An-
tecture Students.', Revista Gestão & Tecnologia de Pro- imation Machine, http://www.musanim.com
jetos, 3(2), pp. 1-23 Martin, E 1994, Pamphlet Architecture 16: Architecture as
Christensen, P and Schnabel, MA 2008 'Spatial a translation of music, Princeton Architectural Press,
Polyphony: Virtual Architecture Generated from New York
the Music of J.S. Bach', Proceedings of the 13th Inter- McCullough, M 1998, Abstracting craft: The practiced dig-
national Conference on Computer Aided Architectural ital hand, MIT press
Design Research in Asia, Chiang Mai, Thailand, pp. Novak, M 2007 'The Music of Architecture: Compu-
501-509 tation and Composition', Media Arts and Tech-
Cobussen, M 2010 'Improvisation as a complex system', nology, University of California, Santa Barbara,
Thinking Sounds, http://www.socasites.qub.ac.uk/ http://www.mat.ucsb.edu/˜marcos/
fschroeder/symposium/2010/Videos/ TheMusicOfArchitecture.pdf, p. 22
Symposium-12.mov Pelletier, L and Pérez-Gómez, A 1997, Architectural Repre-
Ferschin, P, Lehner, D and Oka, Y 2001 'Translating music sentation and the Perspective Hinge, MIT Press, Cam-
into architecture', The third International Mathemat- bridge, Massachusetts and London
ics and Design Conference M&D2001: digital, hand, Pressing, J 1987, 'Improvisation: Methods and Models', in
eye, ear, mind, Geelong, Australia Sloboda, J (eds) 1987, Generative Processes in Music:
Fowler, M 2011, 'Appropriating an Architectural Design The Psychology of Performance, Improvisation, and
Tool for Musical Ends', Digital Creativity, 22(4), pp. Composition, Oxford University Press, Oxford, UK,
275-287 pp. 129-178
Ham, JJ 2003 'Architect as composer: engaging architec- Rittel, HW and Webber, MM 1973, '2.3 Planning Problems
ture students with music in the design studio', De- are Wicked', Polity, 4, pp. 155-169
sign+ Research: project based research in architecture, Schnabel, MA 2012, 'Learning Parametric Designing', in
Melbourne, Australia, pp. 1-10 Gu, N and Wang, X (eds) 2012, Computational Design
Ham, JJ 2004, On Representation and Construction in a Methods and Technologies: Applications in CAD, CAM
Tectonic Design Studio, Master's Thesis, Deakin Uni- and CAE Education, IGI Global, Hershey, pp. 56-70
versity Schnabel, MA 2005, 'Interplay of Domains', in Martens,
Ham, JJ 2005 'Music and architecture: from digital com- B (eds) 2005, Learning from the Past Foundation for
position to physical artifact', Digital design: the quest the Future, Oesterreichischer Kunst- und Kulturver-
for new paradigms, Proceedings of the 23rd Confer- lag, Vienna, pp. 11-20
ence on Education in Computer Aided Architectural Schön, DA 1983, The reflective practitioner: How profes-
Design in Europe, Lisbon, Portugal, pp. 139-146 sionals think in action, Basic Books, New York
Hansen, JH 2015 'Form follows Funk', Studio Jan Henrik Tomara, A, Liapi, M and Parthenios, P 2011 'Translating
Hansen, www.formfollowsfunk.com Urban Environment to Music', Proceedings of the 15th
Holl, S 1994, 'Stretto House', Pamphlet Architecture: Archi- Iberoamerican Congress of Digital Graphics, Santa Fe,
tecture as a Translation of Music, 16, pp. 56-59 Argentina, pp. 530-533
Kleidonas, A 2012 'Lessons in Architectural Improvisa- Van Velthooven, C and Schnabel, MA 2016 'Hyperlocal-
tion', First International Conference on Architecture ization through Architecture and Climatology', Par-
and Urban Design, Tirana, Albania, pp. 1-10 allelism in Architecture, Environment And Computing
Krawczyk, RJ 2012 'Exploring the Visualization of Music', Techniques, (PACT), London, pp. 1-10
Proceedings of Bridges 2012: Mathematics, Music, Art, Xenakis, I, Kanach, S and Lieberman, DJ 2008, Music and
Architecture, Culture, Baltimore, USA, pp. 393-396 Architecture: Architectural Projects, Texts, and Realiza-
Kvan, T 2004, 'Reasons to stop teaching CAAD', in Chiu, tions, Pendragon Press, Hillsdale, NY
ML (eds) 2004, Digital Design Education, Garden City
Publishing, Taipei, Taiwan, pp. 66-81
Pena de Leon, A 2014, Separation of Concerns: Strategies
for complex parametric design modelling, Ph.D. The-
sis, RMIT University

Understanding the Verbal Concepts Appropriated by the
Students in the Architectural Design Studio
Burak Pak1 , Zeynep Aydemir2
1
KU Leuven Faculty of Architecture 2 ITU Faculty of Architecture
1
burak.pak@kuleuven.be 2 aydemira@itu.edu.tr
The main aim of this study is to gain a better understanding of the role of verbal
concepts in the architectural design processes of the students in a studio context.
To serve this purpose, we carried out a 15-week studio in an urban architectural
masters design studio at KU Leuven Faculty of Architecture. We observed the use
of verbal concepts in time during this studio and analyzed the design processes of
the students based on their self-report logs on the studio web platform. Based on
these, we conducted a statistic analysis and a network mapping study. We found
that early concepts provide a starting point for developing fully-fledged
specialized design ideas. Furthermore, a higher number of links between
concepts indicated their importance during the process. In addition, the data
collection and research methods proved to be reliable for mapping the design
process of the students as well as revealing the evolution of the ideas in the studio.
Keywords: Design Studio, Concepts, Crowdsourcing, Web Platform,

Self-reporting, Design Research
INTRODUCTION ent types of these concepts play a role in various

In this study, we use the term "verbal concept" to phases of design and design studio learning.
describe "verbal units of information that design- Motivated by these observations, we aim to de-
ers stress as important" during the design process velop a better understanding of this topic and specif-
(Schön and Rein 1994). According to Schön, design- ically address the following questions:
ers impose a network of 'names' on to the design task
• How can we research the verbal concepts ap-
and the design situation which helps them to frame
propriated by the students during the archi-
the design task and solution (Dorst 1997). Examples
tectural design studio?
of these names (or verbal concepts) for a particular
• Can verbal concept maps and statistics of self
design process is: 'spatial identity, social interaction,
report logs reveal knowledge on these? and
informal networks, and dislocation' (Schön and Rein
how are they related to the design processes
1994).
of the students?
In literature, it is possible to find a plethora of ref-
erences to verbal concepts used in design and their Section 1 will introduce a background review of ver-
possible attributes. However, there are only a few bal concepts in design process. In the following sec-
number of empirical research studies on how differ- tion (Section 2), we will introduce the method of the

study and the hypotheses. Section 3 will reveal the follow (Cross 2006, 34). Initial concepts are not nec-
analysis results from the studio work we carried out essarily consistent, they are alternatively named as
during the Spring 2015 semester. In conclusion (Sec- 'early concepts' that have an effect on the quality of
tion 4), we will discuss the findings and make sugges- design solutions; and successful early concepts do
tions for future research on the role of verbal concep- not necessitate alternative concepts for further de-
tual representations in architectural design and development (Cross 2006, 06).
sign learning. Goldschmidt and Tatsa made an empirical study
on leading concepts (2005). They carried out a de-
BACKGROUND sign studio experiment in which data was collected
In design research, the term 'concept' has more than by an experimenter throughout the semester. They
one meaning. First of all, a 'design concept' may re- inspected students' conceptual ideas quantitatively
fer to a complex construction incorporating mental in the course of development of their term projects
units; an assemblage that gathers the intricacies of and tried to establish possible correlations between
the project as a collection of responses to a particu- the quality of final projects and various parameters
lar problem (Teal 2010, p.301). In this context, a 'de- related to their design ideas. The study shows that
sign concept' is an idea that drives many of the major good ideas can be identified, counted and charac-
preliminary design decisions (Goldschmidt and Sever terized by looking at the links they generate. Lead-
2011, p.139). This idea embodies several features in ing concepts can be associated with good ideas as
the form of an early design defined as representa- they are described as useful concepts that lead good
tions which are essential for developing the concept designs (Goldschmidt and Tatsa 2005, p.603). Ac-
into a fully-fledged design (Cross 2006, p.66). cording to this study, leading design ideas have a
Alternatively, the term 'concept' may refer to higher number of links that are formed among ideas
units of information that designers stress as impor- or decisions. This study suggests that a higher num-
tant during the design process. From a problem solv- ber of links indicate the criticality and importance of
ing perspective, these are 'verbal conceptual repre- the ideas or concepts (Goldschmidt and Tatsa 2005,
sentations' that play an essential role in establishing p.606).
relationships for transforming physical or social rela-
tions into a functional idea (Akin 1986, p.75). Schön RESEARCH DESIGN
(1994, p.26) calls these 'names' or features and re- In order to respond the research questions, a 'living
lations that become the things of a story. Accord- laboratory' was set up in an International Urban Ar-
ing to Schön, designers impose a network of 'names' chitectural Design Master's Studio at KU Leuven Uni-
onto the design task and the design situation which versity, Faculty of Architecture in Brussels during 15
helps them to frame the design task and solution weeks of 2014/2015 spring semester. The partici-
(Dorst 1997). Examples of these verbal concepts were pants were 43 international architecture students (25
identified by Schön in a specific design study as: females and 18 males) in their first or second year of
'spatial identity, social interaction, informal networks master studies.
and dislocation'. In literature, it is possible to find a A web-based research platform for augmenting
plethora of references to verbal and non-verbal con- design learning (Pak and Verbeke, 2012)(Pak and Ver-
cepts used in design and their possible attributes. beke 2014) was set up for collecting information from
According to Waldron and Waldron (1988), ini- the students, which also included the studio brief,
tial concepts are related to the early design phase, schedule, bibliography, program, announcements,
they provide a necessary starting point and could be communication and assignment submissions.
partly fallacious or critical to the procedures that will Every two weeks, an assignment challenge was

Figure 1
Submission of
'diversity-
uniformity' group
for the first group
assignment.
announced which required the students to submit included in the count except the last one.
their work through the online platform. With ev- Identifying and Coding Novel and Leading Con-
ery submission, they were asked to provide up to cepts. Novel concepts are the concepts that are used
five keywords (open question, e.g. connection, infor the first time in a student's submission. They are
tegration, productive) and choose at least one of counted as a novel concept only when they are first
the ten concept-couples suggested by the groups at used. For instance, 'diversity-uniformity' was identi-
the beginning of the semester (multiple choice, e.g. fied as a novel concept when it appeared in the first
density-uniformity, frontstage-sideline, foundspace- submission and categorized as a leading concept for
sterilespace). By doing so, students created self- the second submission.
reporting logs (concepts) on the platform as design
Identifying and Coding Unique Concepts. A
diaries which formed a basis for the source data for
unique concept is used for only one time during the
our research.
semester in the entire studio.
After a review of concept categories in literature,
After documenting all the concepts that are
we decided to focus on two types of concepts: lead-
used, we determined the quantity of the unique con-
ing and unique concepts. A repeating concept is
cepts and their distribution among students. Based
counted as a leading concept in all submissions and

on our literature review we tested six hypotheses. We made a statistical analysis of the novel and unique
Since we have limited space in this conference paper, concepts and their several correlations to examine
we will report only two of them: the hypotheses. To test the hypotheses(1 and 2), we
used Cytoscape software for visualizing network rela-
• H1 Early concepts provide a starting point
tions of concepts suggested by the students.
for developing fully-fledged specialized de-
sign ideas (Cross 2006, p.106) Hypothesis H1: Early concepts provide a starting
• H2 Higher number of links between concepts point for developing fully-fledged specialized de-
indicate the criticality and importance of the sign ideas. In order to test the hypothesis of Cross
ideas or concepts (Goldschmidt and Tatsa (2006), we analyzed how students describe the de-
2005) velopment of their ideas with verbal concepts dur-
ing the semester. We linked nodes of concepts that
are reported by each student to the group concepts
ANALYSIS RESULTS with edges in Cytoscape software to create a concept
With the accumulation of design diaries on the web-
map. Cytoscape is a software designed initially for
based research platform, we collected every verbal
integrated models of biomolecular interaction net-
conceptual representation of design ideas and ana-
works. Within this software environment, molecu-
lyzed their correlations in the context of their nov-
lar species were represented as nodes and their in-
elty, uniqueness, lead, and quantitative limitations.
teractions were represented as links, namely, edges
During the 15 weeks of the studio semester, 479 con-
between nodes (Shannon et al. 2003, 2499). In
cepts were reported through 9 assignment submis-
our study, each verbal concept was represented as
sions. 10 of them were concept- couples derived
a node and linked concepts were represented with
from students' pin-up presentations and suggested
edges in between. For instance, concepts that are re-
by the instructors, and 469 of them self-reported by
ported together for the same assignment were rep-
the students. 240 of 469 were unique concepts re-
resented as nodes and linked with edges; whereas
ported only once. The three most used concepts
concepts that are reported later, but linked to a pre-
were 'connection, integration and productive' (Table
viously reported concept were also represented as
1).
Table 1
The 10 most
frequently used
concepts by the
students.

Table 2
Analysis of a
student's most
concept links based
on edge
betweenness
centrality: the
number of the
shortest paths that
go through an edge
in the whole
network.
nodes and linked with edges to a former concept. gorithm with the following parameters: spring con-
Therefore, all related concepts are linked with edges stant (0.003), compatibility threshold (0.3), the num-
and repeated links are represented thicker. ber of handles (3), maximum iterations (100). With
Cytoscape software provides network analysis a network analysis, we achieved the rates of links be-
function as well as a variety of visual representa- tween nodes. Figure 2 shows that concepts that have
tions of the integrated data. With a network analy- higher rates of connection and more links are gath-
sis we achieved the edge betweenness rates (Girvan ered at the center, whereas concepts with fewer links
and Newman 2002) of links between nodes (Table 2). are in the periphery. Additionally, high rated links are
The concepts 'front stage-sideline' and 'found space- darker and thicker. Concepts with high rated links are
sterile space' are the highest rated concepts in terms used by more students and lead the main studio dis-
of their links; and they are concepts reported for the cussions.
first and third assignments. These results show that
final designs of the students use early concepts as a CONCLUSIONS
basis. We used the setting of an International Urban Archi-
Hypothesis H2: Higher number of links between tectural Design Master's Studio as a living laboratory
concepts indicate the criticality and importance and set an online research platform for collecting self-
of the ideas or concepts. In order to test the hypoth- reports of students as an objective data. We evalu-
esis of Goldschmidt and Tatsa (2005), we analyzed ated the keywords of assignments submitted every
all concepts used in the studio and their links dur- two weeks during the semester as verbal conceptual
ing the semester. We linked nodes of concepts that representations of their works.
are reported together in the same submission (thus We analyzed verbal conceptual representations
relating to a design stage) with edges in Cytoscape in design learning based on two hypotheses (Table 3).
software. Edges are bundled using the built-in al- Our results indicate that it is possible to follow the im-

Figure 2
Concept map of the
whole studio: A
force-directed
graph drawing,
generated through
Cytoscape
Software, using an
SQL query from the
database of the
web platform
containing student
reports.

Table 3
The overall
evaluation of
hypothesis tests.
plications of verbal concepts in the students' design method for collecting, visualizing and analyzing the
processes, and they demonstrate certain patterns: relations between the design concepts.
1. Our network analysis (edge-betweenness

centrality values) indicated that early con-
REFERENCES
Akin, O 1986, Psychology of Architectural Design, Pion
cepts acted as a central reference to the whole Limited
studio. Akin, O and Lin, C 1995, 'Design protocol data and novel
2. Early concepts provided a starting point for design decisions', Design Studies, 16(2), pp. 211-236
developing fully-fledged specialized design Cross, N 2006, Designerly Ways of Knowing, Springer, Lon-
ideas. For each student, the traces of the early don
concepts were visible in the final verbal logs Darke, J 1979, 'The primary generator and the design
process', Design Studies, 1(1), pp. 36-44
of the students. Dorst, K 1997, Describing design: a comparison of
3. The most referenced verbal concepts by the paradigms (Doctoral dissertation), Ph.D. Thesis, TU
students were critical and led the main discus- Delft
sions in studio. Finke, RA, Ward, TB and Smith, SM 1992, Creative Cog-
nition: Theory, Research, and Applications, MIT Press
By using a web-based research platform to collect Cambridge, Massachusetts
self-reports of the students, it was possible to gener- Girvan, M and Newman, ME 2002, 'Community struc-
ate concept maps and statistics which helped to un- ture in social and biological networks.', Proceedings
of the National Academy of Sciences USA, 99(12), p.
derstand how students think conceptually and how 7821–7826
concepts contribute to the design process. Goldschmidt, G 1994, 'On visual design thinking: the vis
Furthermore, in the studio, we observed the im- kids of architecture', Design Studies, 15(2), pp. 158-
plications of verbal conceptual representations to de- 174
sign processes. We found that the verbal concepts in- Goldschmidt, G 2001 'Is a figure-concept binary argu-
troduced in the early stages of the design studio play mentation pattern inherent in visual design reason-
ing', Proceedings of the Second International Confer-
a crucial role in the design processes of the students. ence on Visual and Spatial Reasoning in Design Com-
From a pedagogical standpoint, we believe that dis- putational and Cognitive Approaches, pp. 177-205
covering their effects on different phases can help to Goldschmidt, G and Sever, AL 2011, 'Inspiring design
initiate and trigger new ideas to develop better de- ideas with texts', Design Studies, 32(2), pp. 139-155
signs in a studio setting. Our research is limited in Goldschmidt, G and Tatsa, D 2005, 'How good are good
scale and it is clear that there is a need for more in- ideas?', Correlates of design creativity. Design Studies,
26(6), pp. 593-611
depth research to reveal these effects. We hope that Maccoby, M 1991, 'The innovative mind at work', Spec-
this study will be helpful to designers and researchers trum, IEEE, 28(12), pp. 23-35
as a basis for understanding the role of the verbal Miller, GA 1956, 'The magical number seven, plus or mi-
concepts in design learning as well as a structured nus two: some limits on our capacity for processing

information', Psychological review, 63(2), p. 81
Pak, B and Verbeke, J 2012, 'Design Studio 2.0: Augment-
ing Reflective Architectural Design Learning', Elec-
tronic Journal of Information Technology in Construc-
tion, 17(Special Issue eLearning 2.0: Web 2.0-based
social learning in built environment), pp. 502-519
Pak, B and Verbeke, J 2014, 'Geoweb 2.0 for Participatory
Urban Design: Affordances and Critical Success Fac-
tors.', International Journal of Architectural Comput-
ing,, 12(3), pp. 283-305
Schon, DA and Rein, M 1994, Frame Reflection: Toward
the Resolution of Intractable Policy Controversies, Ba-
sic Books, New York
Segers, NM 2004, Computational representations of words
and associations in architectural designDevelopment
of a System Supporting Creative Design,, Ph.D. Thesis,
Eindhoven University of Technology
Shannon, P, Markiel, A, Ozier, O, Baliga, NS, Wang, JT,
Ramage, D, Amin, N, Schwikowski, B and Ideker, T
2003, 'Cytoscape: a software environment for in-
tegrated models of biomolecular interaction net-
works', Genome research, 13(11), pp. 2498-2504
Teal, R 2010, 'Developing a (non-linear) practice of de-
sign thinking', International Journal of Art and Design
Education, 29(3), pp. 294-302
Waldron, MB and Waldron, KJ 1988 'Conceptual CAD
tools for mechanical designers', Proceedings of Com-
puters in Engineering Conference 1, pp. 203-209

GENERATIVE DESIGN | Design
Concepts & Strategies
Coding as Creative Practice
Corneel Cannaerts1
1
KU Leuven Faculty of Architecture
1
corneel.cannaerts@kuleuven.be
This paper looks into coding as a creative practice within architecture, more
specifically into textual and graphical coding as a practitioner during the design
process. It argues that coding is not a mere tool for designing but a particular
design medium, with its own affordances and resistances. Using code as a design
medium provides a specific form of feedback, it influences the design process and
its outcomes. Code is a technological and conceptual support for design thinking.
In other words, code and coding can be ascribed agency in architectural design.
This research is based on a number of cases from design practice and teaching,
ranging from small design experiments, developing software tools for specific
design projects and teaching workshops. The cases are grouped into three
metaphors, each describing a particular aspect of coding as a design medium.
Keywords: coding, sketching, tooling, structuring
INTRODUCTION tural research both in academia and practice, often

In the last decade coding has become increasingly described as research by design and creative prac-
accessible as a design medium in architecture and tice research (Fraser 2013; Verbeke 2013; Verbeke et
design. Design software is more and more extended al. 2015). In this mode of research, which combines
and customised through the development of script- action and reflection, practice and theory, the act of
ing interfaces, add-ons, plugins, libraries, integrated designing is a substantial and crucial driver for re-
development environments and programming lan- search. The research discussed in this paper uses this
guages (Davis and Peeters 2013). Designing through mode of inquiry, which can be described as the re-
writing custom code has shifted from being practiced search into the 'medium itself' (Van Schaik and John-
by a handful of academics and pioneers to a broad son 2011; Blythe and van Schaik 2013), as it looks into
emerging culture of coding in architecture and de- the role of coding as a design medium in creative
sign (Burry 2011). While there have been an exten- practices of architecture and design. Coding is ap-
sive number of exhibitions, conferences and publi- proached not with the aim towards the fundamen-
cations documenting this shift, they tend to focus tal understanding of a computer scientist or the pro-
either on code as technique or on the design out- fessional understanding of a software developer, but
comes, only a limited amount of publications address rather with the aim of exploring its potential for de-
how code as a design medium informs this culture of sign practice.
coding (Maeda 2004; Silver 2006; Coates 2010; Reas This paper is framed in the two evolutions out-
and McWilliams 2010; Burry 2011). lined above and explores the impact of the increased
Recently there has been a change in architec- accessibility of coding on practices of architecture
GENERATIVE DESIGN | Design Concepts & Strategies - Volume 1 - eCAADe 34 | 397

and design. The aim is to demonstrate how coding tool the feedback it provides to the designer is partly
as a medium informs a design process. The role of visual, and if the designer is fluent enough in writing
coding is explored not isolation but in relationship code it can be used as a sketching tool.
to other media used in a design process. This pa-
per is based on a number of cases from design prac- Figure 1
tice and teaching, ranging from small design exper- Various design
iments, developing software tools for a specific de- experiments and
sign project and teaching workshops. Through re- sketches: Sticky
flection on the projects the cases are grouped into Stuff, Bubble Grid,
three metaphors, each describing a particular aspect Magnetic Field,
of coding as a design medium. Recursive Growth.
SKETCHING WITH CODE

Since the adoption of computers in architectural
practice there has been a debate on whether CAAD
can be used as a sketching medium. The long held
belief (Schön 1983; Garner 1992; Goel 1995; Bilda and
Demirkan 2003) that design ideation was only pos- This notion of sketching, bridging the gap between
sible through hand drawing and CAAD was limited verbal and visual and exploring design ideas through
to refining and documenting already existing design writing code is central to Processing, an open-source
ideas is challenged by more recent research (Jonson programming language, development environment
2005). Schaeverbeke and Heyligen (2012) argue that and community of creative coders (Maeda 2004; Reas
both analogue and digital media can used for sketch- and McWilliams 2010; Reas and Fry 2007; Process-
ing. Furthermore, contemporary practice architects ing, 2016). Programs written in Processing are called
tend to switch between or use different design me- a sketches, and they are collected in a sketch book.
dia simultaneously, design ideation can be located in Its syntax refers to terms used in design practice and
this 'in-between' (Schaeverbeke and Heyligen 2012). software: drawing, stroke, fill, etc. Its simple inter-
The appropriation of coding by architects it cannot face, reference and syntax is deliberately designed
be seen in isolation but as an extra layer of this fluid to reducing the gap between writing code and get-
and diverse use of media in design practice (Meredith ting visual feedback on screen. Its core functionality
and Sample 2013). can be extended, with third party libraries developed
Gänshirt (2007, p.101) makes a distinction be- by the community, into a means for prototyping and
tween verbal and visual design tools, corresponding production.
with complementary ways of thinking: verbal, linear, In my design practice writing sketching through
logical thinking on the one hand, visual-spatial, as- code plays an important role, I have used Processing
sociative thinking on the other. Based on McLuhan's extensively ranging from quick sketches, design ex-
(1964, p.22) thesis that media are ordered hierarchi- perimentation, coding tools for other architects, de-
cally in terms of abstraction, coding can be seen as an signers and artist, writing code for specific fabrication
abstract verbal design medium, whereas sketching is machines (Cannaerts 2015a; Cannaerts 2015b). The
classified as an intuitive visual medium. Interestingly sketches shown in Figure 1, can be compared to doo-
enough Gänshirt (2007, p.101) defines the computer dles, or explorations of specific generative systems.
as being on the intersection between visual and ver- Moments of playing, interacting with the graphical
bal tools. Even though writing code is a verbal design representation of code and tweaking values are alter-
398 | eCAADe 34 - GENERATIVE DESIGN | Design Concepts & Strategies - Volume 1

nated with changes to the code itself. Rather than can take the form of simpler input geometry or low
describing the specific details of these projects, in number iterations or variables. Once an algorithm
this paper I will reflect on the experience of sketching reaches a certain state of development, I tend to in-
with code and how this informs my design practice. crease the complexity, which is often a revealing mo-
Working with code as a design medium provides ment. In graphical programming, a similarly layered
the designer with different kinds of feedback on the feedback exists, allowing work on the geometrical
screen: a graphical window showing the result of the and the algorithmic simultaneously.
running code, a textual one showing the actual code
itself and possibly textual feedback through the con- MAKING AND USING TOOLS
sole. Although the running code can be made to Using code as a design medium affords control over
respond to various inputs, for example mouse and the algorithms beneath the interface of software
keyboard, the design mainly progresses by working tools and, as such, allows designers to develop their
on the code itself. Text-based coding is an unfor- own design tools. Going beyond the intended use
giving medium - forgetting even one character will of a tool or developing your own tools is frequently
lead to a syntax error, and it is often hard to tell from mentioned by practitioners as a main motivation for
the visual feedback alone what is exactly going on in using code as part of their creative practice (Maeda
an algorithm. These limitations can be overcome by 2004). In architecture, this is a prominent argument
continuously testing the code, incrementally build- in many publications on parametric and algorithmic
ing on working versions of the code and using the design (Kilian 2006; Meredith 2008; Fischer 2008;
console to provide textual feedback, or by develop- Burry 2011). This position is most explicitly stated by
ing a debug mode that renders certain information Aranda and Lash (2006) under the title tooling, they
on the screen. Graphic coding interfaces tend to describe a number of algorithmic techniques which
be a bit more forgiving, as code is contained within are illustrated by a version of the algorithm in pseudo
blocks with clearly defined inputs and outputs, but code, a number of experiments and a project devel-
they tend to become quite hard to read once defini- oped with this specific algorithm.
tions get larger. In 2013 I gave a workshop together with Phil
As stated above, sketching with code adds an- Ayres and Hollie Gibbons at CITA, Kopenhagen , Den-
other layer to the multiple design media used in prac- mark to start the Adaptive Aggregates installation
tice, and does not replace sketching with pen and pa- project as the start of the CITA studio master pro-
per but rather complements it. While coding sketch- gram. The workshop conceived as an introduction to
ing on paper can greatly help with visualising ideas computational design that required no prior knowl-
while simultaneously testing them out in code. Dif- edge of coding. Three design tools were developed
ferent from sketching with pen and paper, sketches looking into simulating aggregate structures across a
with code develop incrementally, not by retracing a number of scales. The design tools used text-based
sketch but by building on working previous blocks of coding in Processing and parametric modelling in
code. The reuse of code and gradual increase in com- Grasshopper, while they could be used without al-
plexity allows sketches to be turned into blue prints tering the actual code through user interfaces, they
and actual design projects. introduced various computational design strategies
Code as a medium to develop ideas tends to and provided the source code for students willing to
progress in chunks as parts of the underlying algo- learn how the underlying code worked.
rithm get defined, evaluated and refined. While de-
veloping an algorithm, often happens through a sim-
plified version of the design problem at hand, which

design tools with graphical interfaces. The three in- Figure 2
troduced tools are described in more detail above. Tool 1: material and
The position, orientation and geometry of the com- simulated inflation.
ponents could be described by the three coordinates
and an offset thickness, and all the tools could ex-
port and import text files containing this information.
Rhino was central in the work-flow; it was used to col-
lect and compile the multiple explorations, prepare
the starting conditions and environment meshes as
input for the simulations and assemble the different
strategies into a final design proposal.
Figure 3
Tool 2: Branching
The first tool (Figure 2) looked into the fabrication based on
and inflation of the material component, it was de- connecting logic.
veloped through iterative material testing and sim-
ulated using a particle spring model in Grasshopper
and Kangaroo. The components, consisted out of an
inflated tube were fabricated prior to the workshop
in three scales. The second tool (Figure 3) was coded
in Processing and explored different ways of connect-
ing the components, it allowed for exploring branch-
ing structures based on the logic of connecting the
different components. The user interface allowed for
control of the geometry of the components, the rules
of connection, the noise introduced by small devia- In the tools coded for Adaptive Aggregates work-
tions in the connecting system. The third tool (Fig- shop, simulations could be found on multiple scales.
ure 4) used a rigid body simulation to simulate the Through nesting different simulations, it becomes
pouring or packing aggregations, the user interface clear that each tool comes with its own assumptions
allowed for the position, type and number of compo- and its own requirements and limitations and all pro-
nents to be poured into the scene, an environment vide a different insight in the design task at hand. All
mesh could be used to simulate pouring over differ- of these simulations have their affordances and cap-
ent objects and obstacles. ture materiality in discrete encoded elements, from
In order to overcome the limits inherent in de- the particles and springs simulating the inflation to
sign software, we used different computational de- the compound shapes assembled out of convex parts
sign tools and strategies to develop a spatial propo- in the rigid body simulation. Furthermore, time is
sition. We chose to operate simultaneously in dif- encoded as explicit discrete steps: the algorithm
ferent software environments and use different com- that makes up the simulation computes the resulting
putational design techniques: explicit geometric world one iteration at a time.
modelling in Rhino, visual parametric modelling in Having access to the code that drives software
Grasshopper and text-based scripting in Processing. tools can allow for a deeper understanding of the de-
The preparation resulted in three design tools; com- sign issues at hand and uncover the assumptions in-
putational techniques were made available through herent in the tools. Actively developing this code

allows for these assumptions to be questioned and ing their environments, the forces at play, and other
explored differently. In the Adaptive Aggregations components.
workshop I was often reminded of my own assump-
Figure 4 tions as students were using the tools. To that ex-
Tool 3: Pouring tent, I think the tooling metaphor is useful, but it
simulation. also introduces an opposition between tool making
and tool using, which in my practice of using code
as a design medium are not separate activities, rather
they mutually inform each other. The process of mak-
ing a tool gradually unravels the design problem at
hand. Furthermore, coding your own design tools
does not generally start from a blank canvas but is
instead based on examples, code snippets, add-ons
and libraries often developed by others. The environ-
ments in which you code are obviously tools them-
selves, with their own assumptions, limits and poten-
Figure 5 tial, programmed by someone else.
Material & Digital
Form Finding STRUCTURING CODE
Workshop. One of the main reasons architects and designers
name for appropriating coding in their creative prac-
tice is automating repetitive work (Burry, 2011). Reas
and McWilliams (2010), call repetition the 'computers
talent' and demonstrates this with a number of ex-
amples from art, design and architecture. On could
say that the effort and time spent on coding only be-
comes useful if some degree of repetition and vari-
ation is involved, and it can operate on collections,
lists or arrays of elements. Code has a bias towards
the many and the multiple, and designing with code
tends to shift the focus from the one off and the Much of the syntax of code has to do with struc-
unique to the systemic and the multiple (Burry 2011). turing repetition and meaningful variation. In text-
In the Material & Digital Form Finding workshop based coding this is reflected in defining and declar-
(Figure 5), material experimentation formed the ba- ing variables, loops, conditionals, functions, objects,
sis for coding different a number of bespoke design classes etc., which are all means of efficiently struc-
tools that allowed for exploring spatial potential of turing code and determine the flow of instructions
form finding systems. Conceptually translating ma- passed to the computer (Fry 2010). The elements
terial system into a code requires decisions into how for structuring code are highly hierarchical and are
to structure the code, ie, what are the material com- geared toward modularity and reuse, and splitting up
ponents and how do they relate computational con- a design problem into reusable chunks.
cepts. In this project they were simulated as particle Findings from the Material & Digital Form Finding
spring-models (Kilian 2006), where springs and par- workshop were further developed. The Folded Strips
ticles are modelled as autonomous objects negotiat- Pavilion project (Figure 6) , which extended the rela-
tionship between material and digital to be more it-

Figure 6
Folded Strips
Pavilion: Design
Model, Fabrication
Model, Paper &
Metal Prototype.
erative and cyclical by also including digital fabrica- In visual programming languages, such as
tion. Since the project required digital fabrication the Grasshopper, code is structured as a network of com-
workflow was based on Rhino and Grasshopper as a ponents, where each component computes an out-
more robust modelling environment, which was fur- put based on data it receives as an input. Different
ther extended with Kangaroo, a physics simulation than in text-based programming, the flow of exe-
add-on for Grasshopper and Anemone, which allows cution is explicitly visualised, which provides a clear
recursion within the Grasshopper environment. The feedback of the algorithm. When definitions become
Folded Strips Pavilion used a physics simulation to de- more complex, components can work on lists of data
rive the overall form of a pavilion . The design sim- and nested lists of data, or data trees, it becomes in-
ulated a particle-spring model based on hexagonal creasingly harder to understand what is happening
grid, the grid can be deformed to allow for denser or in the algorithm which necessitates clearly structur-
less dense areas. In order to be fabricated the data ing the definitions through grouping, naming and
tree of the hexagonal grid is reorganised as continu- clustering (Davis et al 2011).
ous folded strips rather than hexagonal cells.

CONCLUSION: AGENCY OF CODE and layered cultures of code, the agency of code lies
While the metaphors used in this paper to describe obviously largely within this shared and cumulative
coding as a creative practice - sketching, tooling, history of code and the people who contributed to
structuring - each highlight a particular aspect of how this.
coding operates as a design medium, none describes The cases discussed in this paper approaches
the use of coding in all its intricacy. Because of its re- coding from the perspective of a designer writing
usability and fluidity, code as a design medium can code as part of the design process. The arguments
switch between these different modes of informing presented here are mainly limited to a conversation
the design process. The cases discussed in this paper between the author as designer and the design at
demonstrate that code is a specific design medium hand. Coding increasingly influences shared prac-
with its own affordances and resistances. In other tices and design collaboration, further research could
words code has agency in the design process and be done in sharing code as collaborative practice.
each of the discussed metaphors unveils part of that Code becomes increasingly a shared language be-
agency (Cannaerts 2015a; Pickering 1995). tween multiple partners in the design process.
The fundamental difference between coding
and most other design media, is the layered-ness of ACKNOLEDGEMENTS
the provided feedback. Coding allows for simultane- The Folded Strips Pavilion project was developed by
ously working on the geometry representing a partic- Corneel Cannaerts in collaboration with Bart Mer-
ular instance the design and the textual representa- mans and VAC machines.The Adaptive Aggregates
tion of the logic that drives that design. Through hav- project was developed by Phil Ayres, Hollie Gibbons
ing sufficient experience in coding and using intu- and Corneel Cannaerts, CITA Copenhagen.All other
itive interfaces, coding can become a form of sketch- sketches and projects by the author.
ing, although it operates through both textual and
geometric manipulations. Working on verbal and vi-
sual representations of design is also reflected in the
REFERENCES
Aranda, B and Lash, C 2006, Tooling, Princeton Architec-
difference between making and using tools. Instead
tural Press, New York
of promoting the first over the latter it is through the Bilda, Z and Demirkan, H 2003, 'An insight on designer’s
switching between both that designing with code sketching activities in traditional versus digital me-
progresses. Both object oriented programming in dia,', Design Studies, 24, pp. 27-50
Processing and data trees in Grasshopper are elegant Blythe, R and van Schaik, L 2016, 'What if Design Practice
structures, computationally but also in visualising the Matters?', in Fraser, M (eds) 2016, Design Research in
Architecture, an Overview, Ashgate, Burlington, pp.
structure of the algorithm, this paper demonstrates
53-69
that those structures can be a conceptual driver for Burry, M 2011, Scripting Cultures: Architectural Design and
design. Programming, Wiley
Cannaerts, C 2015a, Negotiating Agency: Computation
and Digital Fabrication as Design Media, Ph.D. Thesis,
FURTHER RESEARCH RMIT University
Code is a shared and collective medium, the work Cannaerts, C 2015b 'Fabricating Material Intensities:
presented here could not be possible without the Designing and Making Mediated by Computation',
programmers and designers of the programming Creative Practice Conference: Making Research | Re-
languages, interfaces and hardware. Having access searching Making., Aarhus, pp. 151-161
to underlying algorithms in design is empowering, Davis, D 2013, Modelled on Software Engineering: Flexi-
ble Parametric Models in the Practice of Architecture.,
but is only pealing a way the first layer of interfaces
Ph.D. Thesis, RMIT University
of design tools. What is revealed is a highly designed

Davis, D, Burry, J and Burry, M 2011 'Untangling Para- [3] http://www.grasshopper3d.com/
metric Schemata: Enhancing Collaboration through
Modular Programming', Designing Together: Pro-
ceedings of the 14th International Conference on Com-
puter Aided Architectural Design Futures,, Liège, pp.
55-68
Davis, D and Peters, B 2013, 'Designing Ecosystems, Cus-
tomising the Architectural Design Environment with
Software Plug-ins', in Peters, B and De Kestleir, X
(eds) 2013, Computation Works: The Building of Algo-
rithmic Thought, Wiley, pp. 124-131
Fischer, T 2008, Designing (tools (for designing (tools for
…)))), Ph.D. Thesis, RMIT University
Garner, S 1992, 'The Undervalued Role of Drawing in De-
sign', in Thistlewood, D (eds) 1992, Drawing Research
and Development, Longman, pp. 98-109
Goel, V 2005, Sketches of Thought, MIT Press
Gänshirt, C 2007, Tools for Ideas: An introduction to Archi-
tectural Design, Birkhäuser Architecture
Jonson, B 2005, 'Design ideation: the conceptual sketch
in the digital age', Design Studies, 25, pp. 613-624
Kilian, A 2006, Design Exploration through Bidirectional
Modeling of Constraints, Ph.D. Thesis, Massachusetts
Institute of Technology.
Maeda, J 2004, Creative Code: Aesthetics + Computation,
Thames & Hudson
McLuhan, M 1964, Understanding Media: The Extensions
of Man, McGraw-Hill
Meredith, M 2008, From Control to Design: Parametric/Al-
gorithmic Architecture, Actar
Meredith, M and Sample, Hilary 2013, Everything All at
Once: The Software, Videos, and Architecture of MOS.,
Princeton Architectural Press
Pickering, A 1995, The Mangle of Practice: Time, Agency,
and Science, University of Chicago Press
Reas, C and McWilliams, C 2010, Form+Code in Design,
Art, and Architecture., Princeton Architectural Press
Schaeverbeke, R and Heylighen, A 2012 'In Search of the
In Between', In Physical Digitality: Proceedings of the
30th eCAADe Conference, Prague, pp. 49-57
Schön, D 1983, The Reflective Practitioner, Temple Smith
Silver, M (eds) 2006, AD: Programming Cultures: Art and
Architecture in the Age of Software, Wiley
Verbeke, J 2013, 'This is Research by Design', in Fraser,
M (eds) 2013, Design Research in Architecture, an
Overview, Ashgate, pp. 137-160
Verbeke, J, Zupanzic, T and Achten, H 2015 'Digital Tools
and Creative Practice in Architectural Research', Real
Time, eCAADe Conference 2015:, Vienna, pp. 25-28
[1] http://benfry.com/deprocess/
[2] https://processing.org/

Using Processing with Architectural 3D Modelling
Inês Caetano1 , António Leitão2
1,2
INESC-ID/Instituto Superior Técnico
1
ines.caetano@tecnico.ulisboa.pt
2
antonio.menezes.leitao@ist.utl.pt
Although programming was considered a specialized task in the past, we have

been witnessing an increasing use of algorithms in the architectural field, which
has opened up a wide range of new design possibilities. This was possible in part
due to programming languages that were designed to be easy to learn and use by
designers and architects, such as Processing. Processing is widely used for
academic purposes, whereas in the architectural practice it is not as used as
other programming languages due to its limitations for 3D modeling. In this
paper, we describe the use of an extended Processing implementation to generate
three 3D models inspired in existing case studies, which can be visualized and
edited in different CAD and BIM applications.
Keywords: Generative design, Programming, Processing, 3D modeling
INTRODUCTION to designers and architects. Such is the case of the

Only recently have architects considered the use of Processing language (Reas and Fry 2007).
programming in architecture, as they become aware
of its potential, thus introducing it in their design BACKGROUND
practices (Burry 2011). This allowed the automation Processing was created especially for designers with-
of tedious tasks, the exploration of generative pro- out any previous experience in programming. Nev-
cesses, and the generation of complex solutions that ertheless, it has also been used in other fields, includ-
would be difficult and time consuming to produce ing architecture. This programming language is con-
manually. Therefore, algorithms became extensions sidered as a pedagogical language and, in fact, it has
of human thinking by overcoming its potential limita- being taught in several academic courses. Therefore,
tions, allowing the exploration and experimentation this language has grown over the years and has be-
in an alternative realm (Terzidis 2003). The concept come increasingly popular due to its simplicity, to the
of Generative Design (GD) can be defined by this al- academic community support, and to the excellent
gorithmic and rule-based process, through which a documentation available (Fricker et al. 2008).
wide variety of solutions can be created in a short pe- Still, in the architectural practice Processing is
riod of time (Fasoulaki 2008). not as used as other programming languages, such
However, programming is not a trivial task (Burry as Python, Grasshopper, AutoLISP, and VisualBa-
2011). In order to facilitate the use of GD, some sic. The main reasons for this unfortunate situation
programming languages were carefully designed or are the inability of Processing to interact with the
adapted with the aim of teaching programming skills Computer-Aided Design (CAD) and Building Informa-

tion Modeling (BIM) tools that are typically used by Sketchup, Revit, and ArchiCAD).
architects, such as AutoCAD, Rhinoceros 3D, and Re- In order to make Processing suitable for the
vit, and also its shortcomings in 3D modeling oper- needs of architects, 3D modeling extensions to the
ations and transformations, such as sweeping, loft- Processing language were also implemented in the
ing or extruding. This is not surprising, as Processing Rosetta IDE. These extensions include several opera-
was originally intended for 2D drawings and animations for basic 3D modeling, such as boxes, spheres,
tions, running in its own programming environment cylinders, cones, etc., as well as shape forming oper-
and completely isolated from other applications. ations, such as lofting, sweeping and extruding, and
Only recently was Processing extended with sim- also Boolean operations, i.e. subtraction, intersec-
ple 3D operations and ways of exporting the gener- tion, and union of shapes.
ated designs. Some libraries were also developed to In addition to supporting the traditional syntax
add some of the required operations, such as increas- and semantics of the Processing language, our solu-
ing the range of 3D primitives, creating and control- tion extended Processing in three directions:
ling irregular shapes, or dealing with some 3D shape
1. Interactive evaluation, which allows designers
transformations. In the Related Work section we will
to evaluate small fragments of Processing pro-
enumerate and describe some of these libraries with
grams in a Read-Eval-Print-Loop (REPL), en-
more detail.
abling quick experimentation of the scripts
Unfortunately, these 3D operations are still lim-
being developed;
ited in their capabilities. Therefore, the application
2. 3D modeling, an essential extension in order
of Processing in the field of architecture remains dif-
to improve the use of Processing for architec-
ficult, even though it is easy to learn and use. In or-
ture;
der to overcome this situation, we propose an aug-
3. Professional CAD, a connection between Pro-
mented Processing for architects that deals with a
cessing and CAD/BIM tools, supporting the
wider range of 3D modeling primitives (cylinders,
generation of designs in those tools without
spheres, cones, etc) and transformations (extrusions,
suffering from the problems that typically oc-
lofts, sweeps, Boolean operations, etc), which are es-
cur when designs are imported from different
sential in the architectural daily practice, and gener-
applications.
ates their results directly into a CAD or BIM tool.
The most significant advantage of the implementa-
AUGMENTING PROCESSING tion of Processing for Rosetta is the ability to use, in
In the past (Correia and Leitão 2015), we proposed a Processing programs, all the 2D and 3D modeling op-
solution to join CAD and BIM tools with the Process- erations available in Rosetta. Given that these op-
ing language, allowing architects to develop new de- erations access the corresponding operations in the
signs using this programming language while gen- CAD/BIM tool being used, this gives Processing the
erating their results directly into a CAD or BIM ap- capability to directly create shapes in that tool.
plication. This solution was implemented in Rosetta
(Lopes and Leitão 2011), an Integrated Development USING PROCESSING IN ARCHITECTURE
Environment (IDE) for generative design. One main Our implementation of Processing is intended for ar-
advantage of Rosetta is its emphasis on portabil- chitects that learned Processing and want to use it
ity and, unlike other development environments, in their architectural practice. In this section, we de-
Rosetta supports scripts using different languages velop three architectural examples using Processing:
(AutoLISP, JavaScript, Python, Racket, and Scheme) 1. Al Bahar Towers;
and generates identical models in all supported 2. Allianz Arena;
CAD and BIM applications (AutoCAD, Rhinoceros 3D, 3. Quality Hotel Friends.

Additionally, we discuss the advantages and draw- over, these points were also used to generate the sur-
backs of Processing for the architecture practice. For face mesh of the outer layer, which was then orga-
each of the examples we use Processing running in nized into groups of four points (i.e. a squared mesh)
the Rosetta IDE, thereby allowing the visualization so as to facilitate the future development of the tri-
and edition of the generated models in the sup- angular units (Figure 1, on the right). In practical
ported backend applications, including Rhino and terms, this first stage required operations like calcu-
AutoCAD. lus of matrices and surface normal vectors, arrays of
coordinates, and surface creation.
Al Bahar Towers The second stage was the creation of the tower's
Our first example is the Al Bahar Towers in Abu Dhabi, outer skin with the triangular units. First of all, we de-
design by Aedas Architects. These towers are char- fined the geometry of these units and, then, we ap-
acterized by their responsive facade, which is com- plied them on the squared mesh defined in the pre-
posed by several triangular units inspired in the tra- vious stage.
ditional Islamic element "mashrabiya". Nevertheless, To create the units, we started by calculating the
in this paper we will simply focus on modeling these vertices that defined their shape. Then, these ver-
towers facade design, and not on its kinetic proper- tices were strategically linked using lines, which in
ties. turn were used to produce the surfaces. This process
is synthetized in Figure 2 by the images A-B-C.
Figure 1
Secondly, the generation of the outer skin con-
Inner glass surface
sisted in mapping these triangular units along the
(left-side); outer
squared mesh. Finally, we overlapped both inner and
surface organized
outer surfaces so as to produce the final model, which
in groups of four
is visible in Figure 2.
points (right-side).
Note that, apart from the arrays of coordinates,
this second stage required surface creation between
curves, an operation that is not available in the origi-
nal Processing implementation.
Allianz Arena
Our second example is the Allianz Arena stadium in
Munich, designed by Herzog & de Meuron Architects
with ArupSport. As in the previous example, we had
to first define the overall shape of the stadium. There-
fore, in order to obtain this form we used the math-
ematical formula of the superellipse to produce the
We divided this example into two different stages. cloud of points that corresponded to the stadium's
Firstly, we created the inner glass surface with the shape. This was then used to produce, first, the un-
tower's shape and, then, we explored the outer skin, derlying surface of the model and, then, to distribute
which is composed by the triangular units. Before ex- the diamond shaped cells which characterize the fa-
ploring each stage, we had to define the set of points cade. Lastly, we produced each of these cells using:
that corresponded to the tower's geometry, which
together with a surface creation operation produced 1. an array of coordinates and a closed-line
the inner glass surface (Figure 1, on the left). More- defining the diamond-shaped curve of each
cell;

Figure 2
The Al Bahar Tower
3D model using
Processing. On the
left – the units
generation process:
A. the main vertices
that define each
unit shape; B. lines
to link the main
vertices; C. the
creation of surfaces
between lines; In
the middle - One
instance of the
generated model;
On the right – two
shape variations of
its facade.
Figure 3
Allianz Arena
model: a print
screen of the
generated model,
visualized in
AutoCAD, and the
Rosetta IDE with
the corresponding
Processing code.

2. a loft operation between this curve and a they were subtracted from the walls previously cre-
point placed perpendicularly to it. ated, therefore creating the windows openings (Fig-
ure 4, A-B-C).
To sum up, not only did this model use matrices and Summarily, the modeling of this building re-
vector calculations, it also required the loft operation quired 3D primitives, such as boxes and cylinders, to
to generate the stadium envelope pattern. Figure 3 produce the walls and the circular windows, required
shows an instance of the generated model in Auto- both matrices of points and vector calculations to
CAD and the corresponding Processing code. create and place the circular windows, and, finally,
the Boolean operation Subtraction to create the win-
Quality Hotel Friends dows openings. Figure 4 synthetizes this process and
Quality Hotel Friends, in Stockholm, is our shows one instance of the obtained model using Pro-
last example, which was designed by Karolina cessing.
Keyzer+Wingårdhs. This hotel has a straight facade
and it is composed by several circular windows of Evaluation
three different sizes. Note that these three sized win- In this section, we summarize the generation process
dows are strategically placed to create the wave ef- of all three examples and we demonstrate that the
fect visible in Figure 4. modeling extensions added by our solution were cru-
cial in the exploration of these models.
Figure 4 First of all, we synthetized the operations that
Quality Hotel were necessary for each example and, then, we com-
Friends model: A. pared the availability of those operations in (1) the
The creation of the Processing language and Processing Development
cylinders with Environment (PDE), and (2) in our implementation of
different sizes; B. Processing in Rosetta.
overlapping of the Although our sampling is limited, it revealed the
cylinders with the range of operations that are typically used in 3D mod-
facade wall; C. eling. After analyzing the information presented in
subtraction of the Figure 5, we can conclude that the majority of the op-
cylinders from the erations that are essential for the generation of the
wall. previous examples, are not available in the original
Processing implementation. Moreover, we predict
that models that are more complex than the ones
presented in this paper will require even more ad-
In practical terms, we generated this example in two vanced 3D modeling operations. It is thus important
stages. Firstly, we created the hotel walls with a box to fully support in Processing the operations that are
operation and, then, we produced the round win- available in the CAD/BIM tool being used and that
dows. For this, we had to produce the matrix of is precisely one of the important features of our Pro-
points that shaped the facade, and then we orga- cessing implementation in Rosetta.
nized them into arrays of coordinates. These arrays
were then used to distribute the cylinders over the PORTABILITY
walls, and to control each cylinder radius size. More Portability is the ability of a program to be compiled
precisely, each cylinder radius size was controlled by or run in a different environment and, in our case,
the distance between its positioning point and the it allows us to produce identical models in different
attractor point. After generating all the cylinders,

Figure 5
Synthesis of the
needed operations
for each example,
and their
availability in both
Processing PDE and
in our
implementation of
processing to
Rosetta IDE.
CAD tools, such as Rhino, AutoCAD, or SketchUp, the combination of Processing with the other sup-
among others. Using our solution, it is now possible ported programming languages such as Autolisp,
to (1) explore 3D architectural models using the Pro- Phyton, Racket and Javascript.
cessing programming language, since the modeling
operations required are already available, and (2) to RELATED WORK
visualize the obtained models in the different CAD or There are already some libraries that improved Pro-
BIM tools that are essential for architects. Moreover, cessing to deal with more 3D modeling extensions.
it allows the designer to easily change the CAD tool Shapes 3D [6] is a library that extends processing with
that he wants to use (Figure 6). a set of 3D shapes, including ellipsoids, toroids, he-
Finally, Rosetta also promotes portability across lixes, and others. iGeo [5] is a 3D modeling soft-
the supported programming languages, allowing ware library in Java, which has an interface special-

Figure 6
The Al Bahar tower’s
model using
different CAD
applications
(SketchUP,
AutoCAD and
Rhino5).
ized for processing called piGeon. This library in- the architecture community, mainly for those who do
cludes vector operations, NURBS curve and surface not have any programming experience. Both have
geometries, polygon meshes, and 3D model file I/O. the advantage of interacting with at least one of the
ANAR+ is a geometry library for Processing (Labelle et CAD or BIM tools that are used in architectural prac-
al. 2010) that was intended to be a programming in- tice.
terface supporting shape exploration based on para- Grasshopper is a visual programming language
metric variations. Toxiclibs [4] is an independent li- for Rhinoceros. This language is widely used by be-
brary collection for computational design tasks with ginners as it allows them to quickly develop, test and
Java and Processing, which supports 2D/3D vectors, visualise small programs due to the friendly and in-
spline curves, 4x4 matrices, intersection tests, mesh tuitive environment. Recently, it was also extended
container, and OBJ and STL exporters. Finally, Com- to interact with Revit and ArchiCAD via independent
putationalGeometry library [3] allows Processing to plug-ins. Unfortunately, Grasshopper has scalabil-
deal with dynamic mesh generation and rendering, ity problems: the more complex the user program
including isometric contours and surfaces, boundary gets, the more difficult it becomes to understand and
hulls and skeletons. maintain it. The use of clusters can mitigate this prob-
One of the problems with these libraries is that lem but, in practice, it is rare to see programs that take
none of them allows architects to directly interact advantage of the idea. The plug-ins for interacting
with the most used CAD/BIM applications. Therefore, with other tools besides Rhinoceros are very experi-
they have to resort to import procedures which, fre- mental and not yet widely used. For the purposes of
quently, loose some of the geometric information or this work, the major drawback of Grasshopper is that,
transform it into undesirable forms (e.g. converting although it also supports some textual programming
smooth surfaces into meshes). In addition, some 3D languages, the list does not include Processing.
operations are still missing, including Boolean oper- Similarly, Dynamo is a plug-in for Revit strongly
ations and shape transformation operations. Our so- influenced by visual programming languages. It is
lution, besides extending processing with 3D model- also based on a workflow of nodes and connections
ing operations and primitives, also allows the inter- that creates BIM objects in Revit. Dynamo share the
action between Processing and the CAD and BIM ap- same advantages of Grasshopper, but it also suffers
plications typically used by architects. from the same limitations, including the scalability
Dynamo [1] and Grasshopper [2] are program- problem and the fact that it does not support the Pro-
ming languages that are also quite popular among cessing language.

Comparing to Rosetta IDE, Grasshopper and Dynamo chitectural modelling.
might be more intuitive and user-friendly, partic- A final advantage of our solution is that it also al-
ularly for those without programming experience. lows architects to combine Processing with the dif-
Nevertheless, to overcome their limitations when de- ferent programming languages provided by Rosetta,
veloping more complex models, architects are forced such as Python and Scheme. This allows Processing
to use textual programming languages. As a conse- to move from its comfort zone - the design environ-
quence, architects end up dealing with textual lan- ment - into the more complex and demanding archi-
guages even when they were not supposed to. More- tectural environment.
over, programs that combine both visual and textual
programming languages tend to be less organized ACKNOWLEDGEMENTS
and more difficult to understand. This work was supported by national funds through
On the other hand, although Rosetta requires a Fundação para a Ciência e a Tecnologia (FCT) with ref-
more profound initial investment in learning the Pro- erence UID/CEC/50021/2013.
cessing language, it allows the development of pro-
REFERENCES
grams with higher complexity and more levels of ab-
Burry, M 2011, Scripting Cultures: Architectural Design and
straction. Programming, John Wiley & Sons, Ltd, Publication,
U.K.
CONCLUSION Correia, H and Leitão, A 2015 'Extending Processing to
Processing is a simple and pedagogical program- CAD Applications', Real Time: Extending the Reach of
ming language and, hence, easy to learn by design- Computation - Proceedings of the 33th eCAADe, Vi-
ers with no previous programming experience. How- enna, Austria
ever, when it comes to the architects' work, Process- Fasoulaki, E 2008, Integrated Design: A Generative Multi-
Performative Design Approach, Ph.D. Thesis, Mas-
ing shows its limitations:
sachusetts Institute of Technology (MIT)
1. Lack of 3D modeling operations; Fricker, P, Wartmann, C and Hovestadt, L 2008 'Pro-
cessing: Programming Instead of Drawing', Architec-
2. Difficult combination and interaction of Pro-
ture in Computro - Proceedings of the 26th eCAADe,
cessing with the most used tools in the archi- Antwerpen (Belgium)
tectural practice. Labelle, G, Nembrini, J and Huang, J 2010 'Geometric
programming framework, ANAR+: geometry library
Our solution overcomes these two barriers, by aug- for processing', Future Cities - proceedings of the 28th
menting Processing with new design abstractions eCAADe, ETH Zurich, Zurich, Switzerland
and operations, and by connecting it with several Lopes, J and Leitão, A 2011 'Portable Generative De-
CAD and BIM tools. The examples developed in the sign for CAD Applications', ACADIA 11: Integration
paper illustrate the extension and adaptation of Pro- Through Computation - Proceedings of the 31st ACA-
DIA, Banff, Alberta
cessing to the architectural practice, not only by en- Reas, C and Fry, B 2007, Processing: A Programming Han-
abling the use of the most needed 3D operations and dobook for Visual Designers and Artists, MIT Press,
transformations, but also by allowing a direct con- MIT, USA
nection with the most used CAD and BIM applica- Terzidis, K 2003, Expressive Form: A Conceptual APproach
tions. to Computational Design, Spon Press, New York
It is noteworthy that the majority of the 3D prim- [1] http://www.dynamoprimer.com
[2] http://www.grasshopper3d.com/
itives and operations used to develop the examples [3] http://thecloudlab.org/processing/library.html
presented in this paper could not have been done [4] http://toxiclibs.org/
using exclusively the features provided by the orig- [5] http://igeo.jp/p/
inal Processing language, as it does not provide lofts, [6] http://www.lagers.org.uk/s3d4p/index.html
sweeps, cylinders, and other crucial operations for ar-

Using the Phase Space to Design Complexity
Design Methodology for Distributed Control of Architectural Robotic
Elements
Mary Katherine Heinrich1 , Phil Ayres2

1,2
Centre for Information Technology and Architecture, IBT, KADK
1,2
{mhei|Phil.Ayres}@kadk.dk
Architecture that is responsive, adaptive, or interactive can contain active

architectural elements or robotic sensor-actuator systems. The consideration of
architectural robotic elements that utilize distributed control and distributed
communication allows for self-organization, emergence, and evolution on site in
real-time. The potential complexity of behaviors in such architectural robotic
systems requires design methodology able to encompass a range of possible
outcomes, rather than a single solution. We present an approach of adopting an
aspect of complexity science and applying it to the realm of computational design
in architecture, specifically by considering the phase space and related concepts.
We consider the scale and predictability of certain design characteristics, and
originate the concept of a formation space extension to the phase space, for
design to deal directly with materializations left by robot swarms or elements,
rather than robots' internal states. We detail a case study examination of design
methodology using the formation space concept for assessment and
decision-making in the design of active architectural artifacts.
Keywords: phase space, complexity, attractor, distributed control
INTRODUCTION & BACKGROUND et al. 2008). This paper deals specifically with the
The methodology detailed in this paper is devel- problem of the architectural design of such a robotic
oped as part of the EU research project flora robot- system, and the question of how high-level architec-
ica (Hamann et al. 2015, [1]), which uses tightly cou- tural design requirements can be merged with low-
pled plant-robot bio-hybrids to form architectural level rules of robotic control.
artifacts that are self-organizing and evolving, with
distributed control mechanisms. In the flora robot- Active architectural elements
ica project, the robotic controllers incorporate ap- In this context, the architectural design problem
proaches from evolutionary robotics (Bongard 2013), at hand is that of designing controllers for ac-
distributed embodied evolution (Watson et al. 2002), tive/robotic architectural elements that have strong
and interactive evolutionary computation (Secretan self-organizing behaviors.

Existing buildings containing active products ality gap to replicate our simulation results in phys-
(such as automated window shades or automated ical experiments. Having demonstrated that, in the
HVAC) are often controlled centrally, through read- tested conditions, results of plant-robot bio-hybrid
ily available technology known as building automa- controllers evolved in simulation are able to cross the
tion systems or smart building management systems reality gap, we seek to extend the work to evolving
([2]:Siemens 2012). These centralized systems deal distributed controllers. In this paper, we focus on the
with changes in environmental conditions by exist- architectural design problems posed by the context
ing design methodologies of averaging or homog- of distributed controllers evolved in simulation.
enizing (cf. Pavliotis and Stuart 2008), at the scale
Figure 1
of perhaps a wall or a room. In addition to these
Diagram of
industry products, there are existing buildings fea-
bidirectional
turing bespoke active facade elements. There are
communication
many built examples incorporating matrices of LED
links between
lighting, such as the centrally controlled BIX facade
models and
of Kunsthaus Graz (Cook and Fournier 2004). There
analyses in the
are also examples of built facades incorporating ki-
common encoding
netic elements, including Media-TIC (Cloud 9 [3]), IBA
and integrated
Soft House (KVA MATx [4]), BIQ Algae House (Split-
projection.
terwerk [5]), Yeosu Theme Pavilion (SOMA [6]), Al
Bahr Towers (AHR [7]), and MegaFaces (Asif Khan [8]).
Some of these facades feature decentralized control
but, to the knowledge of the authors, do not incor-
porate decentralized communication between ele-
ments. There are some departures from the central- Common Encoding & Integrated Projection
ized communication paradigm at the scale of instal- In the flora robotica project, in order to avoid a single
lation, e.g. Sentient Chamber (PBAI [9]) (Chan et al. model's potential bias toward either the plant sym-
2015). There are also relevant architectural demon- biont or the robot symbiont, we will take a pluralistic
strators, prototypes, or probes (Thomsen and Tamke modelling (Helbing 2010) approach to the definition
2009) in literature (Frazer 1995, Beesley (ed) 2006, of the distributed bio-hybrid controller (see partial
Thomsen et al. 2015). We therefore understand the florarobotica models/analyses in Wahby et al. 2016,
question of designing distributed/decentralized con- Zahadat et al. 2015, Soorati and Hamann 2015, Hein-
trol and communication in architectural elements to rich and Ayres 2016). As a general framework for ap-
be relevant to CAAD (computer-aided architectural proaching the problem of merging high-level archi-
design) state of the art, in addition to its relevance to tectural requirements with these pluralistic low-level
flora robotica. rules of control, we propose a common encoding and
integrated projection (see Figure 1) which will incor-
Evolving robotic controllers porate bidirectional communication between each
In prior flora robotica work (Wahby et al. 2016), we
of the low-level input models and a single common
have used NEAT (Stanley and Miikkulainen 2004) to
encoding of those models, will use the current state
evolve Artificial Neural Network controllers in simula-
of the common encoding to generate an integrated
tion for an evolutionary robotics (Bongard 2013) ap-
projection of a late timestep, will assess the inte-
proach to the control of plant growth and motion in
grated projection according to high-level architec-
a bio-hybrid setup, and successfully crossed the re-
tural objectives, and will use that assessment to pun-

ish or reward the evolution of the low-level input In this paper, we consider the complex sys-
models. tems and behaviors of architecture that is respon-
sive, adaptive, or interactive--in other words, archi-
COMPLEXITY & PREDICTABILITY FOR AR- tecture that contains sensor-actuator systems. We
CHITECTURAL DESIGN consider sensor-actuator systems that are controlled
In the description and study of dynamic systems, the through decentralized control and communication,
'phase space' is the representation of all possible in- thereby allowing self-organizing behavior. We also
stantaneous states that can occur in a physical sys- adopt some of the vocabulary of complexity science,
tem (Butkovskiy 1990, Sayama 2015). Each of these so that 'complexity' refers to the number of possibili-
states, and its associated characteristics, corresponds ties in a system and the length of description thereby
to a single point in the mathematical phase space. required to encompass it (Bar-Yam 1997).
In existing design rhetoric, we sometimes refer
Figure 2 to the parameter space and the solution space of a
Author's illustration model. These spaces can be approximated and un-
of Bar-Yam's (1997) derstood analytically, because any particular com-
mathematical bination of parameters always results in the same
concept, called a unique solution. The phase space is distinct from
complexity profile, these in that it describes the full set of possible states
showing how one that may eventually arise from a set of behaviors, and
system can have therefore, in certain types of systems, cannot be de-
different numbers rived purely analytically (see Bar-Yam 1997). In the
of independent context of architectural design, this means that de-
behaviors at sign solutions must be modeled and simulated be-
different scales, fore visual interrogation and analysis of the design
depending on the can occur.
system's scale of These types of systems, which can be described
randomness, through modelling and the analysis of model results,
coherence or are categorically complex systems (or complex adap- It is sometimes assumed in architectural discourse
correlation. tive systems). The scientific study of complex sys- that incorporating methods of emergence, dis-
tems deals in part with the principles governing the tributed control, and evolutionary robotics will
emergence of such systems from simple components largely equate to unpredictability in the architec-
(Bar-Yam 1997). This paper details an approach of tural result (Frazer 1995, Kolarevic and Malkawi (eds)
applying a particular aspect of complexity science to 2005, Kretzer and Hovestadt (eds) 2014). However,
the realm of computational design (or design com- we know from complexity science that this varies
putation), specifically by employing the mathemati- depending on which scale of the system you are
cal phase space as an approach for architectural de- describing, which characteristics you are concerned
sign, modelling, and simulation. Due to the context with, and at which point in time you are viewing the
of architectural design, we assume that only spatially system. For example, in the classic agent-based Boids
distributed models of systems are relevant to this de- model (Reynolds 1987) of flocking birds, though the
sign methodology, which can be an important dis- positions of agents in relation to one another may
tinction in modelling (e.g. work on the breakdown of be very unpredictable at early timesteps, they will be
mean-field approximation and work on spatially dis- much more predictable at later timesteps after they
tributed models, Werfel and Bar-Yam 2004, Bar-Yam have formed flocks. Similarly, if the system's behavior
and Sayama 2006, Dieckmann and Law 2000). is described at the scale of individual agents within

a flock, it is more unpredictable because each run domness at a certain timestep and scale, making it
of the system will have birds with differing start po- seem too unpredictable for architectural design. At
sitions and orientations. However, if the model is such a state, a comprehensive description of the sys-
described at the scale of the flock, then it is much tem would have to include a description of each in-
more predictable, because the system will tend to- dividual agent in order to describe the behavior of
ward similar flocking behaviors, regardless of start the system. But in that same system, there could be
position. coherent behavior at a different timestep and scale.
The amount of predictability varies greatly by At this state, a comprehensive description could be
system. Some complex systems have attractors, comprised of very little information, because it would
which will cause the system to move toward a partic- need to describe only one behavior: that of the co-
ular configuration or set of configurations, depend- herent group (see random, coherent and correlated
ing on the number of attractors and the expanse of behaviors, Bar-Yam 1997).
the basins associated with those attractors. For in- The balance of complexity and simplicity in a de-
stance, if a particular system has two attractors, and signed system may likely depend on which scale is
the basins of attraction from these two cover the being described, and at which point in time. There-
phase space equally and entirely, then the system will fore, the task of designing architecture that is emer-
tend to one of these two states with equivalent prob- gent, distributed, or evolutionary is not a task of de-
ability, despite an expansive array of possible start signing behaviors with primarily unpredictable re-
conditions (see Hurley 1982, Sayama 2015). In other sults. It is rather the task of designing a relevant
words, the phase space of possible configurations is phase space; it is the task of designing the scale at
predictable, if not the exact configuration, and the which coherent or correlated behavior occurs, with-
expansiveness of that predictable phase space de- out resorting to high-level control.
pends on the complexity of the system. The amount
of predictability of the exact configuration that a sys- DESIGN METHODOLOGY CASE STUDY:
tem with attractors will move toward depends on the CONTROLLERS FOR SOLAR RESPONSIVE
number of attractors and basins of attraction. There- SELF-ORGANIZING GROWTH
fore, in looking to design rules for distributed control The case study selected is the task of designing an
and communication in architectural sensor-actuator exterior plant-robot bio-hybrid system (similar to the
systems, we do not design a solution by finding and bio-hybrid setup in our prior flora robotica work,
evaluating individual instances. We instead work to- Wahby et al. 2016) that responds to solar exposure.
ward designing a system's (i.e. set of behaviors') full The design task of the response is to to shade occu-
set of possibilities (the phase space) and its attractors. piable spaces from the sun by growing taller where
The description used to design an architectural needed, but to only grow where necessary, so as to
complex system should occur at the scale that is per- leave as much occupiable space on the site as pos-
tinent to the particular architectural design problem sible. However, the system should maintain some
at hand (this does not change the scale of rule-based presence, even if the site receives no sunlight, so
interactions; scale of rule and outcome are still dis- that the open area is divided into smaller occupiable
tinct). The system should be designed so that coher- spaces, and so that the system has a chance to re-
ent behaviors or correlated behaviors (see complex- act if the lighting conditions change at a later time.
ity profile and related concepts, Bar-Yam 1997) oc- The controller should assume that the bio-hybrid sys-
cur at that pertinent scale (see Figure 2). As we saw tem generally grows upward, to both align with plant
in the example of the classic Boids model (Reynolds dynamics and to be self-structuring. The bio-hybrid
1987), there could be a prohibitive amount of ran- system should be able to be placed on any site and

correctly interpret whether each element is in sun or both time and space, and the possible states of the
shade. system and the possible transitions it can undergo
are finite (Wolfram 1983, Wolfram 1984, Sayama
Figure 3
2015). We use a Moore neighborhood setup [11], in
Comparison of
which, as each cell determines its state, it looks at the
predetermined and
cells directly below, above, and beside it, and it also
centralized (which
looks at the 4 cells diagonal to it. We use a standard
use methods of
majority rule (Wolfram 1983), in which each cell de-
averaging and
cides its state by conforming to whichever state holds
homogenization,
the majority in its neighborhood.
Pavliotis and Stuart
The majority rule means that, if a cell is dark, 2
2008), and
of its neighbors are dark, and 6 of its neighbors are
decentralized
light, the cell will become light in the next timestep
systems, showing A predetermined controller in this case (which would (see Figure 4). Likewise, if a cell is light, 5 of its neigh-
the conditions bring bio-hybrid growth to a certain height based on bors are dark, and 3 of its neighbors are light, it will
under which they solar geometry derived from latitude and longitude) become dark in the next timestep (see Figure 4). The
either fail or is not sufficient , because the controller would be un- majority rule was chosen because its dynamics are
succeed to able to react to shade cast on the site by clouds, trees, known to often lead to clusters, which are desirable
accurately respond or neighboring buildings (see Figure 3). A central- in the given case study. In this setup, we use dimen-
to site conditions. ized sensor-controller would also not be sufficient in sions of 100 x 100 and periodic space, and in initial
this case because, though the sensor would be able conditions use equal probability of randomized cell
to correctly interpret its local neighborhood's solar state distribution. (All setups are completed in a com-
condition, it would assume that the solar condition bination of PyCXsimulator [12], IronPython [13], and
is the same across the site, and would fail in condi- Grasshopper3d/Rhino3d [14], with reference to the
tions where neighboring buildings (for instance) cast PyCX Project [12].)
shadow on half of the site (see Figure 3). Decentral-
Figure 4 ized sensor-controllers, by contrast, would not fail in
Author's illustration any of the above mentioned scenarios (see Figure 3)
of two examples of and has the spatial consequence and design oppor-
transitions of the tunity of increased heterogeneity.
states of single cells If the system has not only distributed control,
in a Moore but distributed communication, it has the added ad- Extending Phase Space to Formation Space
neighborhood [11] vantage of being able to create spatial clusters as it The phase space of a Cellular Automaton is an enu-
majority rule CA grows, to provide as much occupiable space on the meration of all possible configurations in that setup,
(Wolfram 1983), site as possible. Therefore, with the design problem according to the dimensions of the cellular automa-
based on presented in the case study, it is appropriate to de- ton and the number of possible cell states. For
communication sign a system that features distributed control and instance, in the first part of Figure 5, we can see
strategies from distributed communication. all the possible configurations in a 2-cell by 2-cell
Sayama (2015). The two-dimensional cellular automaton with binary cell
central cell in each Cellular Automata (CA) setup states (two possible cell states). The number of possi-
block of 9 is the We select Cellular Automata [10] as a typology to ad- ble configurations in this setup is few, so they are easy
only cell transition dress this design problem, because it models decen- to imagine, visualize, and analyze. In the second part
depicted. tralized communication in a way that is befitting of of Figure 5, we see a visualization of the phase space
non-mobile architectural elements, it is discrete in of this system (visualization method from Sayama

Figure 5
1) An enumeration
of all possible
configurations of a
2-dimensional
binary state CA with
4 cells, 2) author's
visualization, based
on Sayama's (2015),
of the phase space,
attractors, and
attraction basins of
the CA (Wolfram
2015). Each possible configuration of the system is characteristic of the configuration at a larger scale 1983), and 3) a
subjected to the majority rule (in a Moore neighbor- than the individual cell, ideally at a scale at which visualization of
hood in periodic space), and the configuration it goes some coherence or correlation of behavior is appli- isometrically
to in the next timestep is recorded. Then, a graph cable to the design problem at hand. In the third mapping a
is constructed showing each configuration transition portion of Figure 5, we look at the particular char- formation space of
as an edge connecting the two configurations. Each acteristic of length of boundaries between oppos- the CA onto the
component (cluster) in the graph reveals a basin of ing states throughout the system's entirety, isomet- phase space,
attraction in the system, and the hub of that com- rically mapped onto the phase space to form a for- according to the
ponent (center of the cluster) reveals the attractor mation space. In this mapping, we can see that the characteristic of
(see Sayama 2015). The phase space visualization two large basins of attraction now lead to the same boundary length.
for this system shows that it has two equally large attractor, boundary length "0", and that the 7 states
basins of attraction, going to the attractor configu- which used to be completely isolated can now be
rations of fully dark or fully light. There are several grouped together into boundary lengths "2" and "4".
other configurations that have no predecessors and This method of visualization shown in Figure 5 (for
do not change configuration when subjected to the phase space and for formation space) is fine for this
rule. The only way for the system to achieve these very small system. However, the number of possi-
states is to begin there, and, once there, it can pro- ble configurations in a CA system is defined as the
ceed nowhere else (see Sayama 2015). When using number of possible cell states, raised to the power
these systems in architectural design, however, we of the number of cells (Wolfram 1983, Sayama 2015).
are unlikely to be concerned with the internal states So, while this system shown in Figure 5 contains only
of members of the system, which is what the phase 16 (2ˆ4) configurations, a system with only 16 cells
space description is based upon. We are much more would contain 65,536 (2ˆ16) configurations. It is easy
likely to be concerned with the material manifesta- to see how this visualization approach quickly be-
tion left behind as the system progresses. For this comes unmanageable.
reason, we originate the concept of the mathematical Rather than interrogating every possible config-
formation space, as an extension of the phase space uration in the phase space or formation space, we
intended specifically for the realm of architectural de- must find ways to approximate effectively and make
sign. The formation space is an isometric mapping reliable design decisions. The following lays out rea-
onto the existing phase space (hence being consid- soning for the usefulness of the formation space (and
ered an extension). It is a mapping of a particular related concepts), in architectural design, as motiva-

Figure 6
Progression of two
setups. Top setup
reaches its steady
states very quickly,
bottom setup does
not yet reach its
steady states at the
100th step.
Figure 7
The 100th timestep
of 20 different
setups, with 5
separate simulation
runs shown for each
setup. Setups are
indicated by
number of states (s)
and size of
neighborhood
radius (r). This
side-by-side
comparison is
meant as a method
of visual analysis
exploring
dependencies in
the system through
changing variables.

tion for further work to verify usefulness and find ef- randomized initial condition will fall. But, if we look at
fective approximations. For this case study, we in- the formation space, we can reliably project (perhaps,
clude 20 setups of the CA, and conduct 5 simulation in some cases, even guarantee) the attractor that the
runs for each setup. The 20 setups cover each pos- characteristic will fall near to. Therefore, it is feasi-
sible combination if number of possible cell states (s) ble that using this system, we could design low-level
ranges from 2 to 5 and radius size of neighborhood (r) control rules to reliably produce an approximate de-
ranges from 1 to 5 (these setup choices are strongly sired density, number, and spatial scale of the occu-
informed by PyCX Project [12] and Sayama 2015). In- piable areas present on the site.
dividual timesteps of single runs of two of the se-
tups are shown in Figure 6, and the 100th timestep Figure 8
of every run is shown in Figure 7. In Figure 6, we Visualizing the
see that the upper setup settles on its steady states configuration of the
extremely quickly, while the lower setup has not yet steady states of two
reached its steady states even by the 100th timestep. runs of the same
In Figure 7, we see that setups with smaller neigh- setup, and
Probe study into evolving controllers visualizing the
borhood radii reliably reach a steady state with many
If, instead of designing for a particular set of these cri- boundaries
small clusters, with the initial possible states approx-
teria, we desire that the density and scale of these oc- between binary
imately equally represented. We also see that setups
cupiable areas can adjust, just like height, to achieve states in those same
with larger neighborhood radii tend to end with the
shading on the site, then we could subject the con- two steady
entire system in a homogeneous state, with only one
trol rules of the system to evolution. In Figure 9 we configurations.
of the starting cell states represented.
see the results of an example probe study (Thomsen
If we look at the resulting steady states from two
and Tamke 2009) into evolving the controller in sim-
runs of the same setup, such as those in Figure 8, it
ulation (in Galapagos [14]). The evolution of the con-
is clear that it would be impossible to predict which
troller is rewarded for maximizing open area, shad-
state an individual cell will end up in, or which over-
ing that open area, and maximizing length of bound-
all configuration the system will end up, if the start-
aries. The results of this probe (Figure 9) show dif-
ing conditions are randomized and unknown. In this
ferent results on each site, suggesting response to
way, considering the phase space of the system, its
differences in shade provided by neighboring build-
result seems incredibly unpredictable, and perhaps
ings. The first site, shaded only from the south,
not useful for architectural design. However, if we
results in high density of boundary condition, tall
do not consider the cell states, system configuration,
growths, and open areas in the center, where they are
and phase space, and instead consider the formation
shaded by growths. The second site, more shaded
space and spatially-useful characteristics at the scale
by buildings, results in shorter, fewer growths. The
most relevant to the design problem at hand, the
third site, already very shaded, results in low den-
system suddenly seems drastically more predictable.
sity of boundary conditions, and growths that slope
If we look only at the boundaries between states in
downward and outward, thereby providing little ad-
these two simulation results, and consider the overall
ditional shade to open areas. These preliminary out-
boundary length in the system, we can see that they
comes show promise in fulfilling the high-level de-
are very similar. More precisely, they are of lengths
sign requirements of maintaining open occupiable
1,834 and 1,709; incredibly similar for a system with
space and system growths as spatial dividers, and of
10,000 cells. In short, if we look at the phase space
shading occupiable space that is not already shaded
of this system it is difficult, or impossible, to predict
by neighboring buildings, suggesting that the probe
precisely into which basin of attraction an unknown
into evolving controllers merits further investigation.

Figure 9
Results of probe
study, evolving the
CA majority rule
controller in
simulation on three
distinct sites.
CONCLUSION ACKNOWLEDGEMENTS
In conclusion, we have presented a design method- flora robotica has received funding from the Euro-
ology for applying phase space considerations to the pean Union's Horizon 2020 research and innova-
realm of architecture, as an alternative to methods of tion programme under the FET grant agreement no.
averaging and homogenization. We have presented 640959.
the framework of the florarobotica common encod- The authors would like to thank collaborators
ing and integrated projection. We have originated Heiko Hamann and Payam Zahadat for guidance in
the formation space, as a concept to deal with char- swarm robotics and artificial life, and Mohammad Di-
acteristics of active architectural artifacts at a scale vband Soorati for his idea of common encoding. The
that is pertinent to a specific design requirement. We authors would like to thank Yaneer Bar-Yam, Hiroki
have used the formation space as a framework for de- Sayama, and the NECSI Winter School for instruction
sign analysis of case study simulation results in the in complex systems and modelling. The paper relies
context of a specific design task, and used that anal- heavily on the example of Sayama's PyCX Project.
ysis to construct a probe study into evolving con-
trollers in simulation. The results of the probe point REFERENCES
toward the possibility of future work using the forma- Bar-Yam, Y 1997, Dynamics of complex systems, Addison-
tion space to evolve low-level control rules for high- Wesley, Reading, MA
level design tasks, with specificity to individual site. Bar-Yam, Y and Sayama, H 2006 'Formalizing the gene
In the context of the prior flora robotica work on evo- centered view of evolution', Unifying Themes in Com-
lutionary robotics and crossing the reality gap with plex Systems, Springer Berlin Heidelberg, pp. 215-
222
bio-hybrids, this shows the potential usefulness of
Beesley, P (eds) 2006, Responsive architectures: subtle
the formation space in looking at high-level architec- technologies, Riverside Architectural Press
tural design requirements and merging them with Bongard, JC 2013, 'Evolutionary robotics', Communica-
distributed control and distributed communication. tions of the ACM, 56(8), pp. 74-83
Future work includes pursuing substantiation of this Butkovskiy, AG 1990, Phase Portraits of Control Dynamical
specific potential usefulness, as well as general for- Systems, Springer Netherlands
Chan, MTK, Gorbet, R, Beesley, P and Kulič, D 2015
malization, development, and verification of the for-
'Curiosity-Based Learning Algorithm for Distributed
mation space concept.

Interactive Sculptural Systems', Intelligent Robots evolution through evolutionary complexification', J.
and Systems (IROS), 2015 IEEE/RSJ International Con- Artif. Intell. Res.(JAIR), 21, pp. 63-100
ference on, pp. 3435-3441 Thomsen, MR and Tamke, M 2009 'Narratives of Making:
Cook, P and Fournier, C 2004, A Friendly Alien: Ein Kun- thinking practice led research in architecture', pro-
sthaus fur Graz, Hatje Cantz Publishers ceedings of the conference" COMMUNICATING (BY) DE-
Dieckmann, U and Law, R (eds) 2000, The geometry of SIGN
ecological interactions: simplifying spatial complexity, Thomsen, MR, Tamke, M, Ayres, P and Nicholas, P 2015,
Cambridge University Press CITA Works, Riverside Architectural Press
Frazer, J 1995, An Evolutionary Architecture, Architectural Wahby, M, Hofstadler, DN, Heinrich, MK, Zahadat, P
Association and Hamann, H 2016 'An Evolutionary Robotics Ap-
Hamann, H, Wahby, M, Schmickl, T, Zahadat, P, Hof- proach to the Control of Plant Growth and Motion:
stadler, D, Støy, K, Risi, S, Faiña, A, Veenstra, F, Modeling Plants and Crossing the Reality Gap', 2016
Kernbach, S, Kuksin, I, Kernbach, O, Ayres, P and 10th IEEE International Conference on Self-Adaptive
Wojtaszek, P 2015 'Flora Robotica-Mixed Societies and Self-Organizing Systems
of Symbiotic Robot-Plant Bio-Hybrids', Computa- Watson, RA, Ficici, SG and Pollack, JB 2002, 'Embodied
tional Intelligence, 2015 IEEE Symposium Series on, pp. evolution: Distributing an evolutionary algorithm
1102-1109 in a population of robots', Robotics and Autonomous
Heinrich, MK and Ayres, P 2016 'For Time-Continuous Systems, 39(1), pp. 1-18
Optimisation: Replacing Automation with Inter- Werfel, J and Bar-Yam, Y 2004, 'The evolution of re-
active Visualisation in Multi-Objective Behavioural productive restraint through social communication',
Design', Living Systems and Micro-Utopias: Towards Proceedings of the National Academy of Sciences of
Continuous Designing, Proc. of CAADRIA 2016, Mel- the United States of America, 101(30), pp. 11019-
bourne, pp. 373-382 11024
Helbing, D 2010 'Pluralistic Modeling of Complex Sys- Wolfram, S 1983, 'Statistical mechanics of cellular au-
tems', CCSS-10-009, arXiv:1007.2818 [physics.soc-ph] tomata', Reviews of modern physics, 55(3), p. 601
Hurley, M 1982, 'Attractors: persistence, and density of Wolfram, S 1984, 'Universality and complexity in cellular
their basins', Transactions of the American Mathemat- automata', Physica D: Nonlinear Phenomena, 10(1),
ical Society, 269(1), pp. 247-271 pp. 1-35
Kolarevic, B and Malkawi, A (eds) 2005, Peformative Ar- Zahadat, P, Hamann, H and Schmickl, T 2015 'Evolving
chitecture, Routledge Diverse Collective Behaviors Independent of Swarm
Kretzer, M and Hovestadt, L (eds) 2014, ALIVE: Advance- Density', Workshop Evolving Collective Behaviors in
ments in adaptive architecture, Birkhäuser Robotics, GECCO 2015, pp. 1245-1246
Pavliotis, GA and Stuart, A 2008, Multiscale methods: av- [1] http://www.florarobotica.eu
eraging and homogenization, Springer [2] http://www.usa.siemens.com/intelligent-infrastr
Reynolds, CW 1987, 'Flocks, herds and schools: A dis- ucture/assets/pdf/smart-building-white-paper.
tributed behavioral model', ACM SIGGRAPH com- pdf
puter graphics, 21(4), pp. 25-34 [3] http://www.ruiz-geli.com/projects/built/media-t
Sayama, H 2015, Introduction to the modeling and analy- ic
sis of complex systems, Open SUNY Textbooks, Milne [4] http://www.kvarch.net/projects/87
Library [5] http://www.biq-wilhelmsburg.de/
Secretan, J, Beato, N, D Ambrosio, DB, Rodriguez, A, [6] http://www.soma-architecture.com
Campbell, A and Stanley, KO 2008 'Picbreeder: [7] http://www.ahr-global.com/Al-Bahr-Towers
evolving pictures collaboratively online', Proceed- [8] http://www.asif-khan.com/project/sochi-winter-o
ings of the SIGCHI Conference on Human Factors in lympics-2014/
Computing Systems, pp. 1759-1768 [9] http://www.philipbeesleyarchitect.com/sculpture
Soorati, MD and Hamann, H 2015 'The Effect of Fit- s/Sentient-Chamber/
ness Function Design on Performance in Evolution- [10] http://mathworld.wolfram.com/CellularAutomaton
ary Robotics: The Influence of a Priori Knowledge', [11] http://mathworld.wolfram.com/MooreNeighborhood
Proc. of the 24th Int. Conf. on Genetic and Evolution- [12] http://pycx.sourceforge.net/
ary Computation (GECCO ’15), pp. 153-160 [13] http://ironpython.net/
Stanley, KO and Miikkulainen, R 2004, 'Competitive co- [14] http://www.grasshopper3d.com/

Evolving Parametric Models using Genetic Programming
with Artificial Selection
John Harding1
1
University of The West of England
1
john3.harding@uwe.ac.uk
Evolutionary methods with artificial selection have been shown to be an effective

human-computer technique for exploring design spaces with unknown goals. This
paper investigates an interactive evolution of visual programs currently used in
popular parametric modelling software. Although parametric models provide a
useful cognitive artifact for designers to interact with, they are often bound by
their topological structure with the designer left to adjusting (or optimising)
metric variables as part of a design search. By allowing the topological structure
of the graph to be evolved as well as the parameters, artificial selection can be
employed to explore a wider design space more suited to the early design stage.
Keywords: genetic programming, parametric design, artificial selection,

evolutionary design, design exploration
INTRODUCTION applications shows a desire to allow modelling and

Parametric modelling software for constructing vi- analysis to take place in the same integrated environ-
sual programs is now commonplace in architectural ment.
design. Based on dataflow programming, visual pro-
grams define the development of form through a se- Cognitive Artifact
ries of associated explicit functions, commonly tak- A 'parametric schema' defines the development of
ing the form of a Directed Acyclic Graph (DAG). The form explicitly as well as providing a cognitive artifact
structure of the DAG describes a mapping of num- for the design team to interact with. Parametric mod-
bers to geometry, setting out a possible design space elling therefore shifts focus on final form to the devel-
to be explored when parameters are adjusted (Aish opment of form. As Oxman states (2006, 243): "In dig-
and Woodbury 2005). ital design significant processes that have frequently
A combination of parametric modelling and per- been represented as non-explicit in traditional de-
formance analysis tools allow designs to be evalu- sign models must now be considered explicit".
ated both quantitatively and qualitatively in real-time This differentiates explicit programs that have
when adjusting parameters. This process is now well- a defined hierarchy with implicit ones, for example
known, with multiple 3rd party analysis plug-ins (en- those found in emergent systems. Whilst the latter
vironmental, structural, etc.) being developed for vi- can offer novel design approaches and wide design
sual programming tools such as Rhino Grasshopper exploration (Bentley and Kumar 1999), the explicit
and Autodesk Dynamo. The recent increase in such developmental structure between 'seed' and 'design'

is often lost. Parametric models therefore provide a tially be achieved - essentially turning parametric de-
valuable cognitive artifact for design teams working sign software into a combinatorial shape grammar
at the level of process. with a vast selection of common geometric rules al-
ready at hand.
Parametric lock-in Shape grammars in architectural design have
Although a DAG-based parametric model keeps a seen a resurgence in recent years (Grasl and
record of how building geometry is created, display- Economou 2014), both in terms of alternative meth-
ing this explicitly comes at a price. As Aish and Wood- ods of representation using graphs (Grasl and
bury (2005, 11) state: "nothing can be created in a Economou 2011) and in supporting designers who
parametric system for which a designer has not ex- are still in the process of designing and may not yet
plicitly externalised... this runs counter to the often- have a clear shape grammar in mind (Strobbe et al.
deliberate cultivation of ambiguity that appears to be 2015). Whereas these examples tend to use bespoke
part of the healthy design process." As the DAG be- software applications, the motivation for this work
comes increasingly complicated, so its flexibility relies in working with existing visual programming
duces. The graph can quickly resemble a tangle of methods such as Rhino Grasshopper.
spaghetti, making it increasingly inflexible and un- In architectural design, GP has been explored
suitable for the early stages of design where flexibility in the generation of form by Frazer (1995) and later
is most desired (Davis 2013). Coates et al. (1999) using GP with Lindenmayer
Although theoretically an infinite amount of de- systems. More recently evolving networks through
signs can be generated in a parametric system (Ox- augmenting topologies (NEAT) (Stanley and Miikku-
man and Gu 2015), they are bounded by the topo- lainen 2002) have been employed for the exploration
logical structure of the DAG. To date, evolutionary of novel forms (Clune and Lipson 2011).
methods in parametric design have focused on ad- A recent form of GP known as Cartesian Ge-
justing numeric parameters only, for example when netic Programming (CGP) has focussed specifically on
parametric models are combined with metaheuris- evolving DAGs (Miller and Harding 2008). The attrac-
tics such as Genetic Algorithms (GA) or Simulated An- tion of CGP and its applicability in a wide range of
nealing (SA) (Rutten 2013). fields is partly due to its relatively simple develop-
This paper therefore investigates whether a mental encoding from an integer string to combina-
wider parametric design exploration using evolu- torial structures. This form of encoding has inspired
tionary methods can take place by allowing the topo- the method presented here that reduces a Grasshop-
logical structure of the DAG to evolve in addition per parametric model to a part-integer, part-floating
to the metric parameters? This would help counter point genotype, with the associated phenotype a
parametric lock-in, opening up design possibilities at parametric model.
the conceptual stage by working at a higher level of
abstraction (Harding et al. 2012). Embryo
Embryo is the name of a Rhino Grasshopper plug-in
GENETIC PROGRAMMING FOR DAGS developed by the author that maps a number string
DeLanda (2002, 11) states that if architectural designs genotype to a complete parametric model pheno-
are to enjoy the same degree of combinatorial pro- type. The choice of development within Grasshop-
ductivity as biological ones, we must begin to "think per was partly due to its current popularity. As Vier-
topologically" and avoid too narrow a design search. linger and Bollinger state (2014, 609): "Grasshopper
By applying Genetic Programming (GP) techniques unveils algorithmic design to non-programmers with
on DAG-based parametric models, this can poten- intuitive interfaces". Its organisation of common (and

Figure 1
Encoding process
used by Embryo
when generating a
Grasshopper
definition.
bespoke) explicit functions into 'components' make simple example. A component pool is specified (a),
them similar to shape grammar rules, and the poten- which in addition to the genotype is used by the Em-
tial for combining both human and computer gen- bryo component (b) to generate a parametric model
erated DAGS and analysis tools under one platform (c). The metric genes are mapped directly to the nu-
make it an attractive choice. meric sliders. The functional (shown blue) and topo-
Constructing a graph in Grasshopper can be split logical (shown red) aspects of the genotype indicate
into three categories: External parameters (for exam- how each component is chosen and its inputs are
ple numeric sliders, external geometry, etc.), the pool then connected back to an output respectfully. A
of components in the graph (nodes) and the topolog- form of parameter datatype matching ensures wires
ical structure that forms associations between com- are connected to relevant inputs and outputs.
ponents (edges). These three categories form the ba- Further details on the method used can be found
sis of the genotype used by Embryo when construct- in Harding (2014) which includes the current limita-
ing a parametric model: tions with the approach. This include aspects such
as handling multiple associations per component in-
1. Metric genes: control the parameter values put which is yet to be included. Aspects such as deal-
for generated sliders and have a direct numer- ing with parameter datatype matching and more set-
ical mapping. These can be either integer or tings such as removing unsuccessful components is
floating point values. These metric parame- also discussed but will not be elaborated on here.
ters are the first things generated by Embryo. Clearly the combinatorial possibilities with such
2. Function genes: when a component is added a process is huge, hence a brute force approach for
to the graph, the function genes controls the exploring the design space meaningfully is difficult
type of component is selected from the pool. and other metaheuristics should be considered that
3. The topological genes are integer based and help counter bloat, a known problem when using GP
map the output location for each component (Miller 2001). The explicit embryogeny of the DAG
input when forming the graph. Altering these encoding makes it suitable for evolutionary methods
genes changes the topology of the graph. such as genetic algorithms (Kumar and Bentley 2000),
where evolvability is also a key concern.
If enough genes are not provided then genes are re- The CGP encoding method means that during
peated. Furthermore, if the value of genes are too mutation and crossover, aspects of the original def-
high for the current state of the system then modular inition are maintained. As the process is effectively
arithmetic is used. Using a large gene pool negates evolving a visual program itself, this essentially trans-
this but requires more memory. lates to a form of genetic programming.
Figure 1 gives an outline of the process with a

EVOLVING PARAMETRIC MODELS (Simulated Binary Crossover) and mutation (point)
Evolutionary algorithms commonly use an objective are all utilised. Hoopsnake by Yannis Chatzikonstanti-
function to evaluate designs at each generation. Ar- nou [1] facilitates feedback loops within Grasshop-
tificial selection however replaces this function with per. This is activated following selection, allowing the
selection by human participant(s) who need not population to evolve and pause for user input at each
make their motivations explicit and may change dur- generation.
ing the evolutionary process (Dawkins 1986). Such
Interactive Evolutionary Computation (IEC) methods Figure 2
are capable of exploring complexity without requir- The evolutionary
ing human understanding of the specific process in- process with
volved (Sims 1991, 328). associated
Leaving the objective function open to the de- Grasshopper
sign team allows for qualitative or intangible drivers plug-ins
to be included in the design search. That is not to say highlighted in red.
that quantitative performance criteria cannot be in-
cluded, only that the human participant(s) becomes
an active part of the design search, guiding the evo-
lutionary method through the paradox of choice at
the early design stage (Piasecki and Hanna 2011).
Judgement during evolution therefore takes
place on two forms of phenotypic representation,
both the parametric schema (development model)
and the building model (generated design). The de-
sign team can (if they choose) steer the process to
evolve parametric models that are legible, or capable
of being understood. Maintaining this engagement Figure 2 shows the overall process with associated
gives the evolved models a chance to be further de- special components highlighted in red. An initial
veloped manually following an evolutionary process. population of genotypes and associated parametric
model phenotypes are randomly generated. These
Grasshopper Setup can be viewed by the designer (both in terms of the
Grasshopper proved to be a suitable environment for DAG and generated geometry) and two elite parents
combining Embryo, a metaheuristic algorithm and selected for the next generation. Crossover and mu-
an artificial selection method. By explicitly formulat- tation is then applied to form the next generation.
ing the search algorithm in the same environment The associated implementation in Grasshopper
as the problem is defined, a more flexible reaction is given in Figure 3. Below the horizontal threshold
to special requirements is possible (Vierlinger and line is the metaheuristic process as per the previous
Bollinger 2014, 611). This meant that several third diagram. This includes the manipulation of the popu-
party components could be used to construct an in- lation of genotypes and their use by the Embryo com-
teractive evolutionary approach. ponent (shown bottom right). The parametric mod-
Octopus Explicit (Vierlinger and Bollinger 2014) els generated by the machine are located above the
is a set of tools for forming customised evolution- threshold line. The components used in the gener-
ary algorithms. In this implementation the genotype ated graph are selected from a pool, shown top-left.
format (Octopus solution), components for crossover These could be any Grasshopper components, either

Figure 3
The setup in
Grasshopper. The
model above the
threshold line is
machine-
generated.
Figure 4
Three bespoke
Grasshopper
components for the
Tower Hamlets
project.
standard ones or those created by third parties. Project Application

Note that in this example, some human placed In collaboration with 3DReid Architects (London),
components have also been incorporated during Embryo was originally used to generate paramet-
graph generation, something that Embryo can han- ric models for a residential project in Tower Ham-
dle with outputs being tagged (*). This is one of the lets (Harding 2014). Shape grammar rules were em-
main benefits of using a single environment, as de- bedded as Grasshopper components (Figure 4) with
signers can begin to work at different levels of ab- heuristics both specific for this project and based on
straction (i.e. crossing the threshold). the experience of the architect.

This included two types of residential massing block idential blocks to improve connectivity and contrast
of set width and variable length: one with a set min- the orthogonality of the block layout (c).
imum height (a) and one with an additional vari- The original process involved mostly using a ran-
able controlled by the 'add storeys' input (b). Planes dom seed to generate designs, but only small alter-
for connecting the massing blocks were set as input ations could then be made manually. Essentially, the
and output parameters, allowing different but con- process whilst useful became a task of selecting a par-
strained permutations to be explored. An additional ticular generated model as opposed to developing
third component formed a link bridge between res- one iteratively using cumulative selection.
Figure 5
Six generations of
development.
Gross internal area
is shown for each
design.

In light of this, the original project was therefore re- the design team that can potentially influence the se-
visited, adapted into the evolutionary method as dis- lection process.
cussed in the previous section and a new trial con- Designs with around 40,000m2 GIA based on the
ducted. The generated parametric models therefore number of residential units were deemed to be desir-
lie somewhere in-between a human and machine ap- able. The massing context could be easily referenced
proach, with the process becoming more of a conver- in during the design search due again to the single
sation with the machine able to suggest alternatives Rhino/Grasshopper environment.
but with neither party dictating development. The results of a typical run is shown in Figure
5 with each design having an associated parametric
Results model phenotype (Figure 6). Each design is labelled
In the trial, an initial random population of 8 designs A-H for each generation. In this example, only six
was presented to the design team. The paramet- generations are shown but the process could in the-
ric models generated began with 6 numeric param- ory continue indefinitely. Two parent designs are se-
eters and 12 components. A crossover and muta- lected from each generation as shown highlighted.
tion rate of 0.7 and 0.1 respectively was used. During Some of the motivations behind the selection at
the search, quantitative analysis could be conducted each generation were recorded during the process:
again within Grasshopper. In this simple example, us-
able floor area (as a function of volume) and a mea- 1. First choice simply based on potential for fu-
sure of heat loss (surface area / volume) is offered to ture development.
2. Option D chosen due to the introduction of a
Figure 6
Associated
parametric model
topology at each
generation.

Figure 7
The evolved
definition is further
modified and
shown in context.
block in the x-axis. Option B chosen due to in- Clearly these are simply massing models and some
teresting array of link bridges of the choices made are based on relatively simple
3. Option D again for the same reasons. Option criteria, but the potential for including graph devel-
(F3) with non-working components not se- opment as part of the search has potential if the leg-
lected. Parameter domains reduced and com- ibility of the graph can remain. The clean structure
ponent number increased to 16. Option G se- of machine generated parametric models can help
lected due to possible courtyard space at cen- with this. In reality there is a balance to be struck
tre. between having too few metric parameters to un-
4. Options D&E appear similar with slightly dif- derstand their influence and too many to keep track
ferent parameters. D selected. GIA needs of. As Davis (2013, 76) states: "An ideal parametric
reducing to brief requirement influencing model would encompass all the variations the de-
choice of H. signer wants to explore within the smallest dimen-
5. Heat losses getting quite high hence Option sionality possible."
H selected. Elevated elements providing in-
terest. Ramp on A a possible entrance to CONCLUSIONS
a centrally located building away from exist- This research shows how a combination of paramet-
ing massing. Courtyard created between new ric modelling tools with genetic programming can
buildings and existing. explore wider design spaces in architectural design
6. Option F now has ramps from one location whilst retaining a parametric definition. By opening
to two blocks. Heat loss performance has re- up the topology of the graph as part of the search,
covered. GIA is a little low but further adjust- parametric models are generated as opposed to ad-
ments can be made. justment of numeric parameters by the metaheurisic.
Using an interactive evolutionary algorithm, human-
The motivations during the search were both qual- computer interaction can be used to enhance the
itative and quantitative, and constantly subject to search even when the problem cannot be defined.
change. Once the final design was selected, a para- The example given has shown potential, but has
metric definition exists (as opposed to a CAD model) revealed a number of limitations that need address-
could be taken forward for further development. Fig- ing in future work:
ure 7 shows how the model was developed in terms
of parameter adjustment (a), alterations to the graph • At present only one parametric model can be
(b) and enhancing the massing model (c). The GIA viewed at a time (Grasshopper is not designed
could therefore be increased to the brief requirement for this type of use). Further interface devel-
parametrically. opment is therefore required to display mul-

tiple models in each generation simultane- REFERENCES
ously. Aish, R and Woodbury, R 2005 'Multi-level interaction in
• All generated numeric parameters have the parametric design', Smart Graphics, pp. 151-162
same range giving no hierarchy of scale ap- Bentley, P and Kumar, S 1999 'Three ways to grow de-
signs: A comparison of embryogenies for an evolu-
propriate to the components they are associ-
tionary design problem', Proceedings of the genetic
ated to. and evolutionary computation conference, pp. 35-43
• As discussed in Section 3, multiple edges per Bittermann, MS 2009, Intelligent Design Objects (IDO): A
component input is yet to be implemented. cognitive approach for performance-based design, TU
• It was found that crossover caused a larger Delft, Delft University of Technology
than expected disruption to the parent Clegg, J, Walker, JA and Miller, JF 2007 'A new crossover
technique for cartesian genetic programming', Pro-
graphs. This was likely due to the CGP method
ceedings of the 9th annual conference on genetic and
of encoding and using mutation alone may be evolutionary computation, pp. 1580-1587
more suitable approach in future in line with Clune, J and Lipson, H 2011, 'Evolving 3d objects with a
other CGP applications (Clegg et al. 2007). generative encoding inspired by developmental bi-
• Bloat can still be an issue and generated para- ology', ACM SIGEVOlution, 5(4), pp. 2--12
metric models can sometimes freeze. Incor- Coates, P, Broughton, T and Jackson, H 1999, 'Explor-
ing three-dimensional design worlds using Linden-
porating a maximum calculation time and re-
mayer systems and genetic programming', in Bent-
moving any model that does not compute ley, PJ (eds) 1999, Evolutionary design by computers,
from the gene pool is a possible solution. Morgan Kaufmann, San Francisco, pp. 323-341
• Facilitating groups of components in the pool Davis, D 2013, Modelled on software engineering: flexi-
(i.e. clusters). ble parametric Models in the practice of architecture,
• User fatigue is a known issue with interactive Ph.D. Thesis, School of Architecture and Design,
RMIT University
evolutionary computation. Kruse and Conor
Dawkins, R 1986, The blind watchmaker: Why the evidence
(2015) have shown how intelligent agents can of evolution reveals a universe without design, WW
be used in conjunction with an IGA. Norton \& Company
DeLanda, M 2002, 'Deleuze and the use of the genetic al-
As well as addressing these limitations, future related gorithm in architecture', Architectural Design, 71(7),
research includes using the technique to evolve para- pp. 9-12
metric models for existing CAD geometry in order to Frazer, J 1995, An evolutionary architecture, Architectural
suggest new design directions for models that are Association Publications
inflexible. Here a combination of target based evo- Grasl, T and Economou, A 2011 'GRAPE: Using graph
grammars to implement shape grammars', Proceed-
lution is more suitable, perhaps using other meta-
ings of the 2011 symposium on simulation for architec-
heuristics such as simulated annealing. ture and urban design, pp. 21-28
Grasl, T and Economou, A 2014 'Towards controlled
ACKNOWLEDGEMENTS grammars: Approaches to automating rule selec-
tion for shape grammars', Proceedings of the 32nd
The author would like to thank 3D Reid Architects for
eCAADe Conference, 2014, pp. 357-363
their support during this research. This work is part- Harding, J 2014, Meta-Parametric Design: developing a
sponsored by the 2016/17 UWE VC Early Career Re- computational approach for early stage collaborative
searcher Development Award. practice, Ph.D. Thesis, University of Bath
Harding, J, Joyce, S, Shepherd, P and Williams, C 2013,
'Thinking Topologically at Early Stage Parametric
Design', in Hesselgren, L, Sharma, S, Wallner, J, Bal-
dassini, N, Bompas, P and Raynaud, J (eds) 2013, Ad-
vances in Architectural Geometry 2012, Springer Vi-

enna, pp. 67-76
Kruse, J and Connor, A 2015, 'Multi-agent evolution-
ary systems for the generation of complex virtual
worlds', EAI Endorsed Transactions on Creative Tech-
nologies, 2(5), p. Paper e5
Kumar, S and Bentley, PJ 2000 'Implicit evolvability: an in-
vestigation into the evolvability of an embryogeny',
Late Breaking Papers at the 2000 Genetic and Evolu-
tionary Computation Conference, Las Vegas, Nevada,
USA, pp. 198-204
Miller, J 2001 'What bloat? cartesian genetic program-
ming on boolean problems', 2001 Genetic and Evolu-
tionary Computation Conference, pp. 295--302
Miller, J and Harding, S 2008 'Cartesian genetic pro-
gramming', Proceedings of the 2008 GECCO confer-
ence companion on Genetic and evolutionary compu-
tation, pp. 2701-2726
Oxman, R 2006, 'Theory and design in the first digital
age', Design Studies, 27(3), pp. 229-265
Piasecki, M and Hanna, S 2011, 'A Redefinition of the
Paradox of Choice', in Gero, JS (eds) 2011, Design
Computing and Cognition 2010, Springer, pp. 347-
366
Rutten, D 2013, 'Galapagos: On the logic and limitations
of generic solvers', Architectural Design, 83(2), pp.
132-135
Sims, K 1991 'Artificial evolution for computer graphics',
Computer Graphics 25(4), Siggraph '91 Proceedings,
pp. 319-328
Stanley, KO and Miikkulainen, R 2002, 'Evolving neural
networks through augmenting topologies', Evolu-
tionary computation, 10(2), pp. 99-127
Stiny, G and Mitchell, WJ 1978, 'The palladian grammar',
Environment and Planning B, 5(1), pp. 5-18
Vierlinger, R and Bollinger, K 2014 'Accommodating
change in parametric design', Proceedings of ACADIA,
pp. 609-618
[1] http://yconst.com/software/hoopsnake/

GENERATIVE DESIGN | Applications
Homeorhetic Assemblies
Turning beehive formation dynamics into high-res tectonics
Roberto Monesi1 , Alessio Erioli2

1
università di Bologna 2 Università di Bologna
1
www.emergent-behaviors.blogspot.com
2
www.unibo.it/sitoweb/alessio.erioli
1
mone.monesi@hotmail.it 2 alessio.erioli@unibo.it
This thesis research investigates the architectural and tectonic potential that can
stem from behavioral complexity of collective construction in biological systems
and its dynamic relations with the colony in terms of the continuous construction
and adaptation process over the time. The role model considered as a case study
regards the dynamics of honeycomb formation, and in particular three
fundamental behaviours have been extracted from this biological process:
stigmergic behavior, structural self-stabilization capacity and environmental
adaptability. All these features were then coded into a multi agent system
interacting in an heterogeneous environment and capable of selectively adding
elements to a particle-spring system that is periodically self-adjusting, simulating
material behavior. The outcomes, strongly rich and heterogeneous in their spatial
organization, are characterized by a continuous tectonic of emerging
singularities seamlessly flowing into one another.
Keywords: tectonics, beehive, stigmergy, multi-agent system, robotics
INTRODUCTION gle elements that share the same shape and mate-
Homeorhetic Assemblies begins from the study of rial. The analyzed systems are characterized by the
self-organizing biological systems. The reason for ability to create highly organized structures and effi-
this interest is rooted in principles of ecology, here cient complexity from simple rules and a single ma-
intended as the study of the set of economies (ex- terial, harmonizing form with material organization
change relations) between an organism and its envi- and physiology in a condition of homeorhesis (a sta-
ronment, and efficient resource distribution patterns ble dynamic of development). Their morphological
are a cornerstone of evolutionary success of a sys- and physiological complexity are not the outcome of
tem. An ecology applied to design and contempo- a carefully thought and executed general plan, they
rary fabrication technology beyond mere formal im- rather emerge from the interaction in space and time
plications (i.e. imitating shapes) in the search for ar- of a large number of individuals with elementary be-
ticulated spatial complexity and heterogeneous per- haviors and their ability to perform acts of construc-
formance with a high-resolution fabric made of sin- tion (pick and place, pour, weave, etc.) engaging also
GENERATIVE DESIGN | Applications - Volume 1 - eCAADe 34 | 435

the morphogenetic capacities of a variety of material ment of deposition the material is in viscous state (to
systems. be worked more easily) and the cells are built accord-
ing to the size and shape of the bees' body. The re-
ROLE MODEL sult is a closest-packing pattern composed of cylin-
The role model considered as a case study regards drical cells all similar in size to each other because the
the dynamics of honeycomb formation (see Figure size of worker bees is equally similar. After short time,
1). The honeycomb stereotype, a grid of perfectly during the passage of the wax from viscous to solid
hexagonal cells, is just a particular case of forma- state, the tension in the material tends to change the
tion on a flat surface in homogeneous conditions; cell's original shape, concentrating efforts along the
nonetheless, it has too often been misinterpreted as lines of contact between the various cylinders; from
a general design intent from the bees. This is one of these interactions and the closest packing arrange-
such cases (phyllotaxis is the typical other case) when ment emerges a cell shape that tends to the char-
representation cuts the links between a pattern and acteristic hexagonal pattern. This material capacity
the dynamics of its formation. Cell shape, size and de- becomes essential for the entire system to optimize
ployment depend instead from its primary function stress distribution patterns and ensure great stability
(containment), size and body shape of the construc- to the whole structure.
tor agent and other factors unfolding over time such
as coordination among agents and material behav- STIGMERGIC BEHAVIORS
ior and the variability of environmental conditions Camazine et al.(2003) demonstrates how the way
(Tautz 2004).Three fundamental behaviors have been in which the bees realize their natural hives can be
extracted from the role model: morphogenetic ca- considered as a self-organized biological system, in
pacity of the material, stigmergic behavior and envi- which the resulting structure emerges from the ac-
ronmental adaptability during the growth process. tivity of many individuals, none of which is a leader or
has plans to follow. A fundamental aspect for study-
MORPHOGENETIC CAPACITY OF THE MATE- ing the self-organizing systems is to understand the
RIAL way in which the agents interact with each other. On
Processes and interactions that lead to the realiza- this topic Camazine et al. confirmed the studies pub-
tion of a beehive are multiple and unfold continu- lished by Darchen and Darchen (1994), that demon-
ously in space and time. The physical parameters, strates how communication between the bees dur-
such as mechanical characteristics and their variation ing the construction process happen via stigmergy
in time play a fundamental role in the system. In the (indirect coordination among agents by means of
case of the hive, in every instant of the growth pro- a trace in the environment). The potential of this
cess the structure is stable, and this is possible not type of interaction is evident in many biological sys-
only because self-organization promotes stable pat- tems (such as termite mounds) and has been care-
terns (Camazine et al. 2003), but also because the ma- fully studied in computing by Johnson (2009). Dur-
terial is able to adapt (modifying its shape, its den- ing the hive construction each bee deposits a small
sity and its consistency) to the efforts and tensions amount of wax, in adjacency to the already built cells,
to which it is subjected. Tautz (2004) demonstrates which can be used to continue the construction. The
these morphogenetic capabilities investigating the wax acts thus as an attractor for the other insects
processes necessary for the realization of the hon- and this material directs the building development
eycomb, which is defined as emergent result from process. Stigmergy, then, is used to regulate the se-
the interaction between the function of the cells, the quence and the pattern of construction start points
construction technique and the material. At the mo- as well as affecting the bee trajectories and allows the
436 | eCAADe 34 - GENERATIVE DESIGN | Applications - Volume 1

Figure 1
Example of natural
beehive: these
images from web
([1],[2],[3],[4]) show
the emergence of
the typical
hexagonal pattern
and the great
capacity of the
system to adapt to
the environment.
suitable material concentration for the system devel- possibilities for the system increasing its stability.
opment.
BEHAVIORAL MODELS
ENVIRONMENTAL ADAPTATION All the features of the bees' works were coded into
Another typical property of self-organizing systems, a multi agent system, based on a large number of
and fundamental characteristic of the beehive, is agents which shared behaviours can be divided into
their environmental adaptability. The colony, in fact, three families: physical and mechanical properties,
is able to interact with the environment and eval- stigmergy and environmental adaptation. Since the
uate the principal parameters for its development. final intent is to transfer the process to a different
Darchen and Darchen (1994) demonstrate this great scale and means of fabrication (robotic driven fil-
adaptability studying the system behaviors in rela- ament deposition) the cell as elementary unit for
tion to some essential environmental factors. The growth has been substituted with linear segments.
first one is temperature: bees are extremely suscepti-
ble to these parameter variations (especially larvae) PHYSICAL AND MECHANICAL PROPERTIES
and for this reason the hive is developed along ar- All the behaviors and properties used to simulate the
eas with very low thermal variation. Effect of airflow creation of physical structures and their features are
is the second environmental parameter considered: included into this family, and their influence on the
beehives, in fact, in the case of persistent flows, tend system is visible in Figure 2.
to follow the main directions of the wind; in this way
Material deposition and self-stabilization. Each
the currents are conveyed and thermal fluctuations
agent combines its movement with a continuous
reduced, while also limiting orthogonal stresses on
analysis of the environmental conditions. Whenever
the honeycombs. Another typical characteristic of
suitable conditions are met, it deposits a physical
many self-organizing systems is their ability to take
structure in which every particle is connected to the
advantage from the potentialities that the environ-
nearest among the already existing ones through
ment offers: bees, during the hive construction, are
springs. Thus, the structural system created (which
able to wrap, incorporate or attach to any part of
has specific physical and resilience characteristics)
the environment that is able to afford that operation,
has a tendency to continuous self-stabilization, tend-
thereby transforming them into anchor points for the
ing to a stable configuration. The structure, then,
structure. This behavior allows more development

turns out to be in continuous transformation under Repulsive charges. To ensure the continuity of the
the effect of the external forces. interstitial voids each agent also settles repulsive
charges that guarantee a minimum distance be-
Figure 2
tween different honeycombs. When this distance is
Examples of
not respected the structures tend to connect by in-
morphological
creasing the local stiffness.
variations
STIGMERGIC BEHAVIORS dependent on
Stigmergic behaviors are a fundamental part of the physical parameters
coded system because (although indirectly) they are such as the
the only form of communication between agents. maximum length
and minimum
Cohesion. The first behavior concerns the interac- percentage of
tion of the agent with the particles that are set at a elongation of the
medium-long distance from its actual location. Each springs.
agent is attracted by the visible portion of the already
stabilized material. This behavior affects directly the
agent's movement and its deposition behavior; it is
essential to trigger material concentration (which is
typical of stigmergic systems) and avoid dispersion
(see Figure 3).
Strength increase in time. A fundamental charac- Figure 3

teristic of the studied role model is the alteration of Examples of
the physical and mechanical features of the structural morphological
elements over time. To implement this in the simula- variations caused
tion, each spring increases its stiffness over time. The by the changes of
simulated hardening produces a greater diversifica- the parameters of
tion of the structural behaviour. The result is a de- stigmergic
formability that decreases over time and the progres- cohesion (visual
sive transformation of elements from elastic springs range and
into rigid bodies. Over time, the older parts of the magnitude).
structure become less and less compliant to physi-
cal stress, freezing the particles into place and lock-
ing their surrounding region in a fixed configuration,
allowing the overall development of the structure.
Double section deposit. To ensure adequate struc-
tural redundancy but simultaneously for limit the
number of unnecessary connections each agent de-
Separation. The second stigmergic behavior, in-
posits the material in the form of double cubic sec-
stead, regards the agent's interaction with the nearby
tion. This allows to reduce the maximum length of
particles. Below a certain proximity threshold, the
the springs and, at the same time, also to increase the
particles generate a repulsion force towards the
global structural rigidity.
agent, which will tend to move away (see Figure 4).
This behavior is fundamental for the growth of the

whole system, as it prevents the agents from pen- sary for the construction. As a consequence, the gen-
etrating into the already deposited structures, opti- eral development of the structure has a tendency to
mizing their work and permitting a greater overall de- follow field isovalues.
velopment. Interaction with pre-existing structures. To allow
the adaptation to wide variety of environmental con-
Figure 4
figurations and spaces, each agent is able to read po-
Examples of
tential obstacles and modify its own movement ac-
morphological
cording to them, so as to avoid them. In case there
variations caused
are still the right conditions, each agent continues to
by the changes of
deposit material. The particles, if they are in close
the parameters of
proximity to the obstacle, will adhere to it, by con-
stigmergic
straining to it their own position and becoming an-
separation (visual
chor points for the whole structure.
range and
magnitude). Voids generation. In order to increase the tectonic
and architectural potential of the system and intro-
duce the possibility of design choices, the agents
have been equipped with additional behaviors than
those in the role model. They are able to read vol-
umes and repulsion points (created for design pur-
pose) and change their motion in order to respect
these spaces. These behaviors allow the designer
a possibility of interaction with the system, with-
Environmental adaptation out limiting the properties of self-organization. The
The ability to read and analyze space is a fundamental structure, in fact, in the stabilization phase remains
characteristic for the system's environmental adapta- free to reach its optimal configuration.
tion. In order to make the simulation as plausible as
Material optimization. The deposited structure, be-
possible, the environment (modeled as a finite and
sides tending to stability in every moment of growth,
discrete voxel grid) is considered anisotropic and it is
it is also continuously tested in terms of local and
characterized by different fields (vector and scalar).
global deformation by checking the elongation per-
Vector field. Each agent is able to read the strength centage with respect to its rest length. If this value is
and the direction of the field at its current loca- negligible (which means so is its load), the evidently
tion. This vector component has a direct influence unnecessary element is removed, so as to limit the
on the agent's speed and, consequently, on its move- number of springs to those strictly necessary. From
ment. By altering the agent's movement behavior, a global point of view, in the case of great instability
vector fields have an indirect influence on the de- of more springs, the particles, during their collapse
posited material highlighting constructive patterns. will approach to other parts of the structure and they
These patterns tend to canalize the principal currents will tend, if it possible, to connect to the more stable
and minimizing the horizontal actions acting on the elements.
structure.
Scalar Field. The scalar fields simulate the function Tectonic research
of temperature in nature: each agent reads the sur- The encoding of the multi agent system required
rounding environment and tends to move towards a careful evaluation especially of relations between
areas characterized by values similar to those neces- the introduced behavior (see Figure 5). To do this,

Figure 5
Behaviours
relationship
diagram: the
picture shows
schematically the
behaviours of each
agent, and as the
process that
emerges from these
interactions it is
not-linear.
Figure 6
Examples of
different structural
configuration
emerging from
variations in the
agents behavior
(with increasing
complexity).

at the introduction of any change in the system we tural continuity and to ensure greater stiffness
have generated several phenotypes (see Figure 6) at the node, in the beginning of each spring
based on the properties introduced until then (geno- has been imposed to the robot a small circu-
types). The study of these partial results allowed us lar motion, for increase the deposition of ma-
to progress in our research on continuous tectonics terial and allow a better connection between
until arriving at architectural scale. different elements.
FABRICATION SIMULATION
During the entire research we considered the possi- ARCHITECTURAL APPLICATION
bility to build a prototype through the use of a 6 axis To demonstrate the potentiality of homeorhetic as-
robotic arm and the realization of a nozzle for plas- semblies, the system described was applied to ar-
tic deposition. This step preceded by the study about chitectural scale. To do this it was necessary to de-
hardware (robot and extruder) and material (ABS and fine the boundary conditions. In addition to the
PLA) it required the reorganization of deposition pat- anisotropic space in which to run the simulation
tern through the following phases (see Figure 7). (characterized by scalar and vector fields similar to
those found in reality), two trajectories were de-
1. Initially all the springs have been ordered ac- signed. The mesh constructed from these curves rep-
cording to the Z coordinate of their starting resents the spaces to be respected, and to be left
point (the one on the lower level). empty. In this way, the fundamentals volumes for
2. Then the possible intersections were detected the generation of the pavilion are imposed by the
(between the elements and between ele- designer. The last required inputs were the start-
ments and the robotic arm during its move- ing points from which the structure starts to growth.
ment). After identifying the collisions, the During the growth of the system a lot of information
affected springs were divided at the critical emerge (in the form of data such as stiffness, defor-
points. mation, etc ...) and show the constant interactions
3. At this point all the springs (in an higher num- between the three main subjects involved: the envi-
ber) have been re-ordered according to the Z ronment, the agents and the structure that they re-
coordinate of the starting point. alize. The extraction of these data makes it possi-
4. At the end of this process, to restore the struc- ble to verify the emergence of stabilization pattern,
Figure 7
Above: principal
phases for the
reorganization of
the deposition
pattern. Sorting,
collision detection
and new sorting.
Under: digital
simulation of the
fabrication process
for the prototype.

Figure 8
Patterns that
describe the
changes of
mechanical and
physical properties
of the system
during its growth.
These images show
the evolution of the
deformations (from
green to red) and
the stiffness for the
each springs (from
cyan to magenta).
Figure 9
Principal steps of
the growth of
system.

Figure 10
Top and side view
of the complete
structure.
Figure 11
Side views of one of
the resulting
configurations.
which are fundamentals for the global development the system and the introduction of voids and spaces
(see Figure 8). At the end of the process (see Figure that will be observed by agents during the simula-
9) the coded system produces a morphology that is tion.The outcomes can be defined as high-res con-
not predictable, and it reveals interesting peculiari- tinuous tectonics that appear extremely redundant
ties from an architectural point of view (such as the and heterogeneous in their spatial organization. The
creation openings, cavities etc...) (see Figure 10). The heterogeneous density and the spatial organization
result is then a complex structure, redundant and or- emerge from the process of self-organization and en-
ganized from a material point of view, characterized tail the development of continuous tectonics, cre-
by a continuous tectonic and able to also generate ated for aggregation (and not for juxtaposition) of
a good spatial continuity (see Figure 11). The indi- many elements all similar to each other that undergo
vidual structural elements (all with same shape and a transformation so that the whole is characterized by
material) assume different values depending on their a set of emerging singularities (i.e. surface, column-
position generating different gradient of density and like and truss-like conditions), seamlessly flowing
permeability. into one another. High redundancy is a deliberate
choice, enabled by the possibilities of automating
CONCLUSION construction through the behavioral programming
At the design level, the process resulting from this of robots: a 6 axis robotic arm (for which a custom
research is an example of coexistence between nozzle for plastic deposition is currently being de-
bottom-up and top-down approach. Bottom-up veloped) will be used for prototyping. The introduc-
strategies are the main features: they are manifested tion of this further constraint required the reorgani-
by stigmergic self-organization, the morphogenetic zation of the deposition pattern and had consistent
capacity of material, the environmental adaptation feedback also in the computational simulation phase,
and they highlighting pattern of stability. The top- while not affecting the quality and complexity of the
down strategy, instead, allow us to influence the ex- outcomes. Possible future developments will cer-
pression of self-organization, through the setup of tainly be targeted at overcoming the two main issues

addressed in this research. The first one is the fabri-
cation (especially in large-scale) achievable by opti-
mizing the hardware and using them to prototyping
more elements assembled together, or even through
the use of more robotic arms that can be used in par-
allel. The second fundamental step will be the over-
coming of computational limits appeared in this re-
search. In fact, in these simulations many emerging
effects were limited by the low number of agents and
structural elements (each simulation involved about
200 agents and up to 300,000 springs). The use of a
single CPU therefore seems to be the main computa-
tional bottleneck. However, these problems can be
overcome through the use of technologies such as
GPU or cloud computing. Homeorhetic Assemblies is
an attempt at the application of continuous tectonics
and an exploration of their aesthetic consequences; it
constitutes an example of how they can be harnessed
for the complex articulation of architectural spaces
through high resolution assemblies and material be-
haviors coordinated by the intricate interaction of el-
ementary agents endowed with simple rules.
REFERENCES
Camazine, S, Deneuborg, JL, Franks, NR, Sneyd, J, Ther-
aulaz, G and Bonabeau, E 2003, Self-Organization
in Biological Systems, Princeton University Press,
Princeton
Darchen, B and Darchen, R 1994, Il fantastico mondo delle
api, Piccoli, Torino
Johnson, BR 2009, Pattern formation on the combs
of honeybees: increasing fitness by coupling self-
organization with templates, Division of Biological
Sciences, university of California, San Diego
Tautz, J 2004, 'Honeybee combs: construction through
a liquid equilibrium process?', The Science of Nature,
Naturwissenschaften(91), pp. 350-353
Tautz, J 2008, The Buzz about Bees: Biology of a Superor-
ganism, Springer Science & Business Media, Berlin
[1] http://rsif.royalsocietypublishing.org
[2] www.sawattsdesign.com
[3] www.flickr.com/photos/max_westby
[4] https://twitter.com/LondonBeeKeeper

Monolith: The Biomedical Paradigm and the Inner
Complexity of Hierarchical Material Design
Panagiotis Michalatos1 , Andrew Payne2
1
Harvard Graduate School of Design / Autodesk, Inc. 2 Autodesk, Inc.
1,2
{panagiotis.michalatos|andrew.payne}@autodesk.com
This paper discusses our ongoing research into hierarchical volumetric modeling
and the external forces which are motivating a shift from the traditional boundary
representation (also known as BREP) that has thus far dominated design software
toward a more flexible voxel-based representation capable of describing complex
variable material distributions. We present Monolith; a volumetric modelling
application which explores hybrid forms of digital representations and new
design workflows that extend a designer's ability to describe the material
properties of a 3d model at the mesoscopic and even microscopic scales. We
discuss the inherent complexities in volumetric modelling and describe the design
opportunities which heretofore were unavailable using existing techniques.
Keywords: hierarchical materials, multi-material 3d printing, voxels
INTRODUCTION more than assemblies of discrete components (see

Two technological developments are forcing us to Figure 1). Most design software used by architects
reevaluate the paradigms we use in computer aided has its origins in mechanical engineering, and often
design (CAD). The first is the emergence of new implies an assembly paradigm inherited from the in-
fabrication techniques that enable greater control dustrial revolution. Yet, the needs of the biomedical
over the deposition and mixing of materials at ever industry to describe living matter which exhibits mul-
smaller scales. tiple scales of organization and variable material dis-
Rapid advancements in additive manufacturing tributions, have necessitated new forms of represen-
(AM) and material science are increasingly forcing tation that go beyond the traditional boundary rep-
material design, as a technique, to become an iso- resentation.
lated self-contained practice with little feedback into In a boundary representation, influenced as it
the overall design process. Instead, we argue that were by the products and the demands of industrial
material design is integral to design development manufacturing, each object is understood to be fully
and contend that a new digital paradigm is needed described by its bounding surfaces while the inte-
to support an integrated design process. rior is treated as a solid homogeneous and isotropic
Secondly, 3d visualization and modelling tech- material. Contrastingly, in the biomedical paradigm;
niques are increasingly influenced by developments the interior structure of an object matters. In fact,
in the biomedical field, which because of its sub- the boundaries between components are fuzzy and
ject matter is concerned with distributions of matter highly fused to the point that one cannot speak easily

about the space between traditional CAD modeling The product of our research is called Monolith and is
and three dimensional voxel based image processing a hybrid modeling environment at the intersection of
with the aim of allowing a very fine level of control CAD and three dimensional image editing (see Fig-
over volumetric material distributions, especially in ure 2). It employs visualization techniques borrowed
light of new multi-material 3D printing technologies. from the biomedical field that help reveal and inter-
act with fine inner structures within the material vol-
ume. Figure 1
To achieve this we needed to address three prob- 3d printed model of
lems. One is the data structures for describing these human scull using
continuous material fields, the second is the user in- multimaterial
teraction and design workflows exposed in a soft- polyjet technology.
ware application, and third is the types of outputs
and processing that could link the digital models to
actual fabrication processes.
DATA STRUCTURES
For the data structure, we opted to use a multichan-
nel voxel model. Similar to how an image contains
RGB channels, our data structure contains a series of
channels that define different geometric and mate-
With these as our two starting points, our research rial properties. Each channel represents one type of
and software development focuses on exploring the field that describes a scalar value that changes from
design possibilities of hybrid geometric representa- point to point in space.
tions and design workflows that combine the intu- The most basic channel is the shape channel
itive geometric definitions of existing CAD with the that will determine the solid/void boundary. From
fuzziness and fluidity of voxelized models, and the this field we will extract two nested isosurfaces that
flexibility of parametrized and associative definitions. will determine the outer and inner boundaries of the
In addition, we propose an integrated design envi- solid volume.
ronment where simulation is embedded into the de- A second channel will determine the material
sign process. Finally to realize the promise of hierar- mixing ratio and will be interpreted using three
chical material design that spans scales from the mi- dimensional typographic techniques into a micro-
cro to the macro we explore the possibilities of three pattern with desired porosity and anisotropy. Essen-
dimensional typography as a way to define varying tially, this channel describes the material mixing ra-
material properties. tio at each point in space. For example, a value of 0.0
would mean that the first material would be applied
at the corresponding voxel location while a value of Figure 2
0.5 would be interpreted as a 50% mixture of the two Monolith's user
materials; creating a composite of their color, elastic, interface.
or optical properties (see Figure 3).
Finally three extra channels describe the fields
for volumetric texture coordinates (U, V, and W) that
will enable arbitrary flow of micro-patterns within the
volume.

Figure 3 spond to different weighted arithemtic operations
The shape and between the corresponding voxel values. Geomet-
material ratio rically they correspond to Boolean operations. Addi-
channels form a tion corresponds to union, difference to subtraction,
combined and multiplication to intersection of solids. How-
outcome. ever because the voxel models are fuzzy, they are
Additionally, the data structure is extensible enough characterized by gradients rather than simply solid
to make it easy to exchange information with simu- / void conditions. The blending operations result in
lation modules. For example, supplementary chan- softer blending with some filleting along the inter-
nels can be added to a voxel model which contains sections. Even very complex Boolean and morphing
the base geometric information in order to store the operations that would challenge most boundary rep-
results of a structural analysis simulation. resentation based software, are trivial when using a
Each voxel channel can be used as a drawing voxel representation. In fact, it is the sharp transitions
context. Operations such as anti-aliased drawing of and acute intersections that are difficult to replicate
lines, points, planes, transformed voxel models and within the voxelized model (see Figure 4).
meshes are supported. These operations use three
Figure 4 dimensional super-sampling in order to avoid pixila-
Examples of soft tion and aliasing effects along boundaries.
Boolean operations
between voxel INTERACTION AND MODEL BUILDING
models. Various volume based modeling methods for mate-
rial gradients have been proposed before (Jackson
et al. 1999), but our approach focuses on the fluid-
ity of the user experience trying to provide a com-
prehensivre user interface for volumetric modeling. By using fully parameterized layers we achieve two
The interaction with the model is mediated through things. First, the user can manipulate the model and
a layer compositing paradigm not dissimilar to tra- the series of operations that generate it at a more
ditional image processing applications (Blinn 1994). granular level. This is something that users have
However the layers themselves can host extensive come to expect from design applications. Second,
user interfaces with self-contained workflows. For ex- the use of parameterized layers allows the user to
ample, some layers adhere to a freehand painting interact with a low resolution version of the model
paradigm while others utilize a more controlled ge- while editing and only compute a fine resolution ver-
ometry based interface for field generation. sion for production. This enables the software to
Layers are organized in a stack that is composited achieve interactive rates during modeling and com-
to create the final model. Each layer represents either positing operations.
a modeling operation (e.g. creating a set of fuzzy gra- The user can add various voxel field generator
dients around geometric objects, or voxelize a mesh) layers that allow them to interactively define three di-
or a filtering operation (e.g. Gaussian blur, signed dis- mensional gradients. The five most notable ones cor-
tance map). respond to different creative paradigms.
The compositing of layers is mediated by a com- 1. The formula based layer, generates voxel fields
positing function that determines the type of blend- by a user provided formula (see Figure 5). The user
ing between a layer and the one above it in the types a function of the form f(x,y,z) and the appli-
stack. These compositing functions simply corre- cation evaluates this function at each x,y,z location

throughout the voxel field. In addition any parame- Figure 5
ter that appears in the formula but is not a recognized Formula based
mathematical function (e.g. cos, sin, exp...) or re- generation.
served variable (e.g. x,y,z...) is automatically mapped
to a user controlled parameter and attached to sliders
in the UI. Thus enabling a relatively easy method for
the creation of parameterized volumetric gradients. Figure 6
2. Image based modeling methods rely on user Image sweep
provided greyscale images that are used as cross- generation.
sections in loft or extrusion operations (see Figure 6).
Whereas in brep modeling one lofts curves, in voxel
based modeling one has to extrude images. This
technique can create loft operations with smooth Figure 7
topological transitions. Geometry based
3. The geometric distance field layer enables generation.
the use of conventional geometric primitives (lines,
points, circles, boxes) as base geometries for the gen-
eration of a voxelized field around them. Various
falloff parameters remap the distance of each voxel
to a geometric element so that different gradients are
generated around objects (see Figure 7).
4. The grid editing layer allows a more controlled Figure 8
and structured way of defining arbitrary gradients in Examples of weight
space. The user interacts with a grid of points by as- editing a grid of
signing different weights to different nodes. These control points and
weights are then interpolated in order to generate various
a smooth gradient between the nodes. The user interpolation
can select different interpolation functions that result functions.
in different interpolated boundary surfaces between
the nodes (see Figure 8).
5. The free paint layer allows the user to edit the
field by applying local operations interactively, like Figure 9
"sketching", sculpting or locally distorting the under- Example of
lying voxel model (see Figure 9). localized smearing
Voxel fields created by generator layers can then control in free paint
be manipulated using various filters and effects. A mode.
Gaussian blur applied to a voxel model results in
the smoothing of the level set surface including a Figure 10
reduction of topological complexity (see Figure 10) Example of a
something that cannot be achieved with conven- Gaussian blur filter
tional mesh relaxation methods. applied to voxel
model.

Other filtering layers can be used in order to apply In terms of user perception, the interface emphasizes
some convolution based filters using various kernels. the volumetric aspects of the design; thus moving
Such filters can facilitate operations like erosion or the focus away from surface modelling and the im-
"thickening" of the voxel model. Other filters are used portance of the skin that has dominated BREP based
to create distortions within the voxel fields by remap- free form modeling applications to date. The user is
ping the voxel values to different locations. Voxel presented with a toolset that is geared towards the
based distortions can easily accommodate nonlin- creation of material distributions in the form of gra-
ear transformations, ranging from a simple twist, to dients that permeate space.
topologically complex operations like an inversion
that can turn a model inside out (see Figure 11), or VISUALIZATION
a symmetric self-blending. Monolith's voxel-based approach has also required
that the traditional modes of visualization be revis-
Figure 11
ited. In conventional CAD software, the primary em-
The inversion filter
phasis is on surfaces and the rendering pipeline has
turns an object
been optimized for mesh rendering. However, in
inside out.
Monolith we wanted to be able to visualize in real
time objects that are translucent, have inner and of-
ten quite intricate structures, material gradients, and
even micro-patterns (see Figures 12 and 13). This
has become possible by using pixel shaders that im-
plement a ray marching algorithm (Hadwiger et al.
Figure 12
2008). The shape channel, the material gradient, the
Use of volumetric
UVW map and the micro-pattern are all uploaded to
shaders enables the
the GPU as 3D textures and sampled along the ray
visualization of
paths. Normals are estimated from the gradient of
inner structures.
the transparency field.
The marching ray shader is implemented in
OpenGL Shading Language (GLSL) and High Level
Shading Language (HLSL) and basically computes
the alpha weighted sum of all the points that a ray
Figure 13 leaving the viewer's eye encounters as it propagates
Volumetric shading through the voxel field. This requires a rather compu-
of a medical model. tationally expensive loop within the fragment shader
that has to execute hundreds of 3d texture lookup
operations per pixel.
To further speed up the shader we use an off-
screen low res depth texture that contains either the
bounding box of the voxel field or a low resolution
version of the isosurface (Lorensen and Cline 1987).
This off-screen texture is used as a mask so that pix-
els that do not correspond to rays that would inter-
sect the voxel field are discarded early. In addition
this depth texture determines the starting points for

the marching rays so that we can concentrate most distance will become the W coordinate. This method
of the resolution of the rays within the volume. will allow the patter to flow along a given geome-
Finally the step of the marching along a ray is try (see Figure 17). For instance it might be advan-
slightly randomized. This prevents the appearance tageous for structural reasons to align the inner scaf-
of artifacts that would make the model seem like a folding of a part to the outer surface so that some
series of transparencies parallel to the viewer. At the of edges of the pattern may align with directions of
same time randomization gives a more grainy quality principal stress in some situations.
to the rendered object.
Figure 14
A model with a 3d
RASTERIZATION halftone
In order make hierarchical material design possible,
micro-pattern.
we needed some way to describe the material pat-
terns and their mapping throughout a three dimen-
sional volume (see Figure 14). It was important that
the user should be able to control the microstructure
shape (i.e. its variable porosity) from point to point in
space and its directionality. Control over these prop-
erties enable the realization of materials with inter-
esting anisotropic optical and elastic characteristics.
There are three elements that control the micro
pattern within the volume. The first is the material
ratio channel that determines the macroscopic gra-
dient that we want to achieve. This gradient will be
reinterpreted as some short of variable porosity or
mixing ratio for the two materials that are going to
be 3d printed. The second element is a pattern that Figure 15
will get repeated throughout the volume in order to Micro-pattern map
achieve the three dimensional half toning effect. This compositing and
pattern is repeated and multiplied with the material half-toning.
ratio field. The result of this operation after binariza-
tion is a micro pattern whose porosity depends on
the underlying material ratio field. Finally in order to
determine how the micro-pattern flows through the
volume we can use a set of three channels that would
correspond to the UVW coordinates (see Figure 15). Figure 16
The user can control the flow of the UVW coordi- UVW coordinates
nates using various volumetric extensions of conven- flowing around
tional ways of generating texture coordinates (from imported mesh
a sphere, cylinder or other geometry) (Ebert et al. model.
2002). In addition Monolith enables the user to im-
port a mesh and then it propagates UV coordinates
from the mesh to the whole voxel field along the
singed distance function (see Figure 16). The signed

Figure 17 OUTPUTS
UVW coordinates In order to allow post processing or fabrication of
flow within the monolith models we need to translate them to suit-
volume of a model. able outputs that are compatible with existing BREP
software and printer drivers. The most basic output
is a mesh of the level set surface. However in order to
describe material gradients we added two more op-
tions. One is to export a cascade of closed meshes
that represent different gradient regions along the
material mixing channel. To achieve this we create
isosurfaces of the material ratio channel and then
trim them by the shape channel. This method al-
low the use of monolith model with multimaterial
printers that support gradients indirectly through the
assignment of different materials to different adja-
cent meshes. A second way to translate continuous
voxel models into something resembling a gradient
is to use a hyperfine mesh that represents the variable
Figure 18 porosity pattern of one material mixing into another.
Example of a In addition monolith contains two slicing
micro-pattern with pipelines. The first slicing pipeline is used for the
a 2D raster slice generation of high resolution bitmap stacks intended
preview. to be used with polyjet and other bitmap based 3d
printing technologies (see Figure 18). The second
Figure 19 slicing pipeline extracts various vector based slices
Stress line that are suitable for gcode generation and FDM style
visualization using printing. In both cases the material gradient is inter-
hairline 3d printing. preted as a three dimensional half toning or dither-
ing that macroscopically will resemble one material
mixing into another.
The bitmap slicer can simultaneously slice
through the voxel field as well as a set of curves and a
bounding mesh. In order to efficiently slice through
large numbers of curves we break the curves into
line segments and use an interval tree data structure
The user can define the micro-patterns either to organize these segments so that we can limit the
through simple two dimensional grayscale images number of intersections per slice. The interval tree
(i.e. bitmaps) or through volumetric images that allows us to very quickly retrieve all line segments
they have created within Monolith itself. In that way, that intersect a horizontal slice taking into account
a voxel model can become the rasterization micro- line thickness. Curves embedded in the model can
pattern that determines the material mixing and de- be voxelized at print time allowing for the creation
position of a larger model, thus enabling a form of of very fine patterns throughout the volume without
hierarchical modeling across different scales. the need to create computationally and memory in-

tensive meshes. For example, the stress line pattern plified with significant performance gains. Emerging
can be visualized and printed directly within a single simulation methods such as lattice Boltzman fluid dy-
volume of material (see Figure 19). These stress line namics would eventually enable a Multiphysics sim-
patterns also suggest the optimal reinforcement pat- ulation to run over a voxel model at interactive rates
tern if one were to use multi-material 3D printing to for the user, with real time feedback. The use of topol-
deposit soft and a fibrous materials. ogy optimization too is a natural fit for voxel models.
In fact for the minimum compliance problems the op-
MATERIALIZATION AND DESIGN POTEN- timal solution requires the existence of material stiff-
ness gradients rather than clear boundaries between
TIAL
hard and soft materia (Bendsoe and Sigmund 2004)
The definition and fabrication of material gradients
(see Figure 21).
presents us with new design opportunities but also
require the development of a different design intu- Figure 20
ition (Kaijima and Michalatos 2011). The designer Example of
is expected to think across multiple scales from the anisotropic
micro-pattern to the macroscopic object, and at the transparency.
same time move from the boundary representation
regime that presents objects as sets of surfaces into
a paradigm that favors volumetric material distribu-
tions.
For example by mixing transparent and opaque
materials we can achieve gradients of privacy or op- Figure 21
tical clarity and light diffusion / transmission. If we A reinforced beam
add the micro-pattern into the design process of with a mixture of
these semi-transparent designs we can create microsoft (translucent)
louvres within the volume of an object such that it and hard (opaque)
will seem as more or less transparent (or transmit materials produced
more or less light) depending on the viewing angle using topology
(see Figure 20). optimization.
Mixing two materials of different elastic moduli
we can create a gradient of strength with smooth
transitions between soft and hard parts. This would
avoid certain high stress concentrations along mate-
rial boundaries (see Figure 21). In order to further ex-
plore the possibilities at the intersection of voxel rep-
resentations and structural simulations we added a
static linear elastic analysis and topology optimiza-
tion module within Monolith. This taken to smaller scales enables us to design arti-
Voxel models are, in a sense, easier and more ficial grain within a material (Doubrovski et al. 2014).
straightforward to simulate as they don't require a When we mix a very soft with a very hard material we
discretization step which complicates most finite el- can create patterns that cause the anisotropy in the
ement simulations of brep based models. Because of grain to manifest as a preferential deformation of the
the constant step in the voxel field, certain operations material along certain controlled directions (see Fig-
like the building of stiffness matrices are greatly sim- ure 24).

Figure 22
A chair model with
principal stress
aligned
reinforcement
pattern (side view).
Figure 23
A chair model with
principal stress
aligned
reinforcement
pattern (top view).

CONCLUSION Figure 24
Through the use of new design workflows and visual- Example of artificial
izations, Monolith attempts to achieve a shift in how grain using two
designers approach materiality through digital me- materials of
dia. The assembly paradigm and the boundary rep- different elastic
resentation that we inherited from mechanical engi- modulus.
neering and ultimately the industrial revolution are
no longer adequate in describing the new possibil-
ities afforded by emerging fabrication technologies.
Even the analytical tools have to adapt; moving from
a statistical understanding of materiality through a Figure 25
homogenization process to the possibility of micro- Microscopic images
management of material deposition in a localization of various
approach. This will also enable us to revisit traditional micro-patterns.
materials such as wood in all their complexity and
be able to integrate the material anisotropies, due to
both natural and artificial grain in tectonics that span
a wide range of scales.
REFERENCES
Bendsoe, M.P and Sigmund, O. 2004, Topology Optimiza-
tion, Springer
Blinn, J.F 1994, 'Image Compositing-Theory', IEEE Com-
puter Graphics and Applications, 14, pp. 32-33
Doubrovski, E.L., Tsai, E.Y., Dikovsky, D, Geraedts, J.M.P,
Herr, H and Oxman, N 2014, 'Voxel-based fabrica-
tion through material property mapping: A design
method for bitmap printing', Computer-Aided De-
sign, 60, pp. 1-12
Ebert, D.S., Peachey, F.K., Perlin, D. and Worley, K. 2002,
Texturing and Modeling, Third Edition: A Procedural
Approach (The Morgan Kaufmann Series in Computer
Graphics) 3rd Edition, Kaufmann
Hadwiger, M, Ljung, P, Salama, C.R and Ropinski, T 2008
'Course Notes Advanced Illumination Techniques for
GPU-Based Volume Raycasting', SIGGRAPHASIA2008
Jackson, T.R., Liu, H, Patrikalakis, N.M., Sachs, E.M. and
Cima, M.J. 1999, 'Modeling and Designing Function-
ally Graded Material Components for Fabrication
with Local Composition Control', Materials & Design,
20, pp. 63-158
Kaijima, S and Michalatos, P 2011, 'Intuitive material dis-
tributions', in Legendre, G (eds) 2011, Mathematics
of Space, Wiley
Lorensen, W.E and Cline, H.E 1987, 'Marching cubes: A
high resolution 3d surface construction algorithm',
ACM Computer Graphics, 21, pp. 163-169

Design Exploring Complexity in Architectural Shells
Interactive form finding of reciprocal frames through a multi-agent system
David Gerber1 , Evangelos Pantazis2

1
University of Southern California, School of Architecture & Viterbi School of En-
gineering 2 University of Southern California, Viterbi School of Engineering
1
www.djgerber.com 2 www.topotheque.com
1,2
{dgerber|epantazi}@usc.edu
This paper presents an integrated workflow for interactive design of shell

structures, which couples structural and environmental analysis through a
multi-agent systems (MAS) for design. The work lies at the intersection of
architecture, engineering and computer science research, incorporating
generative design with analytical techniques. A brief review on architectural
shell structures and the structural logic of reciprocal frames is described.
Through the morphological study of reciprocal frames locally we seek to inform
the behavior of a MAS, which integrates form-finding techniques, with daylight
factor analysis (DFA) and finite element analysis (FEA) on a global
configuration. An experimental design is developed in order to explore the
solution space of large span free form shells with varying topologies and
boundary conditions, as well as identify the relationships between local design
parameters of the reciprocal frames (i.e. number of elements, profile) and the
analyses (i.e. stress distribution, solar radiation) for enabling the generation of
different global design alternatives. The research improves upon design
decision-making latency and certainty through harnessing geometric complexity
and structural form finding for early stage design. Additionally, the research
improves upon design outcomes by establishing a feedback loop between design
generation, analysis and performance.
Keywords: Generative design, computational design, multi-agent systems, shell

structures, reciprocal frames, form finding, parametric design
INTRODUCTION AND MOTIVATION have provided architects with opportunities to move

The rapid evolution of digital design in the field of away from standardization and traditional design ap-
architecture and design computing and its' coupling proaches that predicate themselves on symmetry
with new fabrication and construction technologies and repetition. Our work takes as precedent a post-

fordist perspective on the possibilities for architec- fied systems, when the form-force diagrams are vi-
ture, namely a rejection of modernisms' over reliance sualized and made accessible (Rippmann and Block
on the self similar and an acceptance of the con- 2013). Despite advances, the literature indicates
temporary condition for mass customization. How- there remains a lack of integrated and extensible
ever, the work does not take the position of differ- workflows combining multiple analyses and genera-
ence for difference sake but rather more importantly tive design early in the design process. The necessity
sees differentiation as an indicative of performance of such workflows for design through to construc-
and natural design evolution. It furthermore situ- tion of bespoke structures challenge existing tools
ates itself intentionally in a discourse of emergence and methods in part due to the complexity of the
and of the viability, tenability and even perhaps en- design constraints, parameter sets, and often con-
hanced characteristics highly complex and intricate flicting and complexly coupled objectives (Shea et al.
geometry and tectonic systems may have for archi- 2005, Scheurer 2010). Our work in part seeks to tackle
tecture. The availability of parametric and associative this pervasive problem by providing a prototypical
geometry systems has allowed for the invention and integrated workflow but also seeks to further the in-
proliferation of non-Euclidean and un-variegated ge- tegration and adopting of computer science tech-
ometries and has facilitated design variation in accor- niques of great interest and relevance to architecture,
dance with designer intent and here, most critically, that of agent based models or here more specifically
where complexity can be substantiated as higher the multi-agent system (MAS).
performing. The research furthermore takes into ac-
count an era for architecture of an increased interest BACKGROUND
in performance based design models which combine With the introduction and pervasive availability of
analytical and evaluation processes with digital pro- reinforced concrete in the 20th century, an inter-
cesses of form generation and rationalization (Gerber est in the use of thin shells and funicular struc-
2007). Yet our research sees performance as need- tures for architectural purposes becomes prevalent
ing to be balanced with the synthetic, aesthetic and (Pietraszkiewicz and Gorski 2013). This is easily ob-
communicative aspects of all great design, in other served through a survey of the great number of large
words, the informing of form through both the de- span shell structures that were realized in 20th cen-
sign exploration and solution space thinking intrin- tury with functions that vary from stadiums, to fac-
sic to associative parametric and algorithmic genera- tories and airplane hangars to name a few (Adri-
tive design with that of complexly coupled empirical aenssens et al. 2014). Famously, architects like A.
objective functions found in our engineering cousins Gaudi and H. Isler developed apparatuses such as
analytical methods (Mitchell 2005). the hanging chain and fabric model methods in or-
Research has shown that empirically validated der to simulate, test and calculate shell structures
simulations and visualized information can assist de- (Chilton and Isler 2000, Ochsendorf 2010). In sim-
signers in making better design decisions. Exam- ilar analogue fashion F. Otto, studied the behavior
ples include when environmental analyses simula- of grid shells and membrane structures, while struc-
tions are used during design development in order to tural engineers like P. L. Nervi and F. Candela in-
enhance environmentally-conscious and cost effec- vestigated statics and construction techniques and
tive design solutions (Roudsari et al. 2014, Eastman by manipulating them achieved the construction of
1994). Another recent precedent includes the inter- unique thins shell structures such as Agnelli Exhibi-
active design exploration of shell structures through tion Hall in Turin (1948) and the Chapel Lomas de
form finding which demonstrates that complex struc- Cuernavaca (1958) (Tomlow et al. 1989, Burkhardt
tures can be more efficient than reduced and simpli- et al. 1976, Garlock et al. 2008). These researching

practitioners all are credited with laying the founda- der to manage contemporary design demands, archi-
tion of what is called architectural form finding to- tects and designers need to shift away from reduc-
day. More recently, researchers have been develop- tionist models that rely on standardization and seg-
ing computational methods for modeling and solv- mentation of functional requirements and use ana-
ing or "form finding" structures to find equilibrium. lytical simulation tools that provide designers with
Killian among other researchers developed interac- useful performance feedback (Oxman 2008, Malkawi
tive, real-time hanging chain modelling tools, using 2005). Therefore, there is a need to develop interac-
particle spring system solvers(Axel Kilian 2005). Piker, tive and integrative design tools that intuitively allow
developed "Kangaroo" a tool for real time exploration designer to understand the relation between geom-
of funicular networks using dynamic relaxation while etry and performance, promote geometric, material
Block has developed a methodology and tool for and functional differentiation and provide a seamless
the exploration of funicular shells under compression data flow from design to construction for among all
based on Thrust Network Analysis(Piker 2013, Block participating counterparts.
and Ochsendorf 2007).
Our literature suggests that the future of digi- RESEARCH METHODOLOGY
tal design lies in considering the computer as a col- Our interest lies at the intersection of architecture,
laborative partner in the design process that is ca- engineering and that of empirical models to enhance
pable of generating design alternatives and evalu- performance criteria in broad terms. In designing sys-
ating different solutions in response to robust and tems and methods that enable architects to develop
rigorous analytical models of design conditions and free form shell structures that are: computationally
performance (Gerber et al. 2012, Pask 1969, Mitchell generated, intricate and highly varied geometries,
1990). Moreover, the literature indicates that in or-
Figure 1
Experimental
workflow diagram
illustrating the
context,
generation,
analysis, and
materialization and
feedback steps of
top down,
generative,
feedback and
optimization, and
then to material
prototyping.

Figure 2
Diagram describing
the design steps of
the MAS for design
methodology.
and that span continuous open architectural spaces. prototype combines form-finding techniques using
The research contributes to the discourse where the a spring based mesh relaxation solver (Rhinoceros
complex is embraced as opposed to reduced. Here Grasshopper/Kangaroo), with environmental solar ra-
our hypothesis states that by combining interactive diation analysis (Rhinoceros Grasshopper/Ladybug,
form finding techniques with environmental analy- Radiance) in order to drive a multi agent system (Pro-
sis, via a MAS for design approach, designers can cessing, IGEO). Both the global geometry (shell) and
achieve the generation of alternative shell structures, the local (reciprocal frame) are conditioned based on
which perform better across multiple performance a set of user defined environmental parameters such
criteria when compared to purely deterministic and as: geo-location, footprint area, support conditions
or manual design exploration methods. The scope of and material . The research then studies the proper-
the work presented here includes, the defining and ties of the reciprocal frame design alternatives locally,
optimization of structural components, the environ- in terms of different number and shape of elements,
mental efficiencies gained, and, and a discussion of and explores which the relationship between each
the design decision making affordances for the de- design parameter and the structural analysis. Glob-
signer. Another critical objective is to evaluate and ally different support conditions and loading condi-
argue further for how the non-standard can be found tions are studied. The system uses the analyses (i.e.
to outperform the standard structural system. stress distribution, solar radiation maps) and the rela-
Our research methodology includes, the formu- tionships between different design parameters as in-
lation of an integrated workflow for the design of ef- puts for the behavior of a multi agent system (MAS),
ficient shell structures, which are based on the recip- which generates alternative morphologies of recip-
rocal frame structural system (see Figures 1-2). This is rocal frames (See Figure 4) . The morphology of the
developed through a custom MAS which is validated reciprocal frames is informed by the combination of
through a series of incremental experiments and the structural and environmental analyses via a MAS
comparisons. These experiments include the gener- driven optimization. The goal is the generation of
ation of a set of shell models, which are expressed more efficient structures, which span large areas with
and prototyped as reciprocal frames, which are not the use of short structural members (material econ-
solely driven by structural efficiencies but coupled omy) while providing improved environmental con-
with environmental considerations. The software ditions beneath them (energy economy).

Figure 3
Diagram illustrating
the input
parameters and 2
different design
behaviors and
resulting
environmental
analysis.
DESIGN OF THE EXPERIMENT particle engine is informed by user specified environ-

The project explores a design methodology for shell mental parameters (longitude, latitude, orientation,
structures, which combines form finding informed sun position). The generated shells are populated
with environmental parameters and the application with reciprocal frame units which are analyzed struc-
of a modular structural system that of reciprocal turally both locally and globally. Based on the struc-
frames. The principle of reciprocity in structural de- tural analysis on a local level key design parameters of
sign and construction i.e. the use of load bearing el- a reciprocal unit such as: a) the length, and b) thick-
ements to compose a spatial configuration wherein ness of elements are adjusted in order improve the
they are mutually supported by one another has structural performance globally. In this experiment
been in use since antiquity. The design of the ex- we focus only on two aforementioned design param-
periment includes the analytical study of the recip- eters, in Figure 4 we illustrate all key parameters of
rocal unit and the application of reciprocal frames on the reciprocal frames(see Figure 4). The research then
form found shells. The form finding process is occur- implements the above methodology by developing
ring though a MAS based approach where a physics the computational workflow (see Figure 2) that in-

cludes both a generative and an analytical compo- imental design process includes the definition of a
nent informing each other. The generative part of covered area, support conditions and amount of el-
the workflow is developed using Processing and the ements to be used, as well as structural parameter
IGeo libraries while the analytical part is developed such as elasticity, connectivity between mesh points
in Rhinoceros 3D software using Grasshopper and (particles as frame elements) and friction. A shell ge-
specifically a) Kangaroo b) Karamba, and c) Ladybug ometry is generated based on the combination of
plugins (see Figure 1). The workflow is tested by de- force distribution (self weight) and solar path posi-
veloping an experimental design case study, that of tions (shading). Once the geometry reaches equi-
the long span hangar in its global form. The exper- librium, reciprocal frames are applied on the surface
Figure 4
Design parameters
for reciprocal frame
and stress
distribution of
different units
based on the
number of
elements.

at every point of the geometry where the architect RESULTS AND ANALYSIS
and design team controls the valency (number of el- The results of our work so far are on both a local and
ements in one reciprocal unit), as well as the param- global level. On a local level we test different sec-
eterization in the generation of varying element pro- tion profiles, which range from standard to not stan-
files (i.e. sectional profiles). The generated geome- dard and observe that the structural performance of
tries are analyzed both locally and globally where each reciprocal element is largely affected by cross
the design has a graphical overview of the stresses sectional profile. Six different types of profiles are
induced in the profiles assuming a certain material tested including: a circular pipe, a rectangular el-
and load condition (see Figures 3 and 4). For this ex- ement, and 4 types of planar elements with differ-
perimental run we kept this constant as Douglas Fir ent, a) proportions of width over height, and b) with
South Wood and 1kN load distributed over the en- one transformation (torsion along main axis and off-
tire surface area of each beam element. The geomet- set along main axis. We plot the stresses on the re-
ric properties of the results are visually assessed by ciprocal elements with different geometries in rela-
the designer in correlation with their structural per- tion to varying length and thickness (see Figure 5).
formance (see Figure 5). We observe that only one type of element (twisted
planar) fails to meet the requirements for allowable
Case Study: An informed reciprocal frame stress based on the American Building Code while
long span shell structure the rest of the profile types show an almost linear re-
In order to validate the systems and methodology, lationship of their length and thickness to the stress.
the MAS is applied on the conceptual structural de- We note that as the length increases the stress also
sign of a large span hangar structure. We use our increase while as the profile increases the stresses
approach for the design of reciprocal frame struc- on the element decrease. In terms of the unit we
tures with different topologies that span 50m and we observe that as the number of elements increases
compare it with a recently built structure of the same (valency) the stress decreases (see Figure 5). For a
size. We use the comparison between the different unit (reciprocal frame) of n=3 and a pipe profile the
cases as a means for benchmarking efficiencies of our stresses can be as high as 1620 kN while for n=6 it
methodologies and for drawing conclusion for the drops as low as 703 kN (66%) At this stage, for simplic-
further refinement of our system. ity and clarity of the experiment we test only one pro-
Figure 5
Graphs showing
maximum element
stress in relation to
length, thickness
and geometry as
well as the
allowable stress.

Figure 6
Structural analysis
of global
geometries
generated by
varying agent
behaviors. Figure 7
(below): Solar
Radiation analysis
of global
geometries which
have been
generated with
form finding and
with different agent
behaviors.

file type and reciprocal frame unit (n=3) for applying Multi-Agent System for architectural, structural and
it on the global geometry. In terms of global geom- environmental optimization from design to construc-
etry we generate 4 different topological cases based tion.
on the support conditions and whether or not the ge-
ometry has an opening. We test one of the topolo- CONCLUSIONS
gies for both a span of 50 and 100 meters. We ana- This paper presents the application of a MAS for de-
lyze both the form found and environmentally influ- sign approach that supports the integration of struc-
enced geometries and measure the following: prin- tural performance in order to inform the design deci-
cipal forces (compression, stress, moments, displace- sion making of a generative design process, leading
ment and cross section utilization (see Figure 6). The to improved design solutions and in this case long
utilization of the cross section, is the ratio of applied span reciprocal frame structures in an applied hangar
stress over yield strength of the element and using a configuration. The aim of this work is two-fold, a) to
safety factor of 200% we size the elements in order explore how designers could implement non deter-
to achieve a target utilization which is between -50% ministic processes where the precise definition of lo-
to +50%. Additionally, the maximum displacement cal rules can be combined with analytical tools in or-
is measured which is within limits (0.5 to 5cm). The der to lead to globally optimized geometries, and b)
structural analysis shows that geometries with a hole to develop an integrated design methodology based
have a significant increase in moments (130-200% in- on MAS that provides designers with not only ge-
crease) while we observe that the amount of support ometric feedback but structural performance feed-
conditions does not affect significantly the result- back and thus assist them in generating and selecting
ing principal stresses rather their position and shape among well performing design solutions. The pre-
(see Figure 6). The environmentally informed shells sented experimental case illustrates how a MAS for
are subject to higher principal forces as their form is design in architecture approach can be developed
modified from the pure form finding however they and tailored to specific design problems in order to
are performing better in environmental analysis (see assist architects and designers in generating higher
Figure 7). Although the designed workflow is con- performing design solutions in terms of structural
trolled manually by the user in terms of generation and environmental performance.
and transfer of geometrical data for finite element Here the experimental results are presented in
analysis, the intent is to lay a foundation that can the form of three scales and investigations: 1) the
be automated for generation and analysis to form reciprocal frame morphology (valency); 2) the cross
a singular structural agent, which has as an input sectional and element lengths; and 3) the global con-
the analytical results and as an output modifications figuration of a shell structure, the hangar. The ge-
on the defined design parameters. With proper au- ometric configurations of the design solutions may
tomation, ample results can be generated for differ- vary but the coupling of the generative with the an-
ent profiles, different notch styles, varying material alytical and optimizing processes ensures the satis-
parameters, varying cross-sectional dimensions and faction of prescribed structural goals. Our objective
varying frame morphologies. The next steps include has remained to demonstrate the value of the emer-
the development of an autonomous structural agent, gent, non-standard, and geometrically intricate as a
which will be able to analyze, visualize and evaluate viable post fordist solution for form finding and per-
structural analysis data from the generated geome- formative driven design. The system was able to
tries and communicate information to the generative generate shell as reciprocal frame configuration(s),
agent in order to continually optimize the next gen- which provided longer spans based on short self-
erative iterations. The final objective is to develop a similar elements. We have demonstrated that archi-
tects and designers can benefit by implementing an

integrated bottom up design approach for the de- Chilton, J and Isler, H 2000, The Engineer, Thomas Telford
sign of building components that lead to optimal Publishing, London
global configurations. By combining analytical meth- Eastman, C 1994, 'A data model for design knowledge',
Automation in Construction, 3, pp. 135-147
ods with user data, properly formulating and pass-
Garlock, M, Billington, DP and Burger, N (eds) 2008, Félix
ing such data automatically to a generative process, Candela: engineer, builder, structural artist, Princeton
we provide designers with a larger pool of complex University Art Museum Monographs
yet well performing design solutions that could not Gerber, DJ 2007, Parametric practices: Models for de-
be modelled manually. The experimental case study sign exploration in architecture, Ph.D. Thesis, Harvard
and results showed improvements both in terms of Graduate School of Design
Gerber, DJ and Lin, S 2014, 'Designing in complexity:
energy efficiency by generating solutions that in-
Simulation, integration, and multidisciplinary de-
creased the amount of available daylight, as well as in sign optimization for architecture', Simulation, 80(8),
terms of design process as we provide the designer pp. 936-959
a methodology to augment an existing design ap- Killian, A and Ochsendorf, J 2005, 'Particle-Spring Sys-
proach, that of pure form finding while integrating tems for structural from finding', Journal of the Inter-
environmental parameters as well. By analyzing and national Association for Shell and Spatial Structures,
46, pp. 77-84
cross-comparing the results, the continued objective
Malkawi, A and Kolarevic, B (eds) 2005, Performative
is to deduct heuristics and weighting factors to pro- architecture : beyond instrumentality, Spoon Press,
vide intuition and to further model behaviors for our New York
agents in order for the MAS system to be able to au- Mitchell, J 1990, The logic of architecture: Design, compu-
tomatically generate a series of design. tation, and cognition, MIT press, Cambridge, MA
Mitchell, J 2005 'Constructing complexity', Computer
ACKNOWLEDGEMENTS Aided Architectural Design Futures 2005
This material is based upon work supported by Ochsendorf, J 2010, Guastavino vaulting:The art of struc-
tural tile,, Princeton Architectural Press
the National Science Foundation under Grant Oxman, R 2008, 'Performance-based design: Current
No.1231001. Any opinions, findings, and conclu- practices and research issues', International Journal
sions or recommendations expressed in this material of Architectural Computing, 6, pp. 1-17
are those of the author(s) and do not necessarily re- Pask, G 1969, 'The arhitectural relevance of cybernetics.',
flect the views of the National Science Foundation. Architectural Design, 39, pp. 494-496
We would like to give special thanks to all the par- Pietraszkiewicz, W and Gorski, J (eds) 2013, Shell Struc-
tures: Theory and Applications,, CRC Press
ticipants and researchers that contributed to this Rippmann, M and Block, P 2013 'Funicular Shell Design
project; specifically Punit Das and Rhe-seok Kim for Exploration', ACADIA 13: Adaptive Architecture, On-
their contributions in preparing and running the ex- tario,Canada, pp. 1-19
periments and in the illustrations of the results. Roudsari, M, Pak, M and Smith, A 2013 'Ladybug:
A Parametric Environmental Plugin for Grasshop-
per to Help Designers Create an Environmentally-
REFERENCES Conscious Design', 13th Conference of the Interna-
Adriaenssens, S, Block, P, Veenedaal, D and Williams, C tional Building Performance Simulation Association
2014, Shell Structures for Architecture: Form Finding (BS2013), pp. 3128-3135
and Optimization, Taylor & Francis - Routledge Scheurer, F 2010, 'Materialising Complexity', Architectural
Block, P and Ochsendorf, J 2007 'Thrust network analy- Design, 80, pp. 86-93
sis: A new methodology for three-dimensional equi- Shea, K, Aish, R and Gourtovaia, M 2005, 'Towards inte-
librium', International Association for Shell and Spatial grated performance-driven generative design tools',
Structures, pp. 155-167 Automation in Construction, 14, pp. 253-264
Burkhardt, B, Otto, F and Arup, O (eds) 1976, Multihalle Tomlow, J, Graefe, R, Otto, F and Szeemann, H 1989,
Mannheim, Institute for Lightweight Structures (IL), The model, Institute for Lightweight Structures (IL),
Stuttgart Stuttgart

DrAFT: an Algorithmic Framework for Facade Design
Inês Caetano1 , António Leitão2
1,2
INESC-ID/Instituto Superior Técnico
1
ines.caetano@tecnico.ulisboa.pt
2
antonio.menezes.leitao@ist.utl.pt
Architecture has always followed the times and their innovations and, currently,
an architecture based on digital technologies has been emerging and has
increasingly explored architectural facades. In this paper we use DrAFT, a
computational framework for the generation and exploration of facade designs,
to explore a set of different examples of building skins. DrAFT includes a
classification of facades that helps in the identification of algorithms that best
suits each design intent. After combining the algorithms provided by this
framework, the designer can more easily explore the solution space of the
intended design.
Keywords: Generative design, facade design, DrAFT framework, Rosetta
INTRODUCTION specification of facade designs. In practical terms,

Nowadays, the architectural facade is characterized DrAFT - Draft Algorithmic Facade Tool - promotes
by complex shapes and patterns, mainly, due to the the exploration of facade designs and simplifies the
use of new design tools (Pell 2010) which promote adaptation of the generated models to the ever-
further design exploration. The development of Gen- changing design process conditions. Here, we
erative Design tools, particularly, the use of algo- present a collection of examples developed using
rithmic approaches, have had an important role in this framework, thereby demonstrating its usability
the generation of these contemporary skins because and flexibility in facade design, and also showing
they simplified the design of complex and intricate other possible applications.
architectural surfaces, which would not be viable to
produce manually. In addition, they also increase ALGORITHMIC APPROACHES TO DESIGN
the design efficiency and their evolution has been Generative design (GD) is a computer-based ap-
changing, not only the design process, but also the proach to design that creates shapes through algo-
architectural thinking (Kolarevic 2003). rithms (Terzidis 2003). Algorithmic design is a process
Unfortunately, algorithmic approaches do not that explores complex forms from simple and itera-
make facade design trivial. On the contrary, they tive methods/rules while preserving specified quali-
require the rigorous specification of all algorithmic ties (Meredith 2008). For this, architects produce an
steps, a task that requires specialized knowledge and intermediate algorithmic-based description of a de-
that, in many cases, can be quite complex. sign rather than its shape (Leitão 2013).
In this paper, we propose a computational Parametric Design is a specific GD approach that
framework designed to simplify the algorithmic

generates different instances of a design, where each designs to the ever-changing design process condi-
instance represents a particular set of values for the tions and this process can be repeated as many times
design parameters (Barrios 2005), allowing the de- as needed, promoting continuous improvement in
signer to freely explore a large solution space of the the design exploration.
design briefing/program. Therefore, this allows ar-
chitects to continuously evaluate several solutions, Framework Structure
which would be difficult to do with traditional design Our framework takes into account several stages that
methods. typically occur in facade design, thereby dealing with
In spite of their advantages, algorithmic-based the different characteristics of the:
design methods require a disciplined approach
which, in many cases, is difficult to follow. It is impor-
1. Facade's surface, including its shape;
tant, then, to develop strategies that help designers
2. Design units, that together create the whole
implement these methods. In this paper, we address
facade pattern;
this problem and we contribute to the state-of-the-
3. Distribution of the units;
art by proposing a strategy for the development of
4. Articulation between the previous parts, i.e.
algorithmic-based solutions for the generation of fa-
surface and units.
cades.
DRAFT FRAMEWORK For each one, there is at least one categorical di-
DrAFT is a computational framework created to help mension in charge of producing the matching facade
designers in the algorithmic description of facade de- characteristic, which corresponds to a set of related
signs, during the design exploration stage. It is based computational functions.
on a classification of facades composed by different
categorical dimensions that we considered compu- Framework Goal
tationally relevant (Caetano et al. 2015). In practical Note that the goal of our framework is not to provide
terms, the designer combines the main characteris- functions and algorithms to cover an entire range
tics of the idealized design with the categorical di- of facade designs, and neither to limit the facades
mensions which, in turn, guide him in the selection that can be produced. It is rather to reduce the pro-
of the most appropriate algorithms. It is noteworthy gramming effort of the architects at the early stages
that this guiding process is not intended to replace of design, while speeding up the development of
the role of the designer, but to significantly reduce facades using an algorithmic approach. DrAFT al-
his programming effort and, therefore, improve his lows the reuse of algorithms that are already de-
design workflow. veloped and that are typically needed in the explo-
Thereafter, the selected algorithms are com- ration of new designs of facades. This means that,
bined using functional operators, also known as not only can architects generate new designs just
higher-order functions (HOF), i.e. functions that re- by using the selected algorithms, they can also com-
ceive other functions as arguments and/or compute bine these algorithms with some additional scripting
other functions as results (Leitão 2014). The combi- when needed. This often happens when the ideal-
nation of the algorithms produces the correspond- ized design is highly detailed and personalized, re-
ing facade design model which can be quickly modi- quiring a more specific algorithmic description. In
fied as many times as needed, allowing the designer these cases, the algorithms developed can then be
to more easily explore the solution space of his de- incorporated in the framework, thereby further im-
sign. Therefore, this allows designers to adapt their proving the matching process of subsequent facade
designs.

CASE STUDIES COMPILATION Library of Birmingham
We will start with the Library of Birmingham by
Figure 1 Mecanoo, visible in Figure 2. This building has
The selected case a straight facade composed by several overlapped
studies. First row: rings of two different sizes and colours, black and
Library of gold, which results in a unique pattern. First, we anal-
Birmingham, in ysed this facade design and, then, we obtained the
Birmingham, and most appropriate algorithms through DrAFT classifi-
the Yardmasters cation. In practical terms, we selected a set of algo-
Building, in rithms that:
Melbourne; Second
Row: Sheung Wan 1. Produced the flower shaped units (Figure 2-A);
Hotel, in Hong 2. Distributed the units in a regular-grid (Figure
Kong, and Hello 2-B);
House, in 3. Created two layers of units with different sizes
Melbourne: Last (Figure 2-C);
row: Formstelle 4. Overlapped both layers, thus creating the fi-
Building, in Töging nal pattern (Figure 2-D);
am Inn, and
Precinct Energy Lastly, we combined these algorithms using different
Project, in functional operators and HOFs. Figure 2 shows an in-
Melbourne. stance of the obtained model.
Figure 2
Library of Similarly to Shape Grammars, the framework here
Birmingham. The presented can be considered original or analytical
set of algorithms (Pupo et al. 2007). In the original case, the goal is
used to produce the to generate a completely new facade, while in the
model: A. Units analytical case we look for an algorithmic interpre-
shape; B. Units tation of an already existing facade. In this section,
Distribution; C. Two we analyse and generate a set of existing facades to
layers of different demonstrate the analytical capabilities of the frame-
sizes and colours; D. work. Initially, the selected facades were analysed
Layered facade and classified using the DrAFT classification. Figure
articulation. 1 synthetises the selected facades.
The next stage was combining the functions pro-
vided by the classifications. As a result, this com-
position of functions generates the matching facade
model of each project, which we will develop bel- Sheung Wan Hotel
low. In some cases, it was possible to reproduce Sheung Wan Hotel (Figure 3) by Thomas Heatherwick
the facade models simply using the set of algorithms is our next example. As the previous example, we
available. However, more specific designs required also started with a design analysis of this facade to
the development of additional algorithms to comple- then classify it.
ment the provided ones. This facade is straight and it is composed by sev-

Figure 3
Sheung Wan Hotel
model. A. Unit
shape; B. Unit
distribution in a
regular-grid; C. Unit
subdivision into
smaller rectangular
and squared units;
D. Random
application of
materials, metal or
glass; E. Random
depth size.
eral rectangular units (Figure 3-A). These units are an alternated grid. Finally, to create the HELLO ef-
then mapped into a regular-grid (Figure 3-B) and they fect of the facade we used an image with this word
are further subdivided into smaller units with a shape to control the placing of the perpendicular bricks. If
that randomly varies between a squared and rectan- the bricks were coincident with the area of the word
gular geometry (Figure 3-C). Lastly, the units depth HELLO, they were aligned with the facade axis. Oth-
and assigned material vary randomly from a certain erwise, they were placed so as to protrude.
height value of the building's facade (Figure 3-D and
E). Figure 3 synthetizes the combination process of Figure 4
the algorithms and shows an instance of the gener- Hello House model
ated model. produced using
DrAFT framework.
Hello House A. Pattern unit; B.
The following example is the Hello House facade by Unit distribution; C.
OFF! Architecture, visible in Figure 4, which is com- Picture to control
posed by several stacked white bricks placed in two the bricks
different positions: along the facade's surface or per- positioning; D. Final
pendicular to it. pictorial effect.
Firstly, we considered the set of two bricks as the
pattern unit, i.e. an horizontal and a perpendicular
brick (Figure 4-A). Secondly, we distributed them into

Precinct Energy Project Also note that the size of the perforations varies
continuously along the facade, thus reaching its max-
Figure 5 imum at the center and its minimum at the ends (Fig-
Precinct Energy ure 6-C). To produce this effect, the size transforma-
Project model. A. tion of these prisms was controlled by an attractor,
Definition of the which is a point or a set of points that act like virtual
unit; B. Unit magnets. In this case, we used a set of points creat-
distribution in an ing a horizontal line in the middle of the facade and,
alternated-grid. C. then, the size of each prism was calculated according
Pictorial effect with to the distance between its location and the nearest
an image; D. Final attractor point.
effect, which was Finally, the hexagonal prisms were subtracted
then subtracted from the facade surfaces in order to produce the per-
from the facades’ forated surface effect (Figure 6-D). Figure 6 shows an
surface. instance of the obtained model.
Figure 6
As in the previous example, Precinct Energy Project
Formstelle building.
(Figure 5) by PHTR Architects has a pictorial facade
A. Pattern unit; B.
that produces a design similar to an inkblot. How-
Unit distribution in
ever, in this case the inkblot effect is created by the
an alternated-grid;
existence or absence of perforations.
C. Units size
In practical terms, this example has a straight and
transformation; D.
perforated facade, wherein the perforations have a
Facade with a
circular shape. We used cylinders to produce these
perforated
holes, which were placed according to an image rep-
articulation.
resenting the inkblot effect that we wanted to pro-
duce. In the end, the cylinders were subtracted from
the facade surface, thus creating the perforated sur-
face visible in Figure 5.
Note that the process that controlled the place-
ment or not of a cylinder at each position was similar
to the one described in the previous example. Figure
5 synthetizes the whole generation process of this ex- Yardmasters Building
ample, and also the set of algorithms that were used. Our last example, the Yardmasters Building by
McBride Charles Ryan, visible in Figure 7, was also
Formstelle / Campus Netzwerk generated using the algorithms provided by its clas-
The facade of the Formstelle building, visible in Fig- sification. To generate this example we had to pro-
ure 6, was designed by Format Elf Architekten and is duce a facade with a regular shape, a layer with the
characterized by a perforated surface with hexagonal Islamic pattern and also some window openings with
holes. These were produced using hexagonal prisms, an irregular shape. In practical terms, the pattern was
which were then distributed along the facade in an produced by the repetition of a unit (Figure 7-A) dis-
alternated-grid, thereby creating a pattern similar to tributed in an alternated-grid (Figure 7-B). The ob-
a honeycomb (Figure 6-A and B). tained pattern was then used to shape the windows

Figure 7
The Yardmasters
Building model. A.
Units geometry; B.
Units distribution in
an alternated-grid;
C. Creation of the
windows; D.
Applied
articulation, with
the Islamic pattern
on the facade’s
surface; Below: the
Yardmasters’ final
model.
openings (Figure 7- C/D) and, finally, it was applied eration of a large variety of facades by reducing both
on the facade's surface. Nevertheless, we had to de- the programming effort and the time spent.
velop some additional scripting in order to produce
the Islamic units, as they have a more specific shape. OTHER APPLICATIONS
Figure 7 summarizes the generation process of In this section, we present other possible applications
the Yardmasters building, including the definition of of the DrAFT framework. We start by developing an
the surfaces and the pattern units. original facade.
In this example, although we had to develop the As an example, we will consider that we want a
algorithms to describe the Islamic pattern geometry, straight facade with pyramidal elements. To this end,
all the other design parts were produced using the we select the algorithms that respectively generate a
predefined functions. As a result, we can conclude (1) straight surface and (2) pyramidal elements. Let
that the DrAFT framework helps architects in the genus also assume that we want the height of the pyra-

mids to vary according to the distance to a curve. Our changing some of the input values, thereby produc-
framework also provides a function to produce this ing different instances of the same design, or by alter-
type of size variation. Therefore, the algorithms se- ing some of the functions used, thus changing some
lected so far allow us to define the function that cre- design characteristics. Therefore, to modify the type
ates the units. of distribution and geometry of both units and fa-
We set that the units distribution is done in an cade (see Figure 9), we do not need to change the
alternated-grid. In practical terms, the function to rest of the structure, i.e. the functions in charge of
distribute them is a higher-order function, which reproducing the other facade parts.
ceives other functions as arguments: To complete the definition of the facade design,
we can also optimize the generated models. So,
1. The function that creates each unit, as it
imagining that we want to maximize the light pass-
knows how the distribution is done but not
ing through the previous example (Figure 9), we de-
the elements to distribute.
fine as parameters:
2. The function that describes the facade surface
as it requires the set of points on which the 1. the smoothness of the attractor-curve (in a
distribution will be done. range from 0.5 to 6.0);
2. the minimum pyramids openings (between
Figure 8 0.1 and 1.0);
An instance of the 3. the height of the pyramids (with a maximum
pattern produced of 0.5m).
using the set of
algorithms above. A simple but effective optimization algorithm can
then be implemented just by repeatedly sampling
the parameter space, generating the corresponding
facade, and computing the amount of light that it
lets through, saving the values of the parameters that
maximize that amount of light.
For better control of the optimization process we
can fix some of the parameters, e.g., the smoothness,
Figure 9
and optimize for the remaining parameter space.
An instance of the
This allows us to obtain a model that combines a de-
pattern produce
sign that pleases us with values for the remaining pa-
using the
rameters that are close to the optimal (see Figure 10).
algorithms above,
which now include
units with a PORTABILITY
different shape The current implementation of the framework was
(pyramidal) and a done using the Rosetta IDE (Lopes and Leitão 2011).
distribution in an This has the significant advantage of making the
regular-grid. framework portable across the different CAD and BIM
tools supported by Rosetta, allowing us to produce
Lastly, we will define that the elements are applied on identical models in Rhino, AutoCAD, SketchUp, Revit,
the facade's surface and the colour used is gray. The and ArchiCAD. This means that the DrAFT framework
result is visible in Figure 8. is not restricted to a single CAD tool, as it happens
Now, we can simply vary this facade design by with other similar frameworks, thus liberating the de-

Figure 10
On the left: a
graphic with the
areas maximized for
each smoothness
value; On the right:
the selected final
model.
signer from the limitations of any specific software. CAD system that supports .obj, .stl, .collada, and
Moreover, it allows the designer to easily change the .dxf file formats, already attempt to solve the prob-
CAD tool that he wants to use (Figure 11). lems here described. All of them are capable of cre-
Additionally, Rosetta also promotes portability ating grids of points on a surface, mapping elements
across the supported programming languages, al- in different ways, applying attractors to control ele-
lowing the exploration of the framework in different ments size, etc.
programming languages such as Autolisp, Phyton, However, when these tools are manually used
Processing and Javascript. As a result, in order to use in an iterative user-driven process, they can be tire-
our framework, designers can choose the program- some and error-prone. In addition, when they are
ming language that they are more familiarized with, used in an Application Programming Interface (API)
without forcing them to learn a new language. or as plug-ins to a domain-specific programming lan-
guage, such as Grasshopper, a certain level of au-
RELATED WORK tomation is obtained, however, the designer is al-
Some authors have already developed some work in- ways bound to the specific functionalities provided
spired by the wide variety of contemporary facades. by the tool, thus limiting its agency in exploring dif-
Pell (2010), Moussavi (2006) and Velasco et al. (2015) ferent combinations of operations and extending the
tried to organize this variety of designs and each one capabilities of the tool's pre-defined operations. Be-
created a classification of facades based on different sides that, these tools are more used for generic
concepts. Nevertheless, none of the previous classifi- panelization, subdivision, and population of surfaces
cations helps architects with the algorithmic descrip- thus, although they have been used to generate com-
tion of new facade designs. plex facade patterns, they are not fully architectural-
Su and Chien (2016) recognized the existence of oriented which means that they do not directly ad-
some algorithmic patterns in facade designs. These dress relevant concepts in facade design such as ma-
similar algorithmic structures can be reused later to teriality or the tectonic relation between the facade
generate further designs, one of the basic concepts elements.
of our framework. Dynamo for Revit and Grasshopper for Rhino
On the other hand, tools like the Paneling Tools also allow users to implement the functionalities pro-
plug-in for Rhino and Grasshopper, the Lunch Box posed in this paper. However, the freedom allowed
add-on to Grasshopper, and ParaCloud Gem, a stand- by these tools becomes difficult to manage in com-
alone toolkit that adds generative capabilities to any plex facades (Leitão et al. 2012). In these cases,
a more structured and systematic approach like the

Figure 11
Re-execution of the
same model in
different backends
(AutoCAD, Rhino5,
SketchUP and
Revit).
one we propose is more manageable. geometries. In this paper we showed how the DrAFT
In summary, with these tools the architect is lim- framework can help designers generate different fa-
ited by their non-domain specificity or, in order to ex- cade designs. The current implementation was done
tend their capabilities, he needs to build from scratch using the Rosetta IDE (Lopes and Leitão 2011), allow-
the necessary functionalities or use a mix of different ing the exploration of facade designs in different pro-
tools that most of the time are not compatible. Our gramming languages and the generation of the cor-
work extends the state-of-the-art by systematizing responding models in different CAD and BIM applica-
and structuring, in an architectural-oriented frame- tions.
work, the parametric generation of a wide range of The framework uses a classification of facades
facade typologies, and by operationalizing it resort- that guides the selection of the appropriate algo-
ing to a simple algorithmic approach that uses and rithms for each type of facade design. The algo-
combines different functions that directly implement rithms might then be used directly, or might be com-
facade design concepts. bined using functional operators, promoting a sys-
tematic exploration of designs which ultimately aims
CONCLUSION to a higher productivity by improving the time spent
The exploration of architectural facades is not new. in scripting tasks, and adding flexibility to the de-
However, by resorting to recent digital technologies, signers' workflow. Due to the simplicity of the func-
architects can once again focus on facade design, tional composition, this framework accommodates
promoting the exploration of complex patterns and the ever-changing nature of a design process by fa-

cilitating the test of several designs, or instantiations Pell, B 2010, The Articulate Surface - Ornament and tech-
of the same idea, in any design stage. In this paper nology in Contemporary Architecture, Birkhauser
we also demonstrated the framework's flexibility by Pupo, R, Pinheiro, E, Mendes, G, Kowaltowski, D and
Celani, G 2007 'A Design Teaching Method Using
exploring a set of existing facades.
Shape Grammars', Proceedings of the 7th Interna-
In the near future, we plan to expand DrAFT, cov- tional Conference Graphics Engineering for Arts and
ering a wider range of facades. To make this frame- Design, Curitiba, Brasil
work more usable, we are particularly interested in Su, H and Chien, S 2016 'Revealing Patterns: using
conducting a field study of its application, to identify Parametric Design Patterns in Building Facade De-
weaknesses of the proposed processes and opportu- sign workflow', Proceedings of the 21st CAADRIA, Mel-
bourne, Australia
nities for extensions.
Terzidis, K 2003, Expressive Form: A Conceptual APproach
to Computational Design, Spon Press, New York
ACKNOWLEDGEMENTS Velasco, R, Brakke, A and Chavarro, D 2015 'Dynamic
This work was supported by national funds through Façades and Computation: Towards an Inclusive
Categorization of High Performance Kinetic Façade
Fundação para a Ciência e a Tecnologia (FCT) with ref-
Systems', CAADFutures 2015, CCIS 527, Springer, São
erence UID/CEC/50021/2013. Paulo, Brazil
REFERENCES
Barrios, C 2005 'Transformations on Parametric Design
Models: A Case Study on the Sagrada Familia
Columns', Proceedings of CAAD Futures 2005, Vienna
University of Technology, Austria
Caetano, I, Santos, L and Leitão, A 2015 'From Idea to
Shape, from Algorithm to Design: A Framework for
the Generation of Contemporary Facades', The Next
City - 16th CAADFutures, CCIS 527, Springer, São Paulo,
Brazil
Kolarevic, B (eds) 2003, Architecture in the Digital Age - De-
signing and Manufacturing, Spon Press, London, U.K
Leitão, A 2013 'Teaching Computer Science for Architec-
ture', Future Traditions - Proceedings of the 1st ecaade
Regional, FAUP Porto, Portugal
Leitão, A 2014 'Improving Generative Design by Com-
bining Abstract Geometry and Higher-Order Pro-
gramming', Proceedings of the 19th CAADRIA., Kyoto,
Japan
Leitão, A, Santos, L and Lopes, J 2012, 'Programming
Languages for Generative Design: A Comparative
Study', International Journal of Architectural Comput-
ing, 10(1), pp. 139-162
Lopes, J and Leitão, A 2011 'Portable Generative De-
sign for CAD Applications', ACADIA 11: Integration
Through Computation - Proceedings of the 31st ACA-
DIA, Calgary, Canada
Meredith, M 2008, 'Intro', in Sakamoto, T and Ferré, A
(eds) 2008, From Control to Design: Parametric/Algo-
rithmic Architecture, Actar
Moussavi, F 2006, The Function of Ornament, Actar

Elemental Intricacy
Architectural Complexity through Hard and Soft Material Agency
Alexandros Kallegias1 , Eleni Pattichi2

1
Architectural Association School of Architecture 2 EP Architects
1
alexandros.kallegias@aaschool.ac.uk 2 elinapattichi@gmail.com
This paper presents the research completed in AA Greece Visiting School 2016 in
Thessaloniki. The work integrates computational design and digital fabrication,
focusing on aspects of complexity in the making of an architectural interactive
prototype. During this research, the use of computation accommodates the design
and fabrication of indeterminacy and complexity in different scales and levels .
The prototype, Eos, projects on itself the urban characteristics of the city of
Thessaloniki. The aim has been to enable a 1-to-1 scale structure to act as a hub
of information, capable of communicating with human users through interaction.
As the city of Thessaloniki is characterised by specific environmental and urban
conditions, the prototype has been set to express such properties.
Keywords: pattern design, 1:1 scale construction, digital fabrication, design

simulation, interactive prototyping
INTRODUCTION gether 7 tutors and 24 students from different coun-

The gap of complexity that exists between unsimu- tries including the UK, Greece, China, Germany and
latable real world systems and our models of them Cyprus. The participants' background varied in terms
reflects the fact that the real world often behaves of their architectural knowledge, spanning from un-
in ways that directly contradict our ideas about it dergraduate students to professionals in the applied
(Frazer 1995, 117). The paper portrays the design field. The programme's pedagogical approach was
and fabrication of a large scale structure through based on team-based design and learning by experi-
the exploration of human communication coupled mentation, organized through a series of design stu-
with an active built architectural prototype. The case dios, technical tutorials, and design lectures. The ob-
study described in this paper is the outcome of an jectives of the research have been two-fold; complex
investigation which has explored low-tech materials form generation and interaction respectively. Whilst
with high-tech fabrication tools for the realization attention is given to the fabrication of the complex
of a complex 1:1 scale structure during an interna- architectural form that derives from design variations
tional 10-day academic programme in 2016. The proof a singular geometric element, the prototype's abil-
gramme has run as collaboration between the Archi- ity to actively respond to people within its area is also
tectural Association, School of Architecture and the taken into account. It is therefore argued that the fi-
Aristotelion University of Thessaloniki, bringing to- nal outcome is equally informed by the aim to act as

a human communication node and to be generated the contemporary tendencies of architecture are in-
through the manipulation of a fundamental geomet- heritably interconnected with technology. Technol-
ric element. ogy has provided an investigation platform for archi-
The final working prototype is described as a tects for the analysis and simulation of complexity
canopy housed indoors, with dimensions of 3 me- observed in natural phenomena able to solve archi-
ters width, 3 meters length, and 2 meters height. tectural problems. The origin of digital experimen-
The design constraints of the prototype have been tal methodology in natural phenomena proves that
governed by various factors including material, struc- complexity is not an end in itself but the mode and
tural, and interactive aspects. One of the key objec- style to solve complex problems of human inhabita-
tives has been to test the integration of hard and soft tion in space using simple and specific rules (Gruber
material systems in an aggregate configuration. MDF 2011, 110-115).
has been implemented as the material for the rigid el-
ements of the structural system, while Lycra has been COMPUTATIONAL SETUP
applied as the soft material as a skin to improve the The design brief of this research has been the pro-
interactive features of the installation. As the whole posal and construction of a one-to-one scale proto-
piece was devised to work as an aggregate system, type with kinetic and interactive parts that relate to
the connection details have been a significant con- the community's liveliness in the city. Innovative in-
stituent in the design and fabrication of the overall terventions which were informed by bottom-up rules
system. extracted from existing city conditions and activities
formulated the major focus of the design propos-
PROGRAMME STRUCTURE als. These activities focused mainly on the people's
The research programme is formulated into two movement within the city in different moments in
stages; the first stage which runs for 5 days focuses on history, the traffic, the city seafront, the connection to
creating various design models as early trials, while the University campus and the weather's local condi-
during the second stage, the remaining 5 days, the tions. Simulation tools were used to digitally contort
final 1:1 scale prototype is fabricated. The second the city's layout and superpose a sequence of flowing
stage sees all the participants as one team, combin- particles along the urban grid. Hence, digital tools
ing their knowledge gained from the first stage dur- computed a translation of urban information into or-
ing which they work as separate teams. Specifically, ganic geometric compositions and fluid trajectories.
the first stage has included 6 design teams address- Furthermore, a diversity of analogue representative
ing the research's objectives via the creation of mul- models was produced during the form-finding phase
tiple design iterations as a method which enabled of the research. With the completion of this stage, the
the teams to promptly investigate options and filter next step dealt with the realisation of Eos, the final 1:1
through the most successful aspects from each de- scale prototype.
sign proposal. The purpose hasn't been to create a Critical aspect of the research's aim has been
single optimal solution, but rather to swiftly test vari- the connection between form-finding methods and
ations that can provide insights via various pattern- form-making techniques based on a component-
testing. The proposals were developed using Au- type design strategy. Computational composition
todesk's Maya physics software as well as the para- and design analysis enabled the architectural solu-
metric plugin Grasshopper for McNeel Rhinoceros. tion of a sophisticated shell structure, whose com-
Computational tools as such have been an integral plexity was essential for the agency of space and
part of this approach emphasising the way in which structural constraints (Terzidis 2006, 117). The pro-
design and fabrication methodologies that govern totype's form emerges from a set of design explo-

rations that follow the basic principles of digitally out the design explorations (Figure 3). These explo-
simulated pattern formations. More specifically, a se- rations included simulations of cloud particles flow-
ries of forms was investigated through the genera- ing through the city's fabric and were made possi-
tion of doubly-curved surfaces that were then sub- ble through the combinatorial use of computational
divided in triangular parts via recursive technique in design and mechatronic tools that acted as a com-
parallel with aggregation methods of components. mon platform (Maya, Grasshopper and Arduino). This
In detail, the entire form of Eos is based on a low- digital platform of investigation facilitated the incor-
poly design piece, which acted as the base model poration of a vast array of construction criteria into
for its' triangular subdivision with Grasshopper. In a seamless complex composition which was synthe-
turn, each triangulated surface produced the set of sised using the triangle as the starting singular geom-
edge pieces; these were created to act as the struc- etry. The parametric design strategies facilitated the
tural skeleton for every surface. Each edge was then differentiated application of this triangle to build a
aligned with its' opposite in a way that every point complex structural system of polygonal components.
on that side could be connected with a straight line
Figure 1 to the next. (Figure1) Hence, all components' edges
Eos prototype were parallel to each other and vertical to the ground
construction model, to facilitate this system of connections. The triangu-
material and lated subdivision continued onto every surface.
information Eos has been realised with soft malleable fabric
diagram. parts for every surface and hard MDF wooden ele-
ments for each structural edge (Figure 2). The pro-
totype has been originally installed in the city's Uni-
versity campus where it has been exposed to the
people who are visiting or passing through. Whilst
constructed in a fixed yet fluid-like form the struc-
ture keeps its' active character with the integration
of electronics embedded in its' triangular compo-
nents. The use of the open-source platform Arduino
on top of the sophisticated forms designed in Maya
and Rhino Grasshopper result in a shell reactive to ex-
ternal stimuli.
The research included the possible integration
of parameters and constraints, spanning from the
urban and physical environment to the local inte-
rior space. The available materials and computer-
aided fabrication machines played an equally inte-
gral part on the design strategy. Through the digi-
tal generation of component-oriented forms, multi-
ple applications of elements were explored in differ-
ent spatial configurations. Triangular patterns with
a gradational change in size respective to pinned-
pointed locations over a two-dimensional grid were
combined with and fabrication methods through-

FABRICATION DETAILS Figure 2
In regards to the fabrication aspect of Eos, the analyti- Eos prototype
cal software plug-in Karamba for Grasshopper / Rhino fabricated in 1:1
was applied to inform the structure's cells detailed scale.
construction which subsequently led to the extrac-
tion of data for the computer-aided fabrication (Fig-
ure 4). Parametric relationships in reference to user-
oriented and material constraints guided the investi-
gation of the prototype's elemental geometry. Under
the scope of this methodology the application of spe-
cific digital design strategies facilitated the optimiza-
tion and refinement of the structure and its compo-
nents. Specifically, the design details of the connec-
tion system between the components' edges were
generated according to the thickness of the material
and the geometric relationships between the neigh-
bouring components (Figure 5). In regards to the
analogue and digital feedback, experimentation with
physical prototypes informed the digital model. The
minimum and maximum magnitudes and dimen-
sions of the polygonal components in the parametric
definition of Grasshopper were updated according to
the observation of the material's behaviour and the Figure 3
structural components' performance during testing. Eos lighting pattern
Furthermore, the geometry of the various polygonal detail.
planar parts was modified into double curved sur-
faces for the application of Lycra fabric. This dou-
ble curved surface was generated with a digital para-
metric transformation that was used to highlight the
intersections between the polygonal components.
The curvilinear geometrical variations of the edges
around each void intersection of components cre-
ated a dramatic effect in contrast to the components
edges which had a structural role (Figure 6). Addi-
tionally, the articulation of components in distance
to each other addressed as a detail the practical is-
sues of Lycra fabric application. The gap between the
polygonal components was used to fasten the fabric
after the process of assembly and as a design and de-
tail element to highlight the hypostasis of each com-
ponent. Hence, these design strategies combined
structural, constructional, user oriented and contex-
tual parameters in a unified entity under the scope of

a component oriented design. refractions in space. The differentiated orientation of
The parametrically generated structure of Eos re- components in space and the curvilinear shape of the
quired a file to factory continuum for its efficient re- overall geometry enhanced the organic distribution
alisation. The use of parametric software facilitated of light in the surrounding surfaces generating com-
the digital extraction of components in numbered 2- plexity from simple mechanical operations. More-
dimensional spreadsheets for computer aided fabri- over, the application of the laser cut pieces on the
cation in CNC and Laser machines. Engravings on fabric enabled the transformation to be more specific
each component edges and connection systems ac- and directed according to the patterns' orientation.
cording to the digital numbering were used for man- The actual pattern entails greater possibilities of light
aging the complex assembly in the short time frame effects "continuously variable and varying" (Kwinter
of the workshop. In this way, students with the 2002, 9). In this way, the sensorial attributes of the
help of the tutors were able to coordinate the com- installation integrate time in relation to geometry of
plex fabrication and construction of the components the polygonal patterns and real time data of prox-
in pieces and finally complete the assembly of the imity (Figure 8). Nonlinearity and indeterminacy are
entire installation throughout a systematic process. manifested through this system of movement in re-
Such a complex structure could not be realised with- lation to geometry, light and space generating juxta-
out the workflow of data between the digital model positions of lighting effects (Kwinter 2002, 9).
and computer aided fabrication techniques.
Figure 4 123 MDF pieces were connected mainly by using
Karamba a mortise system in combination to fixed steel ele-
screenshot with ments in the prototype's cellular structure. Initially,
displacement the linear MDF pieces were joined into the triangu-
values. lar and rhomboid components which were subse-
quently connected in the overall network of the shell
structure (Figure 7). During this stage, Lycra fabric
was applied on the components along with the pat-
terned cardboard elements. Soft, semi-transparent
fabric was attached to various cardboard opaque pat-
Figure 5 terns. This configuration was coupled to the kine-
Component matics of Arduino servos and sensors. Specifically,
connection detail. Arduino mechatronics were combined with the de-
sign of cut patterns on the Lycra to generate com-
plex lighting effects actuated by the distance sen-
sors. As the sensors send signals to Eos Arduino
"brain", a set of rotating servo motors, installed and
attached on the Lycra skin, begin to move. This mov-
ing mechanism has an immediate effect on the soft
Lycra skin which at first was designed to have a single
layer; however, after testing the overlaying of multi-
ple skins, the double skin was selected as the most ef-
fective one to filter the light with a more drastic effect.
The moving mechanism of servos behind the fabric
transformed the Lycra and generated triangular light

ing area for the moving mechanism. As the person Figure 6
moves closer, simulated projections of Thessaloniki's Detail of curvilinear
urban fabric are activated on Eos' Lycra skin. In par- edge versus
ticular, the city's grid appears initially mapped on Eos component linear
single rhombic component. As the distance sensors edge.
detect people moving the projected grid gets dis-
torted; city's roads and building blocks are gradually
deformed via fluid dynamics. In detail, the system
of interaction involves the detection of movement
while categorising each movement according to its
location. The detection field is distinguished in three
Figure 7
areas; each area correlates to a specific fluid simu-
Construction
lation that varies in scale according to the distance
process.
from the physical shell. Dynamic fluid simulations of
particle swarming the city grid and deforming its ge-
ometry are displayed while the Lycra fabric is physi-
cally distorted by a set of servos connected to it (Fig-
ure 8).
RESULTS
Digital and physical modelling facilitated the explo-
ration of part-to-whole relationships between the
overall structure and its components. This process Figure 8
contributed in the understanding of complexity in Detail of
different levels and scales. These different scales in- interaction.
clude the translation of contextual conditions into
projections that influenced the form-finding of the
overall structure, the user oriented interactive trans-
formations of the components' fabric with the use
of mechatronics, the composition of the shell struc-
ture in relation to the components' geometry and
the structural and constructional aspects of the con-
nection system. The continuum of design process
between large and small scale, digital and physical
realm was explored with the use of a digital work-
Regarding the interactive aspect of Eos, distance sen-
flow of data. Laser-cutting and CNC manufacturing
sors were placed to read people's presence within
techniques enabled a hands�on experience on the
the prototype's area. The sensors were positioned in
diverse range of digital fabrication systems and for-
a strategic part on the structure covering a sensing
mulated the starting point for the physical tests prior
distance of 450 cm. On the computational part, this
to realisation of the final form. The physical tests pro-
distance was divided in 3 different zones; each zone
vided a direct feedback for the prototype's final con-
correlating to a different simulation by Eos as well as
struction methodologies and set the guidelines for
the triggering of its' kinetic parts. The zone which
the translation of the 3-dimensional form into pla-
is the furthest from the prototype acts as the sens-

nar elements. A set of design rules was coded in programme's methodology empowered both partic-
Grasshopper for the digital generation of triangular ipants and tutors to deal with the complexity of con-
and rhomboid cell details as well as for the extraction ceiving Eos' design and equally enabled the smooth
of the 2D spreadsheets used during the fabrication transition from the digital realm to the materiality
process with the digital machines. In this way, stu- of the physical one. In combination with the pro-
dents were able to process the assembly of the ge- gramme's limited amount of time, the entire process
ometrically complex structure with the use of digital contributed to the pedagogical aspect of this initia-
parametric software. tive rendering students more engaged with various
In regards to the timeline of Eos' creation and design and fabrication methods.
due to the prototypical nature of this architectural
project, the different aspects of interaction, simula- ACKNOWLEDGEMENTS
tion, fabrication and analysis were running in paral- The work presented is part of the research under-
lel. On the aspect of interaction, the different zones taken at AA Greece Visiting School 2015. We would
in Eos' sensing and reacting abilities enables an ad- like to thank our tutoring team and students for their
ditional level of interplay where a person may at first great efforts. AA VS Thessaloniki 2015.
activate the shell's moving skin fabric and in turn the PROGRAMME DIRECTOR:Alexandros Kallegias
shell may attract the person to get closer in steps AUTH COORDINATORS:Nikos Kalogirou, Stavros
while different animations are on display. Vergopoulos, Anastasios Tellios
TUTORS:Alexandros Kallegias, Yue Shi, Eirini
CONCLUSIONS Vouliouri, Elina Pattichi, Sylvia Georgiadou, Arsenios
At every point of the design, fabrication and the as- Zachariadis, Alex A. Tsolakis
sembly process, the ability to constantly readjust the PARTICIPANTS:Evdoxia Besmerti, Dimitrios
Eos's overall form respectively to the alterations of its Chatzinikolis, Maria Christopoulou, Marina Di-
smallest detail has been a key driver in recognising mopoulou, Christos Kakouros, Stavros Grigorios
feeble aspects and putting forward design ideas for Kasimatis Voutyras, Dimitris Kollaros, Aikaterini
refinement. Moreover, the choice of working with Maria Kotsampasi, Emmanouil Megalooikonomou,
low-tech material for a demanding, complex proto- Safi Omarov, Niki Papacharalampous, Maria Pet-
type provided useful insights to the unpredictable sani, Artemis Psaltoglou, Styliani Rossikopoulou-
limitations of a system that aims to act as communi- Pappa , Anna Rizou, Eirini Dafni Sapka, Ioannis
cation node. The choice of MDF brought forward the Schoinas, Joanna Sotiriou, Joanne Spyridi, Georgia
hands-on understanding of benefits between adhe- Strinopoulou, Georgia Skartadou, Magda Triantafylli-
sive and mechanical anchoring of the pieces. The set dou, Konstantina Stella Tsagkaratou, Georgios Grigo-
of servo-actuators used to twist parts of the flexible riadis
fabric skin heightened the prototype's active charac-
ter depending on the set's positioning and the simul- REFERENCES
taneous designing, fabricating and analysing pro- Frazer, J 1995, An Evolutionary Architecture, Ellipsis
cesses were critical for the design team to swiftly Gruber, P 2011, Biomimetics in Architecture: Architecture
adapt to the complexity of unexpected results. Over- of Life and Buildings, Springer-Verlag
all, the purpose of the study has been to illustrate Kwinter, S 2002, Architects of Time: Towards a Theory of
architectural possibilities of altering in real-time the the Event in Modernist Culture, The MIT Press
Terzidis, K 2006, Algorithmic architecture, Routledge
built environment through the use of computational
design processes with digital fabrication techniques.
Furthermore, the active learning character of the

COLLABORATION AND
PARTICIPATION
CumInCAD 2.0: A Redesigned Scalable Cloud Deployment
Towards higher impact with openness and novel features
Tomo Cerovsek1 , Bob Martens2

1
University of Ljubljana 2 TU Wien
1
tomo.cerovsek@fgg.uni-lj.si 2 b.martens@tuwien.ac.at
CumInCAD is a cumulative index of publications related to 'Computer Aided

Architectural Design' (CAAD). It includes bibliographic data of approximately
12K records, which were predominantly derived from CAAD-related conferences,
such as ACADIA, ASCAAD, CAADRIA, eCAADe, SiGraDi and CAAD futures. A
brief historical overview of almost two decades of collaboration between the
University of Ljubljana and the above-mentioned CAAD-associations is provided.
After years of successful operation the previous interface became gradually
outdated, which called for new developments to assure continuous support to
open access to scientific knowledge. In this contribution, we explain the existing
status of the systems, its use, and the transition process to a cloud deployment.
Keywords: Open access, Cloud deployment, Bibliometrics, Google Scholar
INTRODUCTION Limited access. Access to CAAD scientific works was

The main goal that propelled the 'Cumulative In- very limited due to a small number (up to 250) of
dex of Computer Aided Architectural Design' (CumIn- printed copies of the proceedings. Conference pro-
CAD) was: (1) to build a 'collective memory' of sci- ceedings are seldom stored in libraries as complete
entific publications from conference proceedings of series, but remain mostly in the participants' book-
CAAD-associations; and (2) to make this memory un- shelves.
conditionally accessible to the scientific community
Almost two decades ago, we responded to the need
as a web-based bibliographic repository of works.
- to prevent data loss and to improve access to
The need for a digital repository of CAAD-related
the works - and set up a web-based bibliographical
publications was initiated by two important issues:
repository, which used Perl as its programming lan-
• Possible data loss guage and a Web Oriented Database (WODA). This
• Limited access made bibliographic records and full texts available to
a wide spectrum of interested researchers and practi-
Possible data loss. There was a fear that valuable tioners. A positive response from the scientific com-
scientific contributions would be lost or at least re- munity contributed to a growing reputation of the
searchers' time could be wasted on unnecessary re- CumInCAD. As the reputation of CumInCAD grew, so
work, or time-consuming literature research and in- did its content, encompassing today over 12,000 bib-
formation retrieval. liographic records and more than 9,000 full texts.
COLLABORATION AND PARTICIPATION - Volume 1 - eCAADe 34 | 485

Three important maintenance aspects for the status METHOD: TRANSITION STEPS
of CumInCAD and its survival are: As a response to the identified problems of the past
Maintenance of bibliographic data. The driving status, the following line of action was set out:
force for a continuing support for a commitment to Agreement on continuous support. We agreed to
collect, organize and make the content available via extend the collaboration between the University of
open access were volunteering activities by partici- Ljubljana and CAAD-associations to assure contin-
pating CAAD-associations and individuals. For quite ued development, support and migration of the ex-
some time, conference organizers have been using isting bibliographic repository. A migration plan was
digital master files in pdf, which are later used for the set up for the second half of 2015. A preliminary we-
repository along with bibliographic data. bometric and bibliometric analysis was performed to
assess relevance and impact.
• Past status: A time consuming preparation of Set-up of a helpdesk system. In order to support
bibliographic data with limited impact. both maintenance of the bibliographic data and mi-
• Goal: To ease bibliographic data preparation gration, we set up a help desk that provides easy and
with a helpdesk and to improve the impact. professional reporting and tracking. The helpdesk
Maintenance of the technical solution. Almost 20 system will also enhance end-user experience in the
years is a considerable amount of time for any IT future and will remain an integral part of CumInCAD.
system, especially in view of current developments From limited to full access. CAAD-associations de-
in web-based (electronic) publishing. A growing cided to upgrade the previous approach of Limited
user-base and amount of content on the one hand Open Access and to progress towards 100% Open Ac-
and continuous technological changes and advance- cess. Open Access to everybody leads to enhanced
ments for the www on the other represent constant visibility. In order to increase impact, we also de-
technical challenges for CumInCAD. cided to technically support google scholar crawling,
indexing and search engine optimization.
• Past status: The hardware and software sys- Porting the system to the cloud. In order to avoid
tem resided on a University of Ljubljana server the problems related to the closed technical system
that wasn't dedicated to CumInCAD only; e.g., that is affected by the hardware and software limi-
some services may require an update, which tations and lack of portability, we decided to go for
could (and did) cause problems to CumInCAD. a dedicated, standalone, flexible cloud solution that
• Goal: To migrate to a solution that is dedi- will allow migration and scalable deployment.
cated and not dependent on the institutional Deployment, testing and re-launching. This part
IT, but open, independent, and scalable. of the project encompassed a complete technical mi-
Maintenance of the open access path. Over the last gration of the entire database content, services that
few years, we established a culture of 'Limited Open were based on a common gateway interface, inter-
Access'. This means that CAAD-associations created preters and supporting libraries, including tens of
tangible "added value" by making their recent full GBs of data. Before a final re-launch we also did a
texts exclusively available to their members (read as complete graphical re-design of the entire interface
registered CumInCAD users). The intention was not and performed tests on different configurations
to generate business; on the contrary, the main aim In the next two sections of the paper we provide
was to keep the repository on a shoestring budget. detailed information on the analysis and results, fol-
lowed by conclusions and the plans for the way for-
• Past status: Limited Open Access ward.
• Goal: Full open access to CumInCAD.
486 | eCAADe 34 - COLLABORATION AND PARTICIPATION - Volume 1

ANALYSIS OF USE PATTERNS on 1st of May 2015. One significant low-use period
The analysis of use patterns is important as it helps us is during Christmas, and at the end of the time-scale
plan technical and organizational improvements of (April 2016), where we see that the number of page
the web repository. For this purpose we used Google views is approaching zero, because we completely
Analytics, a free web analytics service that provides ported the system to a new cloud solution at that
statistics and basic analytical tools for search engine time. Users usually view about three pages per ses-
optimization (SEO) and marketing purposes. The ser- sion, which means that they find what they are look-
vice is available to anyone with a Google account. ing for very quickly, or they quickly find out that the
Once the Google Analytics is active, it starts to content is not appropriate for them. End-users usu-
collect, analyse and format the data to different visual ally stay on the site less than 2 min. We identified
representations. The dashboard data on Figure 1 give about 260 different sessions per day undertaken by
important information on the nature of CumInCAD approximately 200 different users.
end-users: we have over 76,000 users per year, 20% As visible on a diagram we have one significant
or approximately 15,000 returning users (i.e. they re- peak (high-use) of CumInCAD in September. As the
turned to the CumInCAD web site at least once). The number of users is not proportional to the page-
y-axis shows page views (number of pages that were views, we may assume that one of the end-users was
viewed per week) in the period of one year, starting a bot.
Figure 1
An overview of web
access analysis for
CumInCAD from 1.
May 2015 to 30.
April 2016

RESULTS OF MIGRATION Figure 2
There are two important parts to migration: the work Demographic
either invisible to the end-users (back-end) or visible analysis of web
in other systems (e.g., Google), and the work that is access by country
visible to the end-users as an interface (front-end). from 1 May 2015 to
Figure 3 presents an example of an effort that 30 April 2016
is not visible to the end-users of the CumInCAD di-
rectly, but through a Google Scholar interface (once
CumInCAD was opened and ready for indexing, it was
crawled and indexed by Google robots).
Figure 3
Google Scholar
search results
containing
CumInCAD results
Back-end: Towards cloud deployment

We executed the following development steps to-
wards open cloud deployment:
As illustrated in Figure 2, almost 18% of the traffic • Open access search engine friendly solution;
comes from US; in terms of use of the system per • Standalone portable system implementation;
capita, Austrians would be the most active end-users. • Cloud deployment of the system with backup.
We can also conclude that CumInCAD is truly interna-
tional. Open access search engine friendly solution.
The referral analysis also shows that the end- Preparation of a pre-porting review, including bib-
users did not come to CumInCAD from Google, but liographic measures and review of web access stats
they use direct referrals. This means that the imple- and impact. These steps included adjustments of
mentation of Google Scholar would make an impact. metadata. The indexing of the bibliography re-
Based on the use patterns we may also predict quired adjustments of the metadata system, in addi-
how scalable a technical solution should be. On the tion we had to prepare dedicated index files for web
other hand, we can use the Web analytics as an audit crawlers that connect bibliographic records. Testing
trail of the activities on the web systems and allows of the indexing on a Lucene Standalone Engine and
for the measure the affect the developments, and this on Google Scholar; implementation of a novel data
is what we did during the migration and final transi- model for indexing; preparation of initial indexing for
tion to CumInCAD 2.0. the whole repository using pre-processing and XML
records.

Figure 4 Selected cloud deployment - Amazon Elastic Com-
Google Scholar pute Cloud (EC2) - replaced the University server.
public display of The back-up is also provided as a part of the so-
author record with lution. Backups can be used for creation of new in-
the display of stances at different physical location or just for a re-
bibliometric store.
statistics Amazon virtual machines are called 'instances',
which may have different hardware configuration
that match specific requirements of the server. For
an initial set-up one can use Amazon Machine Image
Search Engine Optimization (SEO) that includes in- (AMI), which is a template that contains the software
dexing and referencing by way of the standard configuration (e.g. an operating system, an applica-
Google search engine to assure increased ranking. tion server, and applications). Amazon cloud services
A trial implementation of the customized search ap- allow for easy monitoring of the performance, pro-
proaches triggered technical decisions. We con- viding the benefit of elasticity.
cluded this phase of migration by sending an invita-
Figure 5 tion to CumInCAD users to create a Google Scholar
A cloud service account.
interface that Google Scholar Citations provide a simple way
provides an for authors to keep track of citations to their articles.
overview of An author can trace back information about who is
instance (top) and a citing his/her publications, graph citations over time
completely web and compute several citation metrics. Authors can
based management set up a public profile to appear in Google Scholar re-
of a virtual machine sults when people search for them (see Figure 4).
for CumInCAD 2.0 Standalone portable system implementation.
The second part of the transition was far more de-
manding as it required considerable technical mod-
ifications to be executed on time. This included
porting of a database system to a new standalone.
The database management environment used for in-
put and output (data entry and automatic index file
preparation) was adapted, tested and implemented.
Cloud deployment of the system with backup.
We used the Turnkey Hub implementation and im-
plemented TKLBAM portable backup. Adaptation of
the back-end solution(s) was made available to man-
age specific Linux deployment. The initial instance
was based on open source environment allowing
to be openly transferred to other cloud providers.
WODA is still being used for data entry and automatic
index file preparation for Google Scholar.

Figure 6
CumInCAD 2.0
selected instance
CPU utilization, Disk
I/O and network
traffic
Testing the ported version on Amazon EC2 was siveness and robustness of operations. In this phase
a significant task in order to determine which con- we analysed CPU utilization, read/write activities and
figuration matched the user requirements of respon- network traffic (Figure 6).

Figure 7
Redesign of the
start page
simplified with
minimal menu for
main features.
Figure 8 Front-end: Interface redesign

Listing of search Responsive design considerations guided the re-
hits: Example of design of visual representation and the relaunch of
eCAADe2015- the repository website. There were two main princi-
proceedings. ples in the redesign: (1) simplicity and (2) sufficient
quality of display on multi-modal devices.
Note that we also moved to a new domain,
cumincad.org, in the process of migration, while
all existing references to cumincad.scix.net are re-
directed to a new domain (papers.cumincad.org).
The focus of delivery by way of the interface is di-
rected towards multiple devices (also for mobile de-
vices - see Figure 9) and above all the interface fol-
lows typical search patterns, such as: Certain pro-
ceedings of a certain CAAD-association; All published
papers of a CAAD-association; Papers of a certain au-
thor (personal bibliography), etc.

CONCLUSIONS AND DISCUSSION Maximising attention to the publication output of
The CumInCAD 2.0 cloud deployment proved to be individual authors we strengthen the potential im-
efficient, scalable and independent. The solution is pact by making the bibliographic content available
not tied to an institutional IT infrastructure, but is for crawling and indexing by Google and other search
portable to either private or public clouds. We also engines. In this context, we have strengthened
moved to a new web domain and assured smooth the open publishing coverage and implemented ad-
transition with URL rewrite. As CumInCAD is "owned" vanced indexing and searching. We used the best
by the CAAD-associations without being a profit- advantages offered by both worlds, independent de-
oriented enterprise intended to generate business, velopments, portable cloud solutions with integra-
CAAD-associations now have the power to fully con- tion to the most open commercial developments.
trol the technical part of the CumInCAD solutions, do- For example, a Google team does the daily develop-
main and content. ments on indexing, searching, ranking, integration,
and bibliometrics, while we still do some develop-
Figure 9
ments, e.g. Semantic Web, advanced clustering, data
Listing of search
mining, along with constant maintenance. We be-
hits: conference
lieve that the work invested in this project will result
series.
in higher impact through open access supported by
search engines and services like Google scholar.
Throughout the implementation we observed
the performance of selected configurations of in-
stance and the capacity of selected resources. One
of the great advantages of Cloud Deployment is scal-
ability and extensibility.
Cloud interoperability also prevents cloud ser-
vices and software lock-in. With the help of server
usage tracking and support the end-user experience
is enhanced and technical solutions can be scaled if Figure 10
needed. In order to assure safe operations in terms Advanced search
of data integrity and backup, we perform daily back- interface
ups which enable us to restore the systems to a spe-
cific date. This way we continue to assure two main
goals: safety of valuable bibliographic data and open
access.
ACKNOWLEDGEMENT
We gratefully acknowledge the initial work by prof.
Žiga Turk and the support of Robert Klinc before the
transition, as well as the work conducted by Darko
Malić during and after transition.
Finally, help-desk support services to the CumIn- REFERENCES

CAD end-user community and bibliographic data [1] http://papers.cumincad.org
providers (e.g., editors ) have been implemented.

Collaborative Design with Mobile Augmented Reality
Leman Figen Gül1 , Süheyla Müge Halıcı2
1,2
Istanbul Technical University
1,2
{fgul|halicis}@itu.edu.tr
In early design process, designers employ several media for externalizations of

their design ideas which facilitate decision making, discussion and evaluation.
Especially models are the most common representation tools and are used
generally to understand and make spatial reasoning on the design ideas. Besides
the traditional methods, today, the technological developments bring new ways
for collaboration and design in 3D; the mobile augmented reality (MAR)
technology is one of them. MAR is augmented reality technology which is
provided by mobile devices. This study focuses on the early design process of
collaborative designers' communication when they are working with the MAR
technology. We developed a MAR environment for designers, and conducted an
experiment to understand the key elements of the interaction of the designers with
the interface. The communication and interaction of the designers are analyzed
using the protocol analysis method. The results show that the MAR technology
supports the co-design activities encouraging the designers to manipulate the
created artefact. The results of the study would be indicative for future studies.
Keywords: collaborative design, mobile augmented reality, protocol analysis
INTRODUCTION the co-design situation have been studied by vari-

Augmented Reality (AR) technology combines the ous researchers (Kim and Maher 2005; Seichter and
real and the virtual by providing the 3D real time in- Schnabel 2005). While using the augmented reality
teractivity (Azuma 1997). In the field of design, the technology, collaborative designers would have op-
AR technology is mainly considered as a representa- portunity to interact with each other during the de-
tion tool, which is crucial in a collaborative design sit- sign process which is essential in a co-design con-
uation. As the representation tool for co-designers, text. In the urban design context, Seichter and Schn-
design situations where several designers work to- abel's (2005) indicated how designers could commu-
gether with a design problem, the AR technology nicate and create early design ideas with the AR as a
provides the virtual models which the discussions medium (Seichter and Schnabel 2005). Another im-
would take place on. One of the important features portant features of the employment of the AR tech-
of the AR technology is that, the designers can in- nology is for facilitating the spatial reasoning of the
teract with the design artifact through the graphi- design proposal. Kim and Maher (2005) compared
cal user interface (GUI) and/or with gestures. Using the spatial cognitive impact of the designers with the
the AR technology as a representation tool during co-design situation on the table-top augmented re-

ality system and the graphical user interface. Most construct the design proposal together as a team
of those AR research used the desktops, monitor dis- (Hennessy and Murphy 1999, Seitamaa-Hakkarainen
plays and built-in webcams or the displays projected et al. 2000). In the field of collaborative design,
on the table-top systems for the collaboration. Over the provision of a more efficient communication pro-
the last decade, due to the emergence of smart- cess, especially for the healthy communication, in-
phones and tablets, the cloud technology and the teraction and fruitful presentation, is one of the key
wireless high-bandwidth internet, the use of the mo- elements. In particular, external design representa-
bile augmented reality (MAR) have increasingly be- tions that have a significant role as the communica-
come a common tool. MAR technology provides aug- tive resources are the objects of interaction (Robert-
mented reality technology through mobile devices. son 1996).
This means users with MAR do not have to remain When design thoughts are externalized, each el-
their position due to fixed devices like desktops. The ements of the design artefact contains both the prop-
mobile augmented reality technology gives an op- erties of future interpretations and the negotiation
portunity to the collaborative designers providing a of further developments of the design idea (Gül and
working environment where the designers are able Maher 2009). Sketches and physical models are com-
to freely move around the representation and are mon traditional external design representation tools.
able to visually inspect the design artifact. Although In the collaborative design settings, the design rep-
there are some studies to employ mobile augmented resentations in the form of sketches or models that
reality to understand the user's visual spatial cog- can be pointed to, talked about (Perrya and Sander-
nition (Shelton and Hedley 2004), we do not know son 1998) would play an essential role both in one's
the impact of recent mobile augmented reality tech- conversation with oneself and with others. In partic-
nology on designers' communication and interaction ular, Kvan and Thilakaratne (2003) pointed out that
with the design artifact. models offer benefits of approachability, tangibility,
The aim of this study is to understand the manipulability and collaborative engagement.
changes of designer's communication and interac- With the new developments in the advanced de-
tion with the external design representation while sign technologies for the externalization of the de-
they are moving a physical model making to digi- sign ideas, working with the augmented digital mod-
tal model making using mobile augmented reality. els become more common in the early stages of the
In order to understand the changes of their interac- design process. In addition, the graphical user inter-
tions, we analyze the verbal and visual data of the faces have been increasingly employed between ar-
co-designers that we captured during the design ses- chitects, designers and engineers as the representa-
sions with the Protocol Analyses method (see, Er- tion and communication tools that become afford-
icsson and Simon (1993) for the protocol analysis), able, reliable and easy to use. Those environments
which is widely used in collaborative design settings provide opportunities of the manipulation of the vir-
(Maher et al. 2006; Gül 2007). In this paper, the result tual design artifacts that is not possible before in the
of a pilot study is presented and discussed. co-design situation. In continuum reality and virtu-
ality (Milgram and Kishino 1994), augmented reality
MOBILE AUGMENTED REALITY IN COLLAB- offers new possibilities for the collaborative design-
ers. Studies took attention to mobile augmented re-
ORATIVE DESIGN
ality with the interest of establishing mutual aware-
Collaborative design activity involves effective com-
ness between collaborators (Rekimoto 1996).
munication and working on the shared design pro-
posal in order to determine the design goals, in-
spect the problem space and the constraints, and

Figure 1
The mobile AR
interface. The smart
phones and
Android-İOS tablets
can be used.
METHODOLOGY
In this study, in order to understand the interaction
Figure 2
and communication activities of the AR users, we de-
The pilot study: PM
veloped a marker-based mobile AR (MAR) platform,
(left) and MAR
as shown in Figure 1. In order to analyse and docu-
(right)
ment the collaborative behaviour of designers in two
environments-The
different design settings, the pilot study is conducted
tables’ view is
at the Department of Architecture in the Istanbul
projected on the
Technical University: 1) Co-design with the physical
glass table-top
modelling environment (PM), and 2) Co-design with
(bottom).
the enhanced marker-based mobile AR environment
(MAR). A pair of architects collaborating on two dif-
ferent design tasks with similar complexity are exam-
ined. The assumption of the study is that a compar-
ison of the same architects in two different environ-
ments would provide better indication of the impact
of the environment than using different designers
and the same design task. With these ideas in mind,
the pilot study with two design settings is conducted,
as shown in Figure 2.

A marker-based mobile AR environment is devel- the screen view of the shared tablet is also captured.
oped for the study by using the Unity3D [1] game All the video and audio data are fed into a DVR (Digi-
engine with the Vuforia [2] AR plug-in. In Vuforia, tal Video Recording) system. The data is analysed us-
the library of markers is adapted to match the tar- ing the protocol analysis method that requires tran-
get objects that are the basic primitives; the cube, scription, segmentation and coding the data. We
the sphere and the cylinder objects. In order to aug- have done the transcription and then, the data is seg-
ment the target objects on to the scene, Vuforia plug- mented. Finally, the segmented protocols are coded
in requires the definition of image targets for each of using the customized coding scheme, as shown in Ta-
the target objects. The image targets are the unique ble 1.
predefined 2D shapes, similar to the QR codes, which
are recognized by the device's camera. In the de- CODING SCHEME
veloped application, nine image targets are defined The actions and utterances are recorded, and then
and assigned to the primitive objects. The applica- the transcripts of their dialogue are segmented by
tion also has the interface components to manage using Suwa et al.'s (1998) action and intention rule.
the augmented objects, as shown in Figure 1. The The segmented protocol is coded and analyzed. Ta-
users are able to operate dragging, rotating, scaling ble 1 shows the communication and interaction cod-
and changing the colour of the objects. In addition, ing scheme which has six main categories. In this pa-
Vuforia enables extended tracking, once the image per, we present the results of some of the categories.
target is detected, it keeps the objects on the screen, These are highlighted with gray in Table 1.
even the image target is moved out of the vision of
the device's camera. Table 1
In order to have the collaborative working envi- Communication
ronment, both a shared view of the tablet's display is and Interaction
projected onto the glass-top table and each design- coding scheme
ers' tablet view is shared. A physical model of the site
in 1:500 scale is also provided to help the visualisa-
tion of the given design context. The designers are
able to inspect the site in both physically and virtu-
ally, as shown in Figure 2.
The first phase, (PM) co-design with the physical
modelling tools as the baseline study: the site model
is given in 1:500 scale. The cardboard plates and
small boxes in different sizes, pen, paper, knives, and
some other modelling materials are provided. The
second phase, (MAR) co-design with the enhanced
mobile AR platform: Android and İOS tablets, the
markers, the physical site model in 1:500 scale and
the glass-top table with the shared view of the AR in-
terface are provided to the designers.
The designers had a training session followed by
the experiment. During the experiment, three cam-
eras are used to capture the designers' actions and The first category is the 'design actions' that has
verbalization. In addition, in the second phase (MAR) "physical", "perceptual" and "semantic" codes. The
'physical' refers to discussions on the physical appear-

ance of elements; shape, size, texture, color. The rations of each operation, the beginning of the ses-
second code, the 'perceptual' contains the discussion is on the left, and the end of the sessions is on
sions on the visual spatial relationships of elements. the right. The analysis shows that in the PM session,
The 'semantic' refers to both the functional and con- the 'making' and the 'creating new element' actions
ceptual information of the design idea; discussions are parallel to each other, on the other hand, in the
on the tasks; set up goals, proposing idea and re- MAR session, the 'making' and the 'revising element'
trieve knowledge. The second category contains the actions are parallel to each other most of the time,
generic design process actions: 'analysis', 'synthesis' as seen in Figure 3. The analysis also indicates that
and 'evaluation'. Third category includes the content in the PM session, there are more the 'creating new
of the communications that are the discussions on element' actions than the 'revising element' actions'.
the 'design' idea and the discussions on the 'mak- On the contrary, in the MAR session, the 'revising
ing' of the model. Based on the visual data, we element' actions occur more, as shown in Figure 3.
also analyzed the modelling activity with the cate- We also observe that in the PM session, the design-
gory 'making action', that has two codes: the 'creat- ers prefer to use the given modules that are cubes,
ing a new element' and the 'revising the element' ac- prisims in different sizes. They are able to revise them
tions. The interaction with the given elements and by stacking up, moving, dragging or rotating. In the
the tools-interfaces are also investigated: In the PM MAR session, once the markers are introduced, it will
session, the physical modelling activity consists of stay in the scene (on the physical model and on the
two codes: 'w/module' and 'w/model'. The 'w/mod- tablet's view) and then the designers are able to ma-
ule' refers to the given pre-modeled pieces that are nipulate the elements from the user interface. Thus
the basic prisms. The "w/model" refers to the cre- we can speculate that the MAR technology encour-
ation of the model using the cardboards by cutting, ages the designers to manipulate the created arte-
gluing etc. In the MAR session, the interaction oc- fact.
curs with the interface and with the markers: 'w/- The communication content and the interaction
marker' and 'w/UI'. The 'w/marker' refers to the ma- with the design tool actions are shown along the
nipulation of the marker's location and position on timeline of the sessions in Figure 4. In the PM ses-
the table. The 'w/UI' refers to the manipulation of sion, the design related communication content are
the virtual objects' properties on the user interface not parallel to any of the interaction operations, ex-
on the mobile devices. The 'w/table-top' refers to cept very little parallel actions (the 'w/module' and
the design communication through glass-top table the 'design' actions) occur as shown in Figure 4. In
. In both sessions, there are two more codes; 'note' addition, in the PM session, the 'making' and the
and 'sketch'. 'note' refers to taking quick notes and 'w/module' - the 'w/model' actions are parallel to
'sketch' refers to design activities of any sketching or each other. We may suggest that working on the
drawing. The final category, the 'visual communi- physical model interrupted the designers, thus there
cation' includes the physical activities of the design- is non-concurrent idea verbalization and interaction
ers that would support communication that are 'ap- actions. Thus the making and idea generation be-
proval with eye-contact' and 'gesture'. come separated. In the MAR environment, the design
idea generation occurs parallel to the w/marker and
RESULTS w/UI actions in longer time segments, as shown in
We observe differences in between PM and MAR ses- Figure 4. Thus, we can say that the MAR environment
sions. The communication content and the making supports the idea generation. A speculative finding
actions are shown along the timeline of the sessions might be that the making and idea generation are
in Figure 3. Each horizontal bar represents the du- not interrupted with the MAR interface, thus the en-

Figure 3
The communication
content and
modelling element
actions in the PM
and the MAR
sessions.
Figure 4
The communication
content and the
interaction with the
external design
representation
actions in the PM
and the MAR
sessions.
hanced MAR environment does not increase the cog- Figure 5

nitive load of the designers. This finding needs to be The duration
supported with the experiments with more partici- percentages of the
pants. visual
Figure 5 shows the duration percentages of the communication-
'gesture' codes in both design sessions. There is a the gesture codes
drop in the duration of the gestures in the MAR ses- in both sessions.
sion, as shown in Figure 5. Our observations show
that in the PM sessions, the designers gestured more
when they were explaining their ideas. In the MAR
session, on the other hand, the designers refers to vi-
sual design element on the screen by pointing, tap-
ping etc. Researchers argued that gestures appear
to aid in conceptualization, and in communication
(Kendon 1994). We can argue that the result supports
the findings of (Wesp et al. 2001: 595) that is: 'partici- CONCLUDING COMMENTS
pants describing a picture to an observer do not ges- We have studied designers using two modes of de-
ture as frequently if the picture is visible during the signing: the physical modelling as the baseline study
description'. In the MAR session, the 3D virtual de- and the marker-based mobile augmented reality en-
sign artefact is easy to inspect and communicate on vironment, allowing us to compare their communi-
the screens, thus the designers did not gesture fre- cation content and interaction within the environ-
quently. ments. Based on the results of the analysis and our

observations, we conclude that designers adapt to Gül, LF 2007, Understanding collaborative design in differ-
the MAR environment showing different behavior ent environments: Comparing face-to-face sketching
and communication patterns in each design sessions. to remote sketching and 3D virtual worlds, Ph.D. The-
sis, University of Sydney
Our findings fall into two categories: 1. Facilitating
Gül, LF and Maher, ML 2009, 'Co-Creating External
the idea generation 2. Facilitating the design devel- Design Representations: Comparing Face-To-Face
opments. Sketching to Designing in Virtual Environments',
The study shows that in the MAR environment, CoDesign - International Journal of Co-creation in De-
the designers continuously generate design ideas sign and the Arts, 5(2), pp. 117-138
and work on the development of the design artefact Hennessy, S and Murphy, P 1999, 'The potential for col-
laborative problem solving in design and technol-
at the same time. They show concurrent and longer
ogy', International Journal of Technology and Design
time spans on both activities. Strategically, this is an Education, 9(1), pp. 1-36
important finding because it implies that the intro- Kendon, A 1994, 'Do Gestures Communicate?: A Review',
duction of the enhanced mobile augmented reality Research on Language and Social Interaction, 27(3),
environment into the co-design process conserves pp. 175-200
the essential aspects of designing and let the design- Kim, M and Maher, ML 2005, 'Comparison of designers
using a tangible user interface & graphical user in-
ers communicate and interact within the tools and
terface and impact on spatial cognition', in Gero, JS
each other while working on their tablets. and Lindermann, U (eds) 2005, Human Behaviour in
The second category of impact has to do with Design '05, University of Sydney, Sydney, pp. 81-94
the manipulation and the development of the design Kvan, T and Thilakaratne, R 2003 'Models in the De-
proposal. We observe that while they were working sign Conversation: Architectural vs Engineering',
on the development of the visual design artefact by Proceedings of the Second International Conference of
the Association of Architecture Schools of Australasia,
manipulating the size, position, location of the ele-
University of Melbourne, Australia, pp. 28-30
ments etc. they are also verbally discussing the de- Maher, ML, Gül, LF and Bilda, Z 2006 'Studying design be-
sign related issues. This finding is also important behaviour in collaborative virtual environments', Pro-
cause we can argue that the MAR environment does ceedings of 16th World Congress on Ergonomics
not require extra cognitive load. Milgram, P and Kishino, F 1994 'A TAXONOMY OF MIXED
In conclusion, mobile AR technology is promis- REALITY VISUAL DISPLAYS', Paper presented at the IE-
ICE Transactions on Information Systems
ing for co-designers as both today's and future's inno-
Perrya, M and Sanderson, D 1998, 'Coordinating joint de-
vative design platforms. This study would be a guid- sign work: The role of communication and artefacts',
ance for the further developments for MAR usage in Design Studies, 19(3), p. 273–288
design studies. Rekimoto, J 1996 'Transvision: A hand-held augmented
reality system for collaborative design', Paper pre-
Acknowledgements sented at the Proc. Virtual Systems and Multimedia
The research is funded by the Scientific and Tech- Robertson, T 1996, 'Embodied actions in time and place:
The cooperative design of a multimedia educational
nological Research Council of Turkey (TUBITAK), PN: computer game', CSCW, 5(4), pp. 341-367
115K515. Seichter, H and Schnabel, MA 2005 'Digital and Tangi-
ble Sensation: An Augmented Reality Urban Design
Studio', Paper presented at the Tenth International
REFERENCES Conference on Computer Aided Architectural Design
Azuma, R 1997, 'A Survey of Augmented Reality', Pres- Research in Asia, New Delhi, India
ence: Teleoperators and Virtual Environments, 6(4), Seitamaa-Hakkarainen, P 2000, 'Collaborative designing
pp. 355-385 in a networked learning environment', in Scrivener,
Ericsson, KA and Simon, HA 1993, Protocol Analysis, MIT SAR, Ball, LJ and Woocock, A (eds) 2000, Collabora-
Press, Cambridge tive Design, Springer-Verlag, London, pp. 411-420

Shelton, BE and Hedley, NR 2004, 'Exploring a cogni-
tive basis for learning spatial relationships with aug-
mented reality', Technology, Instruction, Cognition
and Learning, 1(4), pp. 323-357
Suwa, M, Purcell, T and Gero, JS 1998, 'Macroscopic anal-
ysis of design processes based on a scheme for cod-
ing designers', Design Studies, 19(4), pp. 455-483
Wesp, R, Hesse, J, Keutmann, D and Wheaton, K 2001,
'Gestures maintain spatial imagery', The American
Journal of Psychology, 114(4), pp. 591-600
[1] www.unity.com
[2] www.vuforia.com

De Humani Corporis Fabrica - Fabricating Emotions
through Architecture
Maria da Piedade Ferreira1 , Andreas Kretzer2 , José Pinto Duarte3 ,
Didier Stricker4 , Benjamin Schenkenberger5 , Markus Weber6 ,
Takumi Toyama7
1
Fachbereich Architektur Technische Universität Kaiserslautern 2 Fachbereich Ar-
chitektur - Technisches Universitaet Kaiserslautern 3 Faculdade de Arquitectura da
Universidade de Lisboa 4,5,6,7 Deutsche Forschungszentrum für Künstliche Intelli-
genz GmbH (DFKI)
1,2
{m.piedade.ferreira|andreas.kretzer}@architektur.uni-kl.de
3
jduarte@fa.ulisboa.pt
4,5,6,7
{Didier.Stricker|Benjamin.Schenkenberger|Markus.Weber|Takumi.Toyama}@dfki.de
This paper describes an experiment that took place as the last day of the elective
course "De Humani Corporis Fabrica - Fabricating Emotions through
Architecture". The goal of the Experiment was to evaluate the emotional
experience of 1:1 scale architectural settings. The Experiment was part of the
research for a PhD thesis which describes the relationship between architecture
and the body since Vitruvius until the current theories of embodiment. The
referred to thesis proposes the incorporation of the corporeal practices of
performance art in the teaching of Architecture and combines these with the use
of emotion measurement methodologies in order to evaluate the experience of
architectural space and design objects during the design process.
Psycho-physiological changes in the body's sensory perception during the
performances were evaluated through the combined use of biometric technology
(e-health platform), a Presence Questionnaire and a SAM chart. The course was
attended by 4 students who participated as subjects in the Experiment. The
Experiment had the technical support of the DFKI - Deutsches
Forschungszentrum für Künstliche Intelligenz, which provided the machinery
necessary to collect the biometric data during the Experiment and the knowledge
needed to process them.
Keywords: Corporeal Architecture, Performance Art, Emotion Measurement,

Pedagogy

INTRODUCTION
This paper describes an experiment entitled "De trum für Künstliche Intelligenz, which provided the
Humani Corporis Fabrica - Fabricating Emotions machinery necessary to collect the biometric data
through Architecture" that took place as the last day during the Experiment (e-health platform) and the
of an elective course of the curriculum of students knowledge needed to process them.
of Architecture. The title of the experiment is a ref-
erence to the first anatomy treatise with the same Figure 1
name (from Latin to English, "On the Constitution of Subject prepared
the Human Body"), in which its author (Vesalius 1543) for performance
presents dissections of human bodies in order to ex- with biometric
plain its inner workings. This title was suitable for platform and
the experiment also because the original Latin word neutral mask
for fabrica can have the multiple meaning of "fab- installed.
rication, constitution, or construction". It was taken
as the motto of an experiment in which the subjects
would design, fabricate, assemble, and perform in ar-
chitectural settings, taking in consideration that the
design elements in such settings would produce spe-
cific physical and emotional effects on users, or as the
subtitle suggests "fabricating emotions through ar-
chitecture."
The course was attended by students of Architec-
ture and was hosted by the Digitale Werkzeuge at the
Fachbereich Architektur of the TU Kaiserlsautern dur-
ing the Winter Semester of 2013/2014. The goal of
the experiment was the evaluation of emotional re-
action to changes in the sensory perception when a
user is performing the same actions in settings that
have the same design but significantly different di-
mensions. The goal of the course was to increase stu-
dents' awareness of the direct effect that variation in
the values of basic design parameters, such as scale
and dimensions, produces on the human body, while
conditioning movement.
Psycho-physiological changes in the body's sen-
sory perception during the performances were eval-
uated through the combined use of biometric tech-
nology, a Presence Questionnaire (Witmer and Singer
1995) and a SAM (Bradley et al. 1994) chart. The
course was attended by 4 students who participated DESCRIPTION
as subjects in the Experiment. Experiment 4 took For the final presentation, during which the Experi-
place on the last day of the course and had the tech- ment would take place, students were asked to pre-
nical support of the DFKI - Deutsches Forschungszen- pare a performance that explored chosen actions in
settings designed and constructed at 1:1 scale. Stu-

dents also were required to wear black comfortable sensory data, having in mind that our main hypothe-
clothing and a neutral mask, as was the case in the sis was:
previous Experiments undertaken on the course of
this PhD research. (see Figure 1) Experiment 1 eval- • H1 - a user's emotional response as "com-
uated changes in the sensorial landscape when the pelled or not compelled", "positive or neg-
body's range of motion is conditioned by the inter- ative", "aroused or not aroused" and "domi-
action with objects that work as extensions, Exper- nant or dominated" to an architectural setting
iment 2 evaluated changes in sensorial landscape can be evaluated through objective measure-
when the body interacts with architecture models ments of emotion using the Presence Ques-
enhances with sound and smell (immersive mod- tionnaire, the SAM chart and the e-health
els) and Experiment 3 evaluated changes in sensorial platform.
landscape when the body interacts in space condi- Two secondary hypotheses were formulated:
tioned by body extensions and restrictions. Experi-
ment 4 was therefore the last experiment and had the • H2 - the feeling of presence and emotional ac-
aim to extend the previous research to 1:1 scale, rea- tivation can be induced through the perfor-
son why the settings were built with real-life proper- mance in analogical models, in this case, 1:1
ties of weight, sound and texture. Experiment 4 was scale architectural settings;
chosen to be described in this paper as it is also the • H3 - the somatic techniques of "performance
one which is most related to the research tradition in art" and "emotional design" are an effective
movement and ergonomic studies in a design con- strategy to develop corporeal awareness and
text. stimulate the creativity of students and de-
The task that prepared Experiment 4 required signers;
students to choose an architectural setting, for exam-
To verify these hypotheses, the experiment was de-
ple a basic unit such as a dining table and a bench,
veloped considering four stages:
and to design at least two variations of the setting,
for example one with a dining table 1 m high and the 1. Identify the design characteristics that are
other with a dining table 0.5 m high. (see Figure 2) more suitable to induce certain sensations in
Students were encouraged to include movement no- a user, such as "positive, aroused, dominant,
tations in the design process in order to analyse the compelled" or "negative, not aroused, domi-
different stages of basic actions, such as working with nated, not compelled";
a laptop (see Figure 3) or baking a cake in a kitchen. 2. Design an architectural setting so that those
The settings were designed and produced with the characteristics are the most important aspects
combined use of analog and digital tools and were in- of the design;
stalled in a cubic space. This cubic space is intended 3. Perform experiments with users interacting
to function as a living reference to Muybridge´s work with such architectural settings, while being
on the study of the human body in motion. The monitored by camera and assess their emo-
performances with the objects were documented by tional experience through the use of a PQ, a
photography and film. Biometric and emotion mea- SAM chart, and a biometric platform;
surement tools were connected to a laptop via Wi- 4. Process and analyse the sensory data col-
Fi and installed on the performer's body to measure lected to understand if significant differences
while the body was performing with the objects. Re- can be found in the classification and differen-
call that the main goal of this experiment was to clas- tiation between a "compelling-positive" expe-
sify users' response to corporeal stimuli, by analysing rience and a "not compelling-negative" one.

Figure 2
Subject performing
two variations of 1:1
scale setting "Eating
dinner".
Figure 3
Subject performing
two variations of 1:1
scale setting "Office
space".
RESULTS The data is organized according to Presence Ques-

Experiment 4 - De Humani Corporis Fabrica evaluated tionnaire's "Factors" and "Subscales" , as well as SAM's
the emotional experience of 1:1 scale architectural parameters of "Valence", "Activation" and "Control".
settings by analysing changes in the sensorial per- As in the previous three experiments, the final
ception of the user while performing a pre-set chore- values considered in the analysis of the PQ results
ography of actions in settings built for the same pur- were obtained by averaging the ratings assigned by
pose but with significantly different dimensions. The the subjects to each of the questions, according to a
results of this experiment were the answers to the 1-9 point scale. The final values used in the analysis of
Presence Questionnaire (Witmer and Singer 1995) the data collected through the SAM charts were ob-
and the SAM chart, where recall of experience and tained following the same principle, as the three pa-
believability of simulation are systematized. Such rameters of "valence", "activation" and "control" also
data depicts the subject's physiological response and were rated by the subjects using a similar 1-9 point
emotional activation and allows for the evaluation scale. After making these calculations, we obtained
of the setting's ability to condition the user's move- values that qualify each subject's individual experi-
ments and alter his emotional state and perception. ence with the "De Humani Corporis Fabrica" settings,

regarding the parameters of "presence", "emotional • 2 had their attention "very dedicated" during
response", "valence", "activation" and "control". From performance and 2 "dedicated";
the data collected, we can also qualify the experience • 3 found the performance sensually "very en-
of the group of subjects as a whole, by averaging the gaging" and 1 "engaging";
results for the same parameters. • 2 were "very involved" in the visual aspects of
The average value of the answers given by the the performance and 1 "involved", while 1 was
group of 4 subjects to the Presence Questionnaire "not involved";
(PQ) and the SAM chart shows that they have clas- • 2 were "involved" by the haptic aspects of the
sified the experiment "De Humani Corporis Fabrica" performance, 1 "very involved" and 1 "some-
as "compelling", "worthwhile", "moderately pleas- what involved";
ant", "very exciting" and "dominating". These adjec- • 2 were "involved" by the auditory aspects
tives relate respectively to the parameters of "pres- of the performance, while 2 were "not very
ence", "emotional response", "valence", "activation" much";
and "control". • 3 found the performance settings "very con-
Out of the group of 4 subjects which participated vincing" and 1 "convincing";
in the experiment: • 2 were "aware" of events around, while 2 were
"not very aware";
• 4 subjects classified the experience of the per- • 3 were "very aware" of the surrounding space
formance with the "De Humani Corporis Fab- and 1 "aware";
rica" settings as "compelling"; • 3 were "very able" to anticipate use of settings
• 3 subjects rated the emotional experience of just by looking at them and 1 "able";
the performance with the "De Humani Cor- • 2 were "very compelled" to move in the set-
poris Fabrica" settings as "worthwhile" and 1 tings and 2 "compelled";
as "very worthwhile"; • 3 felt that settings and objects "really stimu-
• 2 subjects rated the experience of the perfor- lated" their imagination for the performance
mance with the "De Humani Coporis Fabrica" and 1 just "felt" so;
settings as "very pleasant" and 2 as "pleasant"; • 3 could interact with settings and objects
• 2 rated the experience of the performance "very well" and 1 "well";
with the "De Humani Corporis Fabrica" set- • 3 could manipulate the surrounding environ-
tings as "very exciting" and 2 as "exciting"; ment "very well" and 1 "well";
• 2 rated the experience of the performance • 4 were "very involved in the experiment";
with the "De Humani Corporis Fabrica" set- • 2 were "very distracted" by the suit, mask, and
tings as "dominating" and 1 "very dominat- e-health platform, 1 was "distracted" and 1
ing", while 1 rated it as "not very dominating". "not very much";
Also, out of the group of 4 subjects: • 2 felt the camera didn't interfere with perfor-
mance "at all", 1 "not much", while 1 thought
• 3 could control emotional response to set- that it "interfered".
tings "well" and 1 "very well"; • 2 could concentrate "very well" on the perfor-
• 2 were "able" to control objects in performance and not on the objects and settings
mance and 1 "very able", while 1 was "not very and 2 just "well";
able"; • 4 thought that they had learned new tech-
• 4 found the experimental setting "very stimu- niques which expanded their skills as design-
lating" in imagining the performance to take ers;
place;

• 4 found the experiment a "very good" learn- the most important aspect of the experiment which
ing experience; had to do with the movement of the body while per-
• 2 "lost track of time" and 1 "almost com- forming with the objects, the kinaesthetic sense. It
pletely", while 1 "not at all." also suggests that they were involved by the sense
of vision and hearing. This confirms a high-level of
engagement and suggests that the experience was
DISCUSSION AND CONCLUSIONS immersive, although other results show that subjects
Experiment 4 - De Humani Corporis Fabrica evalu- were somewhat distracted by the suit, mask, and e-
ated the emotional experience of 1:1 scale architec- health platform. Therefore, results support Hypoth-
tural settings by analysing changes in the sensorial esis H2, which suggests that the feeling of "pres-
perception of the user while performing a pre-set ence" and emotional activation can be intentionally
choreography of actions in settings built for the same induced through the performance with analogical
purpose but with significantly different dimensions. models, in this case, architectural settings.
The experimental results support the main research Results also show that most subjects were in-
Hypothesis H1 - a user's emotional response to de- volved in experiment and lost track of time during the
sign objects as "compelled or not compelled", "posi- performance. This suggests that the feeling of "pres-
tive or negative", "aroused or not aroused" and "dom- ence" and arousal situations can be intentionally in-
inant or dominated" can be evaluated through ob- duced in real-space, but further research is necessary
jective measurements of emotion. The results col- to understand which specific design elements are re-
lected through the Presence Questionnaire and the sponsible for this. All subjects rated the experiment
SAM charts show that the majority of subjects ex- as a very good learning experience and found that
perienced a high level of "Presence", "Pleasure" and performance art techniques enhanced their creativ-
"Arousal". Most subjects describe their emotional re- ity and capacity to design. Therefore, results also sup-
sponse to the performance with the objects as "pos- port hypothesis H3 which suggests that the somatic
itive". Results reveal that most subjects could con- techniques of "performance art" and "emotional de-
trol the emotional response to the setting and ob- sign" are an effective strategy to develop corporeal
jects during the performance, but they experienced awareness and stimulate the creativity of students
a low level of control of the performance itself. This and designers.
indicates that the subjects probably didn't feel con- The amount of participants in the experiment
fident with the stability of their designs or perhaps was only four which according to the quantitative re-
their motor-skills got in the way and so they couldn't search tradition could not be understood and treated
perform at ease. Still, results show that the major- as quantitative material. Although a qualitative re-
ity of subjects didn't feel the camera interfered in the search approach would be possible to use in such
performance and felt able to anticipate the use of the a context and eventually providing interesting re-
experimental settings just by looking at them, as well sults, the same methods used in the experiment can
as compelled to move in them, and found them con- be applied to a larger experimental context. Also,
vincing and very stimulating in imagining the perfor- the used approach can be understood as part of the
mance before it took place. educational content of the course and as training
Most subjects reported to be very involved in the of research methods which incorporate innovative
experiment, found the performance sensually very methodologies in the teaching of design and sensi-
engaging, were very involved by visual aspects and tize the students to the emotional impact of design
involved by haptic and auditory aspects, while re- objects in the human body. The results can be seen,
maining aware of events around and the surrounding for that reason as suggestive, as this experiment was
space. This suggests that subjects were focused on

a pilot for future experiments which will involve more Damásio, A. 1999, The Feeling of what happens: Body and
subjects and means. For the same reason, the results Emotions in the Making of Consciousness, Harcourt
are presented in this paper as a list and not as graph- Books, Orlando
Eberhard, J.P. 2009, Brain Landscape. The Coexistence
ics since there would be one graphic per item ad-
of Neuroscience and Architecture, Oxford University
dressed, which would mean 28 graphics to describe Press, Oxford
each of the parameters taken in consideration. As the Gallagher, S. 2005, How the Body Shapes the Mind, Oxford
participants in the experiment were only four, and University Press, New York
the results therefore do not represent a statistical im- Goldberg, R. 1988, Performance Art, from Futurism to the
portance, the authors decided to present them in the Present, Thames&Hudson, London
Kim, M.J., Cho, M. E. and Kim, J.T. 2015, 'Measures of Emo-
form of a list. The experiment described in this pa-
tion in Interaction for Health Smart Home', IACSIT
per shows that PQ, SAM, and biometric technology International Journal of Engineering and Technology,
can be used to objectively discriminate arousal re- 7(4), pp. 10-12
sponses related to "positive" or "negative" emotions, Lang, P.J., Bradley, M.M. and Cuthbert, B.N. 1998, 'Emo-
from the neutral condition, when users experience tion, motivation, and anxiety: brain mechanisms
1:1 architecture settings. On-going research in the and psychophysiology', Biological psychiatry, 44(12),
pp. 1248-1263
fields of IT, psychology and marketing uses an es-
Mallgrave, H.F. 2010, The Architect Brain: Neuroscience,
tablished range of values that also were used in ex- Creativity and Architecture, John Wiley&Sons, West
periment. Results confirm that it was useful to in- Sussex
clude biometric machinery in the experiment, to ob- Vesalius, A. (1543), Garrison, D. and Hast, M. (ed. trans.)
serve how user emotions are triggered while experi- 2003, De Humani Corporis Fabrica, Northwestern
encing design objects, as the participants in the ex- University Evanston, IL USA
Zeizel, J. 2006, 'Inquiry by Design - Environmental / Be-
periment, in this case students of architecture, could
haviour / Neuroscience', in J., Zeizel (eds) 2006, Archi-
see in the computer screen the live feedback of the tecture, Interiors, Landscape, and Planning, W.W. Nor-
physical experience of performing in the design set- ton & Company, New York
tings. This feedback came directly from the analysis
of changes in skin conductance and cardiac acceler-
ations which are strongly correlated with emotional
arousal. Still, the machinery used posed problems,
namely of stability while collecting the results, there-
fore the data collected in the experiment is not pre-
sented in this paper. The aim in the experiment was
to observe peak and limit reactions that would pro-
vide for a wide scale of physiological measures. Nev-
ertheless, more work is necessary to establish solid
conclusions. In future experiments, it is planned to
integrate the use of EEG and to add to the experimen-
tal setting eye-tracking sensing technology to record
the position of the user and where s/he is looking at.
REFERENCES
Bradley, M.M. and Lang, P.J. 1994, 'Measuring emotion:
the self-assessment manikin and the semantic dif-
ferential', Behavior Therapy and Experimental Psychi-
atry, 25(1), pp. 49-59

SMART AND RESPONSIVE DESIGN
Responsive Systems: Foundations and Application
The importance of defining meta-systems and their methods
Goncalo Castro Henriques1

1
Universidade Federal Rio de Janeiro, LAMO, PROURB
1
gch@fau.ufrj.br
Responsive architecture is often considered as one that merely adapts to change.

This reflects its limited and still incipient application in architecture. Due to the
current resource crisis, systemic building management is essential. This article
argues that there are no established processes for creating and managing
responsive architecture. To establish a foundation in this area, it claims that it is
necessary to deepen knowledge about systems, computation, mathematics,
biology and robotics. Despite being a vast subject, it proposes a state of the art of
the systems, investigating how to operate them. A method for generating
responsive systems is tested and implemented in a practical case. Two methods of
adaptation are proposed and tested: static and dynamic adaptation. These
methods reinforce the point that responsive architecture can use not only active
mechanisms, but also passive methods embedded in its form as information. The
research concludes that information management is critical to define what is
designated in software engineering as architecture of the system. Thus, it suggests
that it is necessary to define meta-systems and to define their methods to support
the generation, fabrication, construction and operation of responsive systems.
Keywords: responsive systems, meta-systems, static adaptation, dynamic

adaptation, heuristics
INTRODUCTION building management after construction, which has

New technologies can enable better use of resources been less explored. Personal building(s) manage-
that are scarce and finite. In architecture, the new ment is still considered a luxury in the post-industrial
technological possibilities can allow a shorter, inte- society, available for special buildings and accessible
grated design process, enabling more efficient pro- only to few.
duction. However the most visible expression that Buildings use a significant amount of energy,
emerges from these new possibilities is geometric not only during their construction, but also for their
differentiation. These technologies are also used to maintenance. Precautions must be taken in build-
manage the design process and digital fabrication. ing energy efficiency due to global warming. One
But, seldom are technologies used to manage the of the precautions is the use of shading devices sup-
SMART AND RESPONSIVE DESIGN - Volume 1 - eCAADe 34 | 511

ported by simulation tools (Kirimtat et al, 2015). This Today's society faces serious environmental chal-
article claims that, although shading devices are im- lenges. Edward Mazria studied the relationship be-
portant, the use of simulation is not sufficient, as re- tween architecture, energy and the environment.
sponsive systems should embrace a wider range of Mazria (2003) accounted energy use in three basic
information. Good resource management can only sectors in the US: the architecture, transport and in-
be achieved with new technologies and data man- dustry. According to that study, the architectural sec-
agement. Integrated management of buildings is tor consumes 48% of the energy, while transport and
more than a desire, it is a necessity. The develop- industry consume respectively 27% and 25%. Fur-
ment of mathematics has allowed the formulation of thermore, the architectural sector generates 46% of
general laws described in the general systems theory. CO2 emissions annually, a figure that is already sig-
Although the system concept has long been consid- nificant and likely to increase. If this trend spreads
ered in architecture, it has benefited from the devel- to cities in general, then the architecture should re-
opment of the discipline that studies the control of flect this impact. This justifies the growing concern
systems, cybernetics. In cybernetics, it is indifferent with reducing energy consumption in architecture
whether a system is social, natural or synthetic (De and construction. For a more sustainable environ-
Landa 2011). To define a responsive building, it is ment, buildings should be more adapted to the envi-
necessary to revise the concept of system. Although ronment and use fewer resources. Digital processes
there are references described in the literature about and tools can be used to improve buildings' adapt-
responsive buildings, an inclusive state of the art is ability to the environment.
lacking. This state of the art intends to relate the Developments in biology, computation, archi-
concept of responsive building with the evolution of tecture and design enable the development of re-
mathematical principles that support these systems. sponsive buildings, which unlike traditional static
This paper argues that to define responsive buildings, can adapt dynamically to the environ-
buildings it is necessary to design and develop inclu- ment. This implies an ecological vision of popula-
sive systems. It investigates how to develop a holistic tions, giving emphasis to the group process and in-
system and how to manage complex processes with teraction, in a holistic vision. This can overcome a
a large amount of information. It is expected that the partial view centred on the objects as separate en-
analysis of the state of the art can be useful for the de- tities. To consider buildings in the city context new
velopment of support systems for responsive build- tools are necessary to organise and regulate the flow
ings. of information. The control of information systems
can support the development of responsive architec-
THEORETICAL-PRACTICAL METHODOL- ture. A new class of envelopes can be developed to
adjust and adapt to the environment according to cli-
OGY
matic variations, time of day and geographical loca-
This article will first highlight the importance of en-
tion, with the potential to Customize buildings in or-
ergy in buildings, referring to the studies available.
der to consume less energy.
Then it will investigate the mathematical origins of
the general systems theory and its development.
It will then address cybernetics, the discipline that RESPONSIVE SYSTEMS - STATE OF THE ART
deals with the systems control. The purpose is to in- After the industrial revolution, scientific and techni-
vestigate the relationship between the system con- cal knowledge became specialised. The necessity
cept and emergence of responsive buildings. Finally, to regain a holistic vision of knowledge emerged.
practical applications of these theories will be pre- The concept and the tools are presented in the Gen-
sented and discussed. eral Systems Theory. This theory includes multiple
512 | eCAADe 34 - SMART AND RESPONSIVE DESIGN - Volume 1

branches of knowledge, such as biology, sociology sponse to changing conditions. While this descrip-
or computation. Following the development of this tion successfully captures the essence of the subject,
theory by Von Bertalanffy, many researchers from dif- it does not provide a detailed understanding neces-
ferent fields have contributed to the development sary to build it". He argued that to do so is crucial
of responsive architecture, such as Pask, Alexander, to associate architectural knowledge, artificial intelli-
Weinstock, Negroponte, Menges, Sterk, Mitchell and gence, robotics and structural engineering. The lack
Achten, among others. of knowledge by most architects in these areas jus-
Von Bertalanffy developed the General Systems tified them presenting proposals for responsive ar-
Theory applicable to living systems, social or me- chitecture that are ambivalent, as they have difficulty
chanical (1950-68). This theory mathematically de- in overcoming an aesthetic that could be based on
fines adaptation mechanisms as feedback and equi- something more than events.
finality. Gordon Pask developed the systems of con- Sterk also argues that it is necessary to under-
trol theory (Cybernetics 1969), completing the Berta- stand responsive architecture on different scales. Re-
lanffy control mechanisms with goal-oriented mech- ferred to as the previous model, Negroponte is in-
anisms. If the machines were regarded as closed teresting conceptually, but computational models of
systems, with the development of cybernetics, they this theory are still scarce. It highlights the advances
could be considered as open systems if they were in the fields of artificial intelligence and robotics,
able to learn and adapt to the environment as other in defining high-level control systems (deliberative)
living systems. These definitions allowed a shift from and low-level (self-regulated). It proposes implemen-
closed, unidirectional systems to open multi-causal tation of responsive architecture based on a hybrid
systems. This also supports the creation of collabora- control system (Figure 2) combining high-level com-
tive environments between man and machine to en- putation (deliberative) with low-level computation
able a responsive architecture. (self-regulated). However, it can be argued that Pask
Christopher Alexander in the text "Systems Gen- proposes a goal-oriented approach that already con-
erating Systems" (1968), reflects on the relevance of siders low-level loops within deliberative high-level
the word "system" in architecture. It also reflects on loops (Figure 1). This linkage allows associating the
how to create holistic systems that guarantee forms deliberative decisions with self-regulation. The low-
synchronised with their environment. Alexander ar- level responses can be given via sensors, and when
gues that a system is only considered as such if it a certain value is reached, a higher-level order is exe-
has a generative system governing the relationship cuted. In this threaded process, the feedback loop is
of parts to the whole, and the whole to the parts. It essential.
argues for the importance of generative systems for Sterk aims to combine the deliberative con-
the definition of holistic systems. trol developed by Yona Friedman (1971) with the
Nicholas Negroponte has been a pioneer of re- self-regulated control proposed by Chuck Eastman
sponsive architecture since the 70s, foreseeing how (1972). While the deliberative control requires high-
digital technologies could improve the documen- level computation (i.e. use of a coded language,
tation systems and design, generative systems and and the user knows the system's operation), the self-
support intelligent spaces with built-in computers. regulated control does not require the use of a coded
Tristan-d'Estree Sterk argues that today, this theoret- language and may be controlled by machine code. In
ical understanding is not enough to implement re- self-regulated control, simple reflective procedures
sponsive architecture. According to him "The respon- are used, and the system state is determined by the
sive architecture is often defined as a type of archi- interaction among the parts. Mike Weinstock refers
tecture that has the ability to change its shape in re- to the importance of mathematics in the theory of

Figure 1
Simple feedback
mechanism
(Adapted from:
Bertalanffy, 1969)
and goal-oriented
mechanism
(Adapted from:
Pask, 1969).
Figure 2
Hybrid Control
Model for a
responsive
architecture,
(adapted from
Sterk, 2006). It
combines
high-level
processes
(deliberative) and
emergence. Weinstock proposes application of this latest reference reinforces the need mentioned by
low (self-regulated).
theory (2004) to architecture to support a new mor- Sterk that architects need more extensive knowledge
From left to right: 1
phogenesis of form of various areas to define such systems.
Components of a
Henry Achten (2014) identifies an amount of
discrete model of
terms related to responsive buildings, such as: build- Generating responsive systems
architecture
ing automation system, smart home, sentient build- To support a responsive architecture, it is necessary
including the wants
ing, adaptive building, dynamic building, interac- to create systems that manage systems, i.e. meta-
and needs (n/w),
tive building, kinetic architecture, intelligent build- systems with the characteristics to be introduced into
with spaces (sp)
ing and portable building. Achten provides a def- these responsive systems. Pask, in the cybernetics
and structures (st);
inition for each of these terms, concluding there is theory, states that architects have to generate sys-
2 High-level user
no possibility of unanimous classification for each tems that generate buildings, rather than just de-
decisions (n/ w),
of these definitions, because each one is related to sign buildings. Alexander, analyzing systems archi-
space (SP) structure
a specific technological application. He states that tecture, agrees - arguing that, to generate systems, it
(ts) and , within
this may be related to the recent application of these is necessary to design generative systems that enable
each, there are
technologies, which have not been absorbed into the them. Authors from different fields of knowledge re-
low-level decisions;
common language. Achten seeks to characterise the fer to the need to define meta-systems and their gen-
3 Responsive
interaction of responsive systems with users. He ad- erative principles. Sterk draws attention to the point
architecture
vocates the use of recent developments in artificial that, to control these meta-systems, understanding
applied to network
intelligence, with the use of multi-agent networks ca- of different fields is needed. However, despite the
buildings, repeating
pable of supporting analysis, design and develop- model proposed by Sterk, it is not yet clear how this
the above
ment of complex systems in all project phases. This could be implemented.
combination, but
on a larger scale.

De Landa states that a system has capabilities that be implemented in practice.
arise from the interaction of the parts that exceed the
sum of the individual properties. He states that, for Meta-System
these capabilities to emerge, it is necessary to define To generate a responsive system, a meta-system is
the procedures and variables of the system. He also proposed that guides the search in the space of solu-
states that the restrictions applied to the variables in- tions. The meta-system is organised at different lev-
creasingly define the systems. He refers to algorithms els of abstraction. These levels have different mathe-
as the essence of software, comparing mathematics matical definitions, reducing the search space of the
metaphorically to a type of software. He highlights problem. This method was developed with a practi-
the definition of gravity proposed by Richard Feyn- cal application - the Tetra Script responsive system -
man in three stages: 1 - force and acceleration, 2 - that mediates interior and exterior conditions to as-
force fields, 3 - singularities and solution space (mini- sure daylight quality and quantity.
mum, maximum and variational characteristics). This
Figure 3 change is related to the move of quantitative math-
Proposal to ematics for qualitative mathematics, from values to
formalise a parameters. This change is also related to the move
responsive from Euclidean geometry and the metric systems to
meta-system the topological geometry of local spaces. This math-
(Henriques 2012), ematics allows setting associative behaviour among
defining the the parts through parameters that can be encoded
general search algorithmically.
space for the Examples of natural systems and biology have
particular been incorporated into computation to define open The meta-system proposed is organised at the fol-
(deliberative systems, which are necessary to establish a founda- lowing levels: General System; Specific System and
hierarchy), and the tion for responsive systems. Open systems have the Practical Case. The dynamics of the meta-system is
particular for the ability to retrofeed through feedback loop mecha- multi-causal, from the general to the particular, and
general (distributed nisms and to conform, reconfiguring the relationship from the particular to the general. At each of these
logic, emergence). among the parts. Living systems possess a hierarchy levels variables and restrictions are defined. Going
with different levels of definition, which allows con- from the general to the particular, the variables are
trol of complex systems, such as the human body, progressively less abstract and have more restric-
skeletal system, muscular system, tissues and cells. tions: some shift from free variables to dependent
The relationship among systems is not only hierarchi- variables. However, the low-level variables (practical
cal, but also emergent, from the whole to the parts, case) can also influence the Specific and the General
and from the parts to the whole. Systems.
Thus, the meta-system has a hierarchical nature
METHOD FOR A RESPONSIVE ARCHITEC- (the whole to the part), and the emergent one (the
TURE part to the whole). While the hierarchical nature is
This study intends to contribute to the current state associated with deliberative decisions, the emergent
of the art in the field by relating disperse contribu- nature is more associated with behaviour distributed
tions and by presenting a research about how to by self-regulation. The deliberative and distributed
define, support and generate a responsive system. nature of the system will be more explicit when ap-
For this purpose, there will be a description of the plied. The keyword for defining meta-system, as in
method used for generating a responsive system to natural systems, is feedback loop.

A Practical Case, Tetra Script system tributed network of building membranes.
The Tetra Script responsive system proposes a mem- As initial input, the specific system has a contin-
brane to mediate interior and exterior space. It can be uous NURBS surface, and its local topological space.
defined as a set of components that control internal The surface is populated with developable compo-
daylight according to external factors (geographical nents to assure change of the interior conditions.
location, year, month, day, hour, sky type) and inter- These variables are parametrically linked through an
nal demands (functional requirements of illuminance algorithm. When the specific system is bounded, it
and visual contrast, daylight types and light-shadow can be instantiated in a family of solutions (Figure 4).
patterns). Responsive derives from the Latin "respon- This family of objects, used in an oriented program-
sīvu>> designating one that responds quickly and ming language, is designed as a class for which ac-
appropriately to a situation. The proposed system re- tions and methods are defined.
sponds in real time, which justifies the use of such an The practical case can be instantiated in a specific
adjective. The development of the system took place building. Tetra Script pavilion is an instantiation of
in a PhD research. This article intends to reflect on the specific case, attributing values to the variables.
this previous research to draw useful conclusions for To find these values, a static adaptation process is
the development other responsive systems. used. To manage the goals of the system after con-
This project required conceiving a meta-system struction, dynamic adaptation is used, which will be
to define how to generate and design a daylight con- detailed further below. After setting the geograph-
trol system. The meta-system is defined by succes- ical location and the input surface, the number and
sive approximation levels, starting from a generic so- characteristics of the components are defined.
lution with a large search space. The space of solu-
tion is constrained by progressive definition of the Figure 4
variables, which become dependent in a less abstract Tetra Script System
definition. The meta-system has three levels: general (Henriques 2013):
system, specific system and practical case. The vari- instances of the
ables and factors used are progressively constrained same specific case.
to find a discretization or solution to a particular sit- As a specific case, Tetra Script is defined with a quad-
uation, in a pavilion that was constructed, and where rangular pyramidal component with a base, with mo-
the automation was tested. bile faces, easily manufacturable. Tetra Script pavil-
ion is an instantiation of a particular system, restrict-
Variables, domains and restrictions ing the variables. The configuration process can find
The system organisation will be detailed together the best relationship among the variables to satisfy
with the variables, domain and restrictions used. certain conditions, defining a progressively limited
General System variables are more abstract; setting search space of solutions. An emergent way to set
the search space and focus: to develop architec- the system would be to first choose a type of mate-
ture with a protective membrane that establishes a rial, and then define the components, the base sur-
holistic relationship between the outer and the inner face and, finally, the general system objectives. But,
space, which is customisable through digital tech- an emergent solution can also be achieved through
nologies. It seeks to promote the explicit use of algo- interaction among the components using dynamic
rithms by setting parameters, discarding the implan- adaptation.
tation topography and context. This general system The meta-system levels are divided into sub-
can be used to define other objectives, such as the levels. In the practical case developed, after set-
organisation of interior spaces or to establish a dis- ting the values of the general variables, a specific
system was set containing sub-level groups of vari-

ables that, for space limitations, are summarised as tems. As the author presented in previous publica-
follows. The interaction between each group of tions, these processes find parallels in living systems,
variables determines a solution for each individual particularly in adaptation through natural selection
case. The group of variables is defined thus: exter- and behavioural adaptation.
nal variables of the object (such as geographic lo- In static adaptation, the relation among variables
cation, date, sky type and respective domains, con- is established progressively discretization values. As
straints and dependencies), the specific variables of mentioned, the general system level defines goals to
form (surface type, dimensions, component type and be enabled by the use of digital processes. To find
number), variables of the form in the practical case a solution using the system, an asynchronous pro-
(surface area, parametric space, domain, number of cess was conducted using CAD-CAE-CAM-AUT digital
pyramids, height and aperture value), and internal processes. Digital processes were used to generate
variables or functional requirements (illuminance, vi- and manipulate surfaces using algorithms, to man-
sual comfort, thermal comfort and space heuristics). ufacture components, to test component types and
To determine certain conditions it is necessary to find groups, to simulate and assess environmental perfor-
algorithms to control the parts and set the relations mance of the solutions generated, and to automate
among variables. and develop implementation strategies for the com-
ponents. A diagram is proposed to describe the inter-
System methods: static and dynamic adap- action among the digital tools and processes found
tation in previous publications (Henriques et al. 2012). The
This research proposed two methods to control the search was made with a preference for using explicit
meta-system adaptations, which are referred to as algorithms, but the search process for solutions can
static and dynamic adaptation. Static adaptation be further automated and improved.
controls the generation of geometry until it is baked, The dynamic adaptation controls the relation-
manufactured and constructed. The dynamic adap- ship among the parts to ensure the systems' desired
tation controls the relationship among the parts af- performance. After construction of the Tetra Script
ter the system is constructed and operating. It allows pavilion, the variable that can be controlled is the de-
a real-time adaptation of the system in the practical gree of opening of skylight panels, which thus deter-
case developed. This methodology can be used to in- mines the formal setting of the pavilion. There is a
stantiate different solutions and to develop other sys- multitude of possible combinations of apertures for
Table 1
Definition of
meta-system levels
(Henriques 2013), in
the Tetra Script
case.

controlling the amount and type of daylight. The rations, and, to assess performance, a daylight vali-
dynamic adaptation allows finding aperture settings dated software (Radiance) is used. Simulations were
that meet light requirements, including glare and vi- performed considering all the variables described
sual contrast. A more detailed description of the pro- above. Values were assigned, and system perfor-
cess can be found in previous publications. mance tested through illuminance program evalua-
Heuristics are used to find space solutions with tion setting the Tetra Script performance database.
certain features that meet the desired requirements. An interpolation method was developed in order
Heuristics are a mathematical rationale used to iden- to predict the illuminance behaviour at other dates,
tify good solutions in a short time, rather than opti- times and in different sky conditions considered in
mal solutions that do not always exist and that take the studies. This method is used to evaluate and
longer to be identified. The set of heuristics to de- choose the most suitable aperture setting for the
scribe behaviour patterns is based on simple rules required indoor conditions. A visual interface was
and typify possible solutions, thereby reducing the developed to control the internal conditions of the
search space. The optimisation research could pro- pavilion in real time by modifying the opening of sky-
vide better solutions, but results would take a long light panels.
time to converge, since the search space of all pos-
sible configurations to undergo analysis is vast. The Shared Control Interface man / machine
proposed heuristic search method includes qualita- Algorithm control, feedback and adaptation. The
tive and quantitative properties of natural daylight. user can choose what factors he values most to cus-
The 5 heuristics used are described graphically in Fig- tomise the space, defining the quantity and quality
ure 5. of light inside the pavilion through an interactive pro-
cess. The customisation was precisely one of the ob-
jectives of the research performed. To select a partic- Figure 5
ular factor, the user is informed of the range of pos- Heuristics rendered
sibilities of the system to achieve the same amount preview (above)
of light with different qualities. This choice may be and schematic
performed in real time using the proposed interface. representation
A control system is proposed to control the sky- (below). From left
In the practical case, the interaction among groups light apertures (Figure 6). The system uses different to right : H0
of variables is considered: the external variables (ex- types of inputs: from the user (manual) and from the uniform aperture,
ternal to the object), the specific variables of form, computer (automated). The user can choose the ac- H1 maximum direct
the practical case variables and the internal vari- tivity by defining the minimal illuminance required. light, H2 maximum
ables of space or functional requirements. For each This information is transmitted to the algorithm that diffuse light, H3
group, it is necessary to define the "sub-variables" then computes all the possible configurations. The incident sunlight,
domain and bounds. In other words, it is necessary computer has instructions to obtain the time and lo- and H4 Hybrid low
to build a system that approaches a mathematical cation information from the computer that is running incident + diffuse
model establishing relationships and boundary con- the program, and both inputs - required illuminance light. (Henriques et
ditions. Heuristics are defined parametrically, testing and time-weather conditions - are fed into the heuris- al. 2012).
the boundary conditions throughout the simulation tic search calculation, where values are interpolated.
to find the system space of possible solutions. To find Then, the choice of the most adequate solution can
solutions using this system, it is necessary to establish be made automatically or manually. But, the signifi-
the visual comfort requirements and external refer- cant feature of this scheme is to add a feedback loop
ence conditions. using light sensors to update the database. This pro-
The parametric system is used to find configu-

Figure 6
Shared Control
System man /
machine Tetra
Script pavilion
(adapted from:
Henriques 2012).
cess would enable refinement of the expected values RESULTS AND DISCUSSION
for each situation. Although this process is not yet This paper analyzed and synthesised contributions
implemented, it is expected that its use would allow seeking operational concepts for the development
the system to correct values and, in a way, "learn with of responsive architecture. A previous applied case
experience." study was analysed looking for a general method to
The meta-system has a dual nature, which is de- design responsive systems. Although the proposed
liberative and inductive, thereby creating space for task is complex and wide, the research aims to con-
possibilities ("state spaces") enabling solutions that tribute to this discussion. What becomes evident
are beyond the property of each part, but result from during this attempt is the importance of computa-
the interaction of the parts. Thus, both the meta- tion for the development and implementation of sys-
system, at different levels, as well as control algo- tems.
rithms, seek to enhance the immanent, the virtual, The creation of responsive systems requires
that only when explicit can become knowable, dis- meta-systems and the use of computation to con-
covering new uncharted possibilities. struct a formal model that can approach a proba-
bilistic model used in physics. While classical physics

presents a model in which each object interacts Henriques, G.C., Duarte, J.P. and Leal, V. 2012, 'Strategies
alone, the probabilistic physics tends instead to value to control daylight in a responsive skylights system',
the interaction among the parts combined in a set Automation in Construction, 28, pp. 91-105
Kirimtat, A et al 2016, 'Review of simulation modeling for
as a model. Probabilistic physics also considers sev-
shading devices in building', Renewable and Sustain-
eral causes simultaneously, ideally multi-causality able Energy Reviews, 53, pp. 22-49
that is absent from deterministic models of restricted De Landa, M 2011, 'Real Virtuality', in Menges, A and
causality (or single causality). The result of multi- Ahlquist, S (eds) 2011, Computational Design Think-
causal interaction is predictable within certain pa- ing, John Willey & Sons Ltd
rameters, but is not determinable a priori. To enter Mazria, E 2003, 'It’s the Architecture, Stupid!', maga-
zine Solar Today, American Solar Energy Society, May
this discussion, architects need to know about organ-
/June, pp. 48-51
ising information or meta-systems and their meth- Menges, A and Ahlquist, S 2011, Computational Design
ods. Otherwise, they can only access systems as a Thinking, John Willey & Sons Ltd, Chichester, United
black box, entering information as an input receiving Kingdom
only the output. In this framework, the simple use Negroponte, N 1975, Soft Architecture Machines, MIT
of the computer - or computerisation (as defined by Press, Cambridge, MA
Pask, G 1969, 'The Architectural Relevance of Cybernet-
Terzidis) is not enough, nor is it sufficient to define a
ics', Architectural Design, September issue No. 7/6, ,
system as a model of representation of the reality. Al- John Willey & Sons, London, UK, pp. 1-9
though the computer is still necessary as a tool, the Sterk, T 2005, 'Building upon Negroponte: a hybridized
use of computation to define meta-systems, defining model of control suitable for responsive architec-
methods and actions used to design and operate sys- ture', Journal Automation in Construction, 14, pp.
tems, becomes relevant. 225-32
ACKNOWLEDGEMENTS
My gratitude goes to everyone who contributed to
the development of the Tetra Script system. With-
out their support this system could not exist. Part of
this research was funded by grant no BD/39034/2007,
awarded from Fundação Ciência e Tecnologia (FCT),
Portugal. The author acknowledges Tristan d'Estrée
Sterk kind permission to cite and republish his work.
REFERENCES
Achten, H 2014 'One and Many: An Agent Perspective
on Interactive Architecture', Proceedings 34th Annual
Conference of the Association for Computer Aided De-
sign in Architecture (ACADIA), Los angeles, pp. 479-
486
Alexander, C 1968, 'System Generating #Systems', in
Alexander, C (eds) 1968, Original Architectural De-
sign, December Issue, No. 7/6, John Willey & Sons Ltd,
London
Bertalanffy, LV 1969, General System Theory: Founda-
tions, Development, Applications, George Braziller,
New York

Material Agency and Physical Boundaries
Christina Doumpioti1
1
Architectural Association | OCEAN Design Research Association
1
cdoumpioti@gmail.com
The research builds on the relationship between matter and energy and the idea
of boundaries as sentient interfaces capable of affecting our bodily experience
and perception of space due to their inherent physical attributes. Two key issues
addressed are firstly, the revisiting of the architectural boundary as a
thermodynamic zone and secondly, the identification of material behaviour in
relation to energy stimuli. It is argued that the transient behaviour of materials
can offer an instrumental reconsideration on how architecture establishes spatial
articulation through boundaries and this is demonstrated through a design-led
project.
Keywords: thermal field, responsive materials, passively active materials, heat

transfer, thermodynamic, matter and energy
INTRODUCTION: MATERIALITY IN ARCHI- chitecture (Thomas 2007). In this binary relation mat-
TECTURE ter becomes a passive agent on which form is being
As architects one of our basic tools for organising imposed.
space is by constructing and distributing boundaries, Since the Renaissance and most recently in Mod-
defined by facades, doors, walls, and floors among ernism, architectural elements were conceived as
others. Those boundaries control, converge and dis- pure, texture less and geometrically recognisable ob-
seminate activities within an enclosed defined space jects. This perception has not changed a lot with the
that separates the inside from the outside. On the advent of postmodernism and the current more ex-
other hand, the boundary for physicists is a dynamic pressional architectural formations.
element, a transitional zone between matter and en- The introduction and utilisation of technology -
ergy, which remains changeable over time (Adding- such as the heating and air-conditioning systems- re-
ton and Schodek 2005). sulted in a consequent separation between building
According to Michelle Addington (2010), geom- performance and building morphology while ampli-
etry is privileged in architecture due to "an a priori be- fying architectural "discreteness" (Hensel 2013).
lief that perception is rooted in geometry". As a con- A primary assumption that this paper is mak-
sequence of that, the materiality of architecture being is that any construct is not only the visible static
comes an "aesthetic artefact" rather than an "instru- form, but also the invisible heterogeneous space that
mental" agent intertwined with energy qualities. it conveys through matter-energy relationship. The
Indeed, both the concept of Aristotelian hylo- choice of materials in relation to the environment be-
morphism -morphing of matter- and Plato's theory comes an important task for the creation and mod-
on the purity of form are long-standing believes in ar- ulation of spatial characteristics informed by differ-

ences of light, temperature, sound and humidity. sible with natural ones.
They can be distinguished in two broad cate-
RESPONSIVE ARCHITECTURE gories: 1) those that change their properties in re-
In the current era, human, technology and nature sponse to external stimuli, and 2) those that pro-
are fused together by exchanging information. The vide energy transformation functions. In the fist cat-
pureness of each one is not discernible. The post- egory, materials have intrinsic response to specific
human approach extends the notion of the body to stimuli, while in the second the response can be com-
cyborg, the notion of nature to techno-nature and putationally controlled or enhanced (Addington and
the notion of materiality to meta-materiality. Schodek 2005). Additionally, 3) a third category in-
This is similarly evident in architecture, where the cludes passive materials that without themselves un-
use of smart materials, ubiquitous systems and tech- dergoing change, they display different effects under
nological machines, "challenges the long-standing altered environmental conditions, i.e. dichroic films
conception of the building as an object autonomous and fibre optics.
from its environment and governed by disciplinary
interiority" (Harisson 2013). Attributes of responsive materials
The desire for a responsive architecture capa- With the aim of familiarising with their behaviour,
ble of transforming, interacting and communicating limitation and potentials, a series of testing was initi-
with the environment and its users is evident in differ- ated on readily obtainable responsive materials (Fig-
ent periods of time within architecture, influenced by ure 1).
systems theory and cybernetics. The architecture of This change from one state to another can re-
that period -mainly during the 60s- has been prone to veal the presence and concentration of various sub-
new materialism, which acknowledged the "diversity stances which are not otherwise perceptible. For
and untidiness of the world as well as the social and example photochromic materials illustrate change
psychological experience" (Sadler 2005). Architects of light intensities, thermochromic demonstrate
have designed and prototyped the essence of an al- change in temperature gradients, hydrogels can ex-
tering architecture by integrating electronics, com- pand in response to humidity levels or even to pol-
puters and circuits in architectural components. lution levels, and shape memory polymers change
Through the investigations of the past and the shape in response to temperature or electricity.
present it is often evident that material inertia has These connections can be customised according
caused a problem in realising dynamic, living archi- to different needs and wills. Especially nowadays that
tecture. Advancement in material science and engi- data sensor networks are proliferating, the connec-
neering paves the path towards conceptualising ar- tion between materiality and urban environmental
chitectural projects, which are open to communica- data is becoming more crucial. Responsive materials
tion with their physical environment. thus, can acquire a profound role on raising aware-
ness on environmental conditions by visualising in-
visible environmental agents.
RESPONSIVE MATERIALITY Furthermore, this change of state can become
Materials with computational logics, referred to as a mediator of passive microclimatic adaptation. Ex-
smart, responsive, or active, are composite materi- amples include the altering optical characteristics
als where bits (computation) and atoms (material- of thermochromic materials which not only exhibit
ity) meet (Ishii 2012) and which can easily change temperature differences, but also have the potential
between two or more states. They are mainly en- of enhancing thermal regulation due to the absorp-
gineered materials whose properties are defined in tion or reflectivity of particular colours (Figure 2). For
labs, expanding the range of interaction that is pos-

Figure 1
Small-scale models
of different
responsive
materials such as
photochromic that
change with light
intensity,
thermochromic
applied on wood
and fabric,
phosphorescent,
shape memory
polymers, hydrogel
in latex and liquid
crystal
thermochromic.
Figure 2 Similar propositions have been made in the past but

Models based on have not yet been developed or applied further. The
thermochromic ink Cloudgel, a thermo-chemical device introduced by
mixed with plaster Day Chahroudi is based on chemical adaptation on a
(a,b,c) or silicon (d). molecular level. The main material, sandwiched be-
The colour changes tween two glass panels, changed from transparent
either due to when temperature was low to white when the tem-
directed hot air perature increased and it described "an autonomous
(a,b,d) or warm solar powered temperature control system" (Davies
water passing 1981).
inside model (c). Ultimately, a remarkable feature that responsive
materials have is that they behave similar to biolog-
ical systems: they are inseparable from the energies
they convey, the energies that change their state and
the energies they absorb and emit. They eventually,
configure autopoietic systems, which are open to in-
instance white is the best choice of colour for an ex- formation, while their organisation remains closed,
terior wall, to minimise heat gain in the summer, as and they do so through their one material make-up
it is a poor absorber. If the surface can turn black at and without being comprised by discrete mechanical
night it should radiate most of the energy absorbed elements (Doumpioti et al. 2010).
since black is the ideal emitter of long wave thermal
radiation, ε=1. This is applicable to buildings of arid
Material limitations and potentialities
At their current form however, available respon-
climates with large diurnal temperature variation.
sive materials exhibit the cause-and-effect behaviour

they were designed to have. They form determi- tecture, and the sounds, smells, temperatures, radio
nate systems whose transformation is closed, since waves are the ephemeral elements that define the
there is no new element being generated and single- software (Haque 2004).
valued since they convert only to a single state All architectural elements are dynamic in the way
(Doumpioti 2011). their material make up mediates energies and al-
Although they comprise holistic systems with in- though they do not move or change position, they
tegrated sensing and actuating capabilities they can- are concurrently active since their behaviour results
not process information and consequently they can- "in the creation of qualities the world lacks... seem-
not gain cognition. Their response is autonomous ingly at rest, they are secretly at work" (Leatherbarrow
and independent of user's will. By missing the pro- 2009).
cessor that a mechanical system has, smart materials
cannot exhibit multiple and versatile behaviours and Vernacular architecture and the distinc-
that limits their capacity for connectivity. tion between logical, physical and sensorial
The establishment of multidisciplinary teams boundaries
that bring together designers, with engineers and From simple systems to more elaborated ones, ver-
material scientists may give new directions to what nacular architectures around the world serve as pro-
the material can do, and to attributes linked to found examples on the relationship between (pas-
projected interaction at a larger scale. Recent re- sive) materials and climatic energies. Their geometry,
search, which takes inspiration from nature to de- materials, openings and orientation are strategically
velop programmable 3d printed multi-materials with articulated for environmental mediation.
self-transformational properties, is very promising to- One such example, the Ondol, developed at
wards this end (Correa et al. 2015). about 1,000 BCE was a radiant heating floor for the
Moreover, open source information on the topic traditional Korean house, the Hanok. The hot air
with DIY processes and resources, is a means to open originating from the kitchen fireplace placed at the
up this field of research to a broader community exterior of the house was channelled through cav-
(Kretzer and Hovestadt 2015). ities to the under floor area and finally it was ex-
hausted through a chimney. Cavities were shaped
Passively Active Materials and Physical by either clay or bricks and were covered by large
Boundaries slabs of stone. The clay helped in sealing the combus-
While the initial aim of the research focused on the tion gases from penetrating the living space, while
potential utilisation of smart materials in architec- the stone slabs were becoming warmer by absorbing
ture, their transient behaviour has furthermore con- heat and radiating it above (Figure 3).
tributed to new insights on materiality not necessar- Besides its functionality as a thermal system, the
ily bound to dynamic response but to the capabili- Ondol stimulated the creation of social patterns with
ties of all materials to be active even when no change the "eldest occupying the warmest zones of the floor"
is noticeable. Any material construct is not only the (Moe 2010).
visible static form, but also the invisible heteroge- The under floor heating system has been utilised
neous space that it conveys through matter-energy extensively in vernacular architecture spanning from
relationship. the Roman hypocaust, to the Chinese Kang, and the
In relation to this, architect Usman Haque makes Spanish Floria, to name a few. Currently, the tech-
a distinction between the hardware and software of nology is becoming popular again with the develop-
architecture, where the walls, roofs and floors are ment of under floor heating system based on water
the static elements that define the hardware of archi- or electricity. What distinguishes the vernacular sys-

Figure 3
From left to right: i)
Perspective section
of the Hanok house
and the Ondol
system underneath,
ii) fluid dynamics
analysis of heat
flow, and iii)
thermal analysis.
The hot air tem though, is the synergetic relationship between the users and the existing microclimate by altering
originating from different elements in order to achieve adequate inte- the temperature differences according to their pref-
the outdoor rior thermal quality. All, structure, materials, geom- erence and creating emergent climatic patterns.
fireplace is etry, and other design decisions are interrelated as- While climatic control and regulation is most of-
channelled through pects of the overall system. ten associated with building interiors, the project as-
cavities underneath The boundaries described by vernacular archi- pires to its extension to exterior environments for the
the living area. tecture are physical boundaries as opposed to logi- amplification of public activities and social interac-
cal or "discrete" ones (Hensel 2013). While the logical tion.
boundary is an entity of visual demarcation and spa- Although, as described earlier, heat transfer and
tial articulation delineated in static formations, the thermal modulation played an important role in the
physical boundary describes a zone of fluctuations past, there are only scarce examples in contemporary
that emanate from the interaction of the material sur- architecture. Two profound representatives of this
face and the energies that surround it. In the first in- field are Sean Lally (2009) and Philippe Rahm, who
stance, form is untainted by matter and matter is sub- foster through their ideas new relationships between
ordinated to form. In the second, matter and energy urban space and climate and propose through their
animate architecture and become perceptible from projects, activities -augmented by heat- that might
all our senses. otherwise be assumed to need a conditioned interior.
CASE STUDY: THERMAL FIELD, PASSIVE Interface Components

The interface of the thermal field consists of a dis-
AND ACTIVE MATERIAL ENERGY TRANS-
tribution of small heat sources (peltier units) acti-
FORMATIONS vated by microcontrollers (arduinos). These devices
The thermal field project is a study on radiant sur- can cool down or heat up depending on the direc-
faces, similar to the Korean Ondol described above. tion of the current running through them. When the
The research initiated by experimenting with ther- thermoelectric units are activated their heat is trans-
mochromic materials and evolved into the idea of an mitted through a highly conductive material (copper
interactive version of a radiant thermal surface com- sheets) to the upper ceramic surface, which is treated
prised by passive and responsive materials by consid- with thermochromic ink (Figure 4).
ering their distinct yet complementary attributes. The connection between matter and energy
Similar to Raynar Banham's (1984) renowned alters according to the local materiality, ranging
parable of the fire as a source of spatial differentia- from energy transfer (thermoelectric units and heat
tion, the project aims at the creation of a thermo- conductive tiles) to phenotypical change (ther-
dynamic field as a space of social interaction in ur- mochromic materials) and to energy exchange
ban settings. It aims at acting as a mediator between

(through conduction, convection, radiation) be- is always indicated visually by the emerging patterns
tween the surface, the environment and the users. caused by the thermochromic ink.
While at this stage the microcontrollers are pre- The physical properties of the upper stratum are
programmed, the aim is for the system to be locally crucial in relation to the thermodynamic behaviour
defined by the users who, through their devices, can of the system. Their behaviour can be programmed
select the range of temperature transferred to the in their material properties and by altering two main
tiles and consequently conducted to their own body. attributes: heat conductivity and specific heat.
While thermal conductivity refers to the ability of
the material to transfer heat, the specific heat prop- Figure 4
erty indicates how much thermal energy a material Components of the
can hold in its molecular structure for a given mass. thermal field
Different combinations of materials where tested interface: peltier
and placed in relation to their role within the system. unit, copper and
The peltier units were positioned within wooden thermochromic
frameworks, and star-shaped copper sheets were ceramic tile.
placed on top, as copper is highly conductive. Tiles
made of different material combinations then ab-
sorbed the heat transferred by the copper sheets.
The tiles were made by a) liquid crystal sheet b) plas-
ter, c) fibre composite plaster or, d) plaster with Phase
Changing Material (PCM), testing their heat transfer
behaviour. The latter is a material with unique phase
changing properties as it can change from solid to liq-
uid and vice versa according to temperature differ-
ences, and acts as a temperature buffer.
The material combinations and consequently
the thermodynamic behaviour of the overall surface
can thus, vary extensively (Figure 6).
Figure 5
Contribution and Further Development of Patterns emerging
Thermal Field due to the
Within a conceptual frame, the dual aim of the thermochromic
project is firstly to stimulate an interrogation of the paint on the tiles
purely formal perception of the architectural bound- and the internal
ary to a dynamic and sentient one and contribute to difference of
a field of knowledge that sees architecture beyond its material
static limitations towards one that is energy-related. conductivity.
Thermal Behaviour In this sense the project aspires to combine the atmo-
Due to its heat-sensitive qualities the interface be- spheric with the performative qualities of the matter-
comes a data visualisation apparatus, revealing infor- energy relationship.
mation and affecting modes of occupancy (Figure 5). Further development will focus on the following
The thermal behaviour of the interface -how fast aspects: the first is the connectivity between user
it gets heated, for how long and on which location- and interface and furthermore the heat transfer at-

tributes of the surface in relation to its materiality and envision a future reality where materials play a key
geometry. The latter is an investigation on how ge- role in defining highly dynamic and responsive envi-
ometry and spatial qualities affect heat transfer and ronments. Materials that grow and self-organise by
distribution. Finally, an additional aim is to rethink having integrated processing capacities may lead to
the current energy consumption of the interface with unseen possibilities on spatiality by overcoming the
the utilisation of distributed energy harvesting de- static and pre-determined qualities of the existing.
vices capable of converting ambient energy to elec- The second remark is that despite the utilisation
tricity. of responsive materials, architecture would profit by
reconsidering the attributes of traditional materials
Figure 6 and integrate their 'passively active' behaviour. The
Some variables that make up of materials at a micro-scale as well as their
affect the interface’s organisation and assembly at a larger scale can yield
performance, from highly heterogeneous thermal or acoustic and lumi-
top to bottom: a) nous spatial environments. Likewise the notion of
Difference of the boundary can shift to something that is dynamic
internal material and closer to the physicists' idea of an ever-changing
conductivity. The one.
star-shaped copper
transmits heat. b)
Difference of
REFERENCES
Addington, M and Schodeck, D 2005, Smart materials
external material and Technologies: for the architecture and design pro-
conductivity and fessions, Architectural Press, Oxford
heat capacity: Banham, R 1984, The architecture of the well-tempered en-
liquid crystal vironment, University of Chicago, Chicago
surface vs plaster, c) Correa, D, Papadopoulou, A, Guberan, C, Jhaveri, N, Re-
difference of ichert, S, Menges, A and Tibbits, S 2015, '3D-Printed
Wood: Programming Hygroscopic Material Trans-
internal formations', 3D Printing and Additive Manufacturing,
temperature from 2(3), pp. 106-116
thermoelectric Davies, M 1981, 'A wall for all seasons', RIBA, 88(2), pp.
device. CONCLUSION 55-57
Thinking with responsive materials poses some in- Doumpioti, C 2011 'Responsive and Autonomous Mate-
rial Interfaces', Proceedings of the 31st Annual Confer-
triguing questions to architecture. Are we to con-
ence of the Association for Computer Aided Design in
tinue designing similar forms with new materials or Architecture (ACADIA), Banff (Alberta) , pp. 217-230
is it time to generate new connections between mat- Doumpioti, C, Greenberg, E and Karatzas, K 2010 'Em-
ter and form? Moreover, have we exploited the po- bedded Intelligence: Material Responsiveness in Fa-
tentialities stemming from traditional materials and cade Systems', Proceedings of the 30th Annual Confer-
their distinct properties in relation to acoustic, lumi- ence of the Association for Computer Aided Design in
Architecture (ACADIA), New York
nous and thermal energies? How can matter and en-
Haque, U 2004 'The choreography of sensations: Three
ergy define the formal aspects of architecture? case studies of responsive environment interfaces',
The paper poses a twofold observation: firstly is VSMM Conference Proceedings
that responsive materials can pave the path for new Harrison, AL 2013, Architectural Theories of the Environ-
connections between the static built form and sur- ment: Posthuman Territory, Routledge, Taylor & Fran-
rounding dynamic events. The steps taken by fields cis Group, New York
Hensel, M 2013, Performance-oriented design: Towards an
like material science and bioengineering make us

inclusive approach to the architectural design and the
environment, Wiley, Chichester
Ishii, H, Lakatos, D, Bonanni, L and Labrune, JB 2012,
'Radical atoms: beyond tangible bits, toward trans-
formable materials', Interactions, 19(1), pp. 38-51
Kretzer, M and Hovestadt, L 2014, ALIVE Advancements in
Adaptive Architecture, Birkhauser
Lally, S (eds) 2009, Architectural Design: Energies: New
Material Boundaries, Wiley
Leatherbarrow, D 2009, Architecture oriented otherwise,
Moe, K 2010, Thermally active surfaces in architecture,
Sadler, S 2005, Archigram: Architecture without Architec-
ture, MIT Press

Using Tensegrity and Folding to Generate Soft Responsive
Architectural Skins
Sherif Abdelmohsen1 , Passaint Massoud2 , Ahmed Elshafei3
1
The American University in Cairo, Ain Shams University 2,3 The American Uni-
versity in Cairo
1,2,3
{sherifmorad|drpassaint|a.elshafei}@aucegypt.edu
This paper describes the process of designing a prototype for a soft responsive
system for a kinetic building facade. The prototype uses lightweight materials
and mechanisms to generate a building facade skin that is both soft (less
dependent on hard mechanical systems) and responsive (dynamically and
simultaneously adapting to spatial and environmental conditions). By combining
concepts stemming from both tensegrity structures and folding mechanisms, we
develop a prototype that changes dynamically to produce varying facade patterns
and perforations based on sensor-network data and feedback. We use radiation
sensors and shape memory alloys to control the prototype mechanism and allow
for the required parametric adaptation. Based on the data from the radiation
sensors, the lengths of the shape memory alloys are altered using electric wires
and are parametrically linked to the input data. The transformation in the
resulting overall surface is directly linked to the desired levels of daylighting and
solar exposure. We conclude with directions for future research, including full
scale testing, advanced simulation, and multi-objective optimization.
Keywords: Soft responsive systems, tensegrity, folding, kinetic facades
INTRODUCTION develop low-cost and energy-saving kinetic building

This paper explores the design of kinetic façade sys- façade skins.
tems within the framework of soft architecture ma- Early examples of responsive systems included
chines (Negroponte 1975), where advanced comput- systems that enabled responsiveness by means of a
ing brings in possibilities for "living" in a "meaning- programmable façade such as the Aegis Hyposurface
ful" and responsive man-made environment, thus al- by dECOi (Goulthorpe et al. 2001), programmable
lowing for a quality of architecture that incorporates audio-visual interior settings such as the Freshwater
different needs of building occupants altogether, in- Pavilion by NOX (Lootsma and Spuybroek 1997), and
cluding climatic comfort, spatial requirements and responsive behavior through the changing form of
social interaction. Our main focus involves a "soft re- a cloud such as the Blur project (Diller and Scofidio
sponsive system" (Khoo et al. 2011) that capitalizes 2002). These "hard" mechanical approaches were
on lightweight materials and tensegrity structures to challenging for adoption in large scale architectural

skins and façade systems in terms of reliability, effi- We explore the use of soft responsive systems within
ciency and durability. They also overlooked the func- this holistic framework that extends beyond just per-
tional component of responsive architecture, espe- formance and aesthetics to satisfy user needs "as a
cially that which relates to environmental and struc- set of ever changing conditions" (Sterk 2005). We
tural integrity (Sterk 2003). put forward that architectural space is changing from
More recent work on responsive architecture a static modular state into a fluid topological state
capitalized on functional integrity, soft architectural that continually responds to human activity, social
components and material properties as opposed behavior and interaction, therefore addressing some
to highly mechanistic and complex components. of Negroponte's initial inquiries concerning human-
Oosterhuis and Sterk adopt functional approaches environment relationship, user life style, and envi-
to responsive architecture using actuated tenseg- ronmental control. In this context, we focus on de-
rity structures and pneumatic muscles, where archi- signing building façade skins that are both soft and
tectural skins respond to actual structural, climatic responsive. We implement this soft responsive ap-
and spatial conditions (Oosterhuis 2003, Sterk 2005). proach in order to respond dynamically and simul-
Khan capitalizes on the unique material properties taneously to multiple objectives such as maximizing
of elastomers to develop and construct responsive daylight and minimizing solar radiation. We specif-
structures computationally (Khan 2009). Other ap- ically use the principles of tensegrity and folding to
proaches use digital and physical computation to de- generate a kinetic façade, where members in ten-
velop elastic modular systems as a second skin to ex- sion are triggered through the sensor-network mech-
isting buildings (Khoo et al. 2011). This approach anism to produce varying patterns and perforations
capitalizes on material behavior exploration to de- in the resulting façade system.
sign morphing skins that respond to climatic con- We describe our approach below. We first experi-
ditions especially sunlight and provide aesthetically mented with tensegrity structures to build our mech-
compelling shading devices and building envelopes. anism logic. We then integrated a folding mechanism
Other advanced approaches to soft responsive to generate a working prototype with different sce-
systems implemented more sophisticated materials narios that respond to sensor-network data.
and mechanisms, such as the ShapeShift project
(Kretzer 2011), which used electro-active polymers WHY TENSEGRITY LOGIC?
to develop kinetic and responsive membranes, and In line with the main scope of this paper, which in-
the Media-ICT building (Ruiz-Geli 2011), which re- volves exploring soft responsive systems, it was es-
sponds to changing climatic conditions and moder- sential to study structures that serve the lightweight,
ates sunlight to the interior space using an ETFE ki- efficient and dynamic nature of such systems.
netic façade system. While these projects offer many Tensegrity, as a structural principle and logic, was
advantages for building skins and facades in terms seen as potentially serving that purpose. Little liter-
of performance and aesthetics, they tend to over- ature exists that addresses architectural applications
look the full spectrum of necessary conditions and of tensegrity structures in building skins and facades,
criteria. Sterk devises a hybridized model of control but rather it focuses on geometry, structural integrity
for responsive systems that extends to include: (1) and aesthetic appeal. Tensegrity structures, as de-
"user input", including the possibility of manipulat- scribed by the early pioneers Fuller and Snelson (Lal-
ing building responses, (2) "building structure", which vani 1996), comprise struts or bars in compression,
describes building responses to environmental con- in a network of strings, cables or tendons in contin-
ditions, and (3) "spatial responses", which involves uous tension. It is the properties of this integrated
the partitioning of internal space (Sterk 2005). system that give it its unique behavior of lightness,

foldability, deployability and strength. Another very in terms of cost and resources. Simultaneously, the
important characteristic of tensegrity structures is overall configuration is highly stabilized, where com-
the relatively flexible and easy shape control, where pression struts lose stiffness while tendons gain stiff-
the basic shape of a tensegrity structure can be con- ness upon loading.
siderably altered and deformed with minimal change One of the features that we capitalize on as well
in the potential energy of the structure based on a in this paper is multi-functionality, where both com-
given applied force. pression struts and tendons can represent different
Moreover, literature highlights a number of char- and diverse elements simultaneously. For example,
acteristics and benefits that are specific to tensegrity they can be load bearing members. They can also be
structures; namely natural inspiration, deployability, thermal insulators or electrical conductors. They can
efficiency, stability, and control (Skelton et al. 2001). yet represent a sensing elements, where they are able
Tensegrity structures are mostly derived from natu- to measure length or tension, or actuating elements,
ral and biological inspirations, where tensegrity be- such as nickel-titanium wire. This feature allows for
havior has been observed in cell biology. Transfer- a much larger role for the designed facade system,
able characteristics from nature can therefore possi- where it can possibly - using the appropriate selec-
bly be seen in tensegrity structures at different scales, tion of material or geometrical configuration - regu-
where smart structures could be developed to con- late thermal and electrical properties of the building
trol energy flow and motion by means of a proper se- envelope.
lection of geometry, parametric functions, and actu- The aforementioned features allowed for an ex-
ators. tensible set of properties, including (a) wholeness,
Deployability is another direct benefit that can where surfaces have the ability of responding as
be easily achieved and is significant when attempt- a whole rather than in parts, so local stresses are
ing to design responsive systems. Typically, high transmitted uniformly and are absorbed through
strength materials would tend to exhibit little deploy- the structure; (b) elasticity, where the overall struc-
ability or displacement capabilities. However, in the ture could be deformed or displaced but retains
case of tensegrity structures, the ease of attachment its original shape when necessary; (c) expandabil-
and detachment of the compression struts and their ity, where the structure is stable by itself, and so
connections renders the flexible and large displace- its sub-components could be joined together to cre-
ment of tensegrity basic components and their com- ate larger and more complex systems; (d) foldability,
pact assembly and storage a relatively light task with where surfaces could exhibit folding as a property,
considerable savings. This characteristic is highly sig- while requiring minimal energy in order to change to
nificant in complex buildings with sophisticated fa- a new configuration, and (e) uniqueness, where the
cade skins. structure has no redundant parts, implying that all
Another yet immediate benefit of tensegrity the sub-components of the structure are completely
structures is the efficient distribution of its members necessary for its overall stability.
and components through longitudinal members
which are usually organized in a non-conventional PRELIMINARY TESTING WITH TENSEGRITY
and non-orthogonal manner to achieve maximum In an attempt to reduce the highly mechanistic and
strength with minimal mass. As material is typically complex nature of kinetic facade skins, we first de-
only needed in specific locations that address struc- veloped preliminary testing with basic concepts of
tural loading points, and not just arbitrary and un- tensegrity. We used a 3-strut T-prism 30cm X 30cm
necessary locations, this would result in an efficient X 50cm tensegrity module to generate a linear sys-
three-dimensional configuration with high savings tem surface and experiment with concepts of float-

ing compression inside a net of continuous tension. sible configurations of the responsive skin, with only
After a brief physical testing and experimentation minimal alterations of input variables. This was seen
phase, we used Grasshopper and the Kangaroo plug- initially to lead to a variety of daylighting levels and
in to simulate the 3-strut T-prism behavior, where solar exposure inside the architectural space. As op-
both the compressed struts and the prestressed ten- posed to, for example, a group of hard mechanical
sioned members, or tendons, were represented (see piston movement to move or rotate merely one sur-
Figure 1). In this simulation, we could alter the ten- face panel, a slight change in input tension values
sion in each of the tendons to visualize the resulting across tendons of the overall structure could achieve
behavior in different orientations. The color code in the same result more efficiently, with significant cost
Figure 1 represents the tension values for each of the and time savings.
tendons. This allowed us to explore the scope of vari- Figure 3 shows the Grasshopper definition that
ation in configuration and transformation that could we used to define the transformation of the 3-strut
potentially take place within a building skin of depth module according to altering the values of its three
30cm. tendons. We used the Kangaroo plug-in to identify
the resulting geometrical configuration of the mod-
Figure 1
ules. Input parameters included the length of each
The 3-strut T-prism
of the compressed struts, in addition to the length of
30cm X 30cm X
the tendons at rest and in tension. As the prototype
50cm tensegrity
was meant to stand vertically being a kinetic facade,
module used for
it was important to take into consideration many fac-
preliminary testing
tors. The plug-in allowed for accounting for many
and generating a
of these, including gravity, static and kinetic friction,
linear system
stiffness, plasticity, and others. This contributed to
surface.
an informed decision making process in the design
of the prototype module, the dimensions of its pan-
els, the location of its main anchor points, its weight,
and consequently the type of lightweight material to
In order to generate the required diversity of be used in the construction of its final configuration.
patterns for the ultimate purpose of responding to In order to fully develop a framework for the ge-
sensor-network data related to daylighting and so- ometry and control of a responsive facade prototype,
lar radiation, we introduced parametric variations the paper builds on two main systems as a depar-
at three levels: (1) the modular level, (2) the one- ture point: tensegrity structures, and folding. It is im-
dimensional network of interconnected struts, and portant to distinguish here between pure tensegrity
(3) the two-dimensional network of interconnected structures and geometrical configurations that share
struts. At the modular level, each 3-strut module can tensegrity logic and characteristics. We are more in-
rotate and change its position and size, as shown in terested in tensegrity logic to generate the kinetic
the previous figure. At the 1-D network level, ma- surface properties using linear tensegrity structures
nipulating each of the three tendons per strut in- in a computational medium. It is not the aim of this
duces transformation along the linear axis of connec- paper to stay within "pure tensegrity" per se. Tenseg-
tion (see Figure 2). At the 2-D network level, the ma- rity structures are just considered as a basic departure
nipulation of the tendons across the system surface point to develop new designs, as they comprise sev-
results in a holistic transformation in aesthetic and eral interesting properties, especially when design-
shading patterns. This complex system of paramet- ing a responsive surface.
ric variations results in a highly diverse range of pos-

Figure 2
Scenarios of
resulting
configurations as a
result of
manipulating three
tendons per strut,
allowing for
transformation
along the linear axis
of connection of
struts.
Figure 3
Extract from
Grasshopper
definition used to
define the
transformation of
the 3-strut module
based on
manipulating its
tendons (color
coded).

FRAMEWORK FOR PROPOSED SOFT RE- levels according to time of day and year. See Figure
SPONSIVE FACADE PROTOTYPE 4 for a generalized framework for generating soft re-
After preliminary testing with tensegrity structures, sponsive facade prototypes.
we came to the following conclusions. Working with The framework builds on three main stages: (a)
tensegrity allowed for producing a variety of curved environmental analysis, (b) modeling and simula-
surfaces. The generated surfaces were not only two- tion of facade prototype infrastructure, and (c) phys-
dimensional, but extended to three-dimensional sur- ical testing and control. In the first stage, daylight-
faces. This was due to the fact that we used para- ing analysis using DIVA for Rhino, and solar radia-
metric variations at the modular level, and the one- tion analysis using Ladybug plug-in for Grasshopper,
dimensional as well as the two-dimensional net- are conducted in order to identify optimized values
works of interconnected struts. Controlling these for percentages of perforations for the given facade
surfaces was allowed by manipulating the values for screen. These values are given for both annual and
each of the tendons in all panels. The main prob- daily data and are fed into the Arduino microcon-
lem however, which did not allow for full control of troller to physically control the facade prototype.
the panels, was that tensegrity behavior comprises In the second stage, the mesh infrastructure of
both tension and compression, and therefore does the facade prototype is modeled using Grasshopper.
not allow for control on each and every surface of Simulation is conducted using Kangaroo plug-in for
a given panel. We thus incorporated folding as a the mesh and its folded mechanism , which is built
mechanism for generating the prototype. We subdi- based on tensegrity logic. Scenarios of configura-
vided the given panel into a group of surfaces and tions and folding behavior are explored within the
applied the tensegrity logic together with a folding computational medium, and then the model is phys-
technique on its 30cm X 30cm panels. Folding, within ically constructed. Shape memory alloys are used
tensegrity logic, allowed for a sequence of patterns as the prestressed tensioned members of the overall
across the overall structure, and a hierarchical setting surface.
where the overall panel was divided into a number Upon applying electricity to the shape mem-
of surfaces that transform in geometry, leading to a ory alloys, their lengths are transformed to respond
dynamic pattern that opens and closes, allowing for to the environmental data coming from the sensor-
different solid-void ratio along its folds and side aper- network. The Arduino microcontroller regulates the
tures. general transformation logic of the folded mecha-
We saw two components as key in controlling nism based on the input data, resulting in a spectrum
the tensegrity mechanism to allow for the required of configuration scenarios: (a) open mesh, (b) semi-
parametric adaptation: (1) radiation sensors, and (2) closed mesh, and (c) closed mesh (see Figure 5).
shape memory alloys. Based on sensor data from The configuration possibilities for the prototype
the radiation sensors, the lengths of the shape mem- are constrained to the required panel depth in front
ory alloys connected along each of the 3-strut mod- of the building exterior wall, which was confined to
ules are altered using electric wires, and are para- 30 cm in this example. Changing the depth would
metrically linked to the input data. The transforma- typically increase the number of potential configu-
tion in the resulting overall surface is directly linked rations and the nature of the resulting curved sur-
to the desired levels of daylighting and solar expo- faces, and consequently the solid-void ratio and per-
sure, where the specific parameters pertaining to the centages of perforations for the overall prototype.
percentage of perforations, orientations of tensegrity A physically tested sample is illustrated in Figure 6.
modules, and angles of solar exposure are fed in real The physical prototype comprises a number of fixed
time to the system to continually acquire the desired points on each of the subdivided panels, in addition

Figure 4
Basic workflow of
the proposed
responsive facade
prototype.
Figure 5
Possible
transformations in
the surface of the
proposed
responsive facade
prototype. Left:
Open mesh, Center:
Semi-closed mesh,
Right: Closed mesh.

to a folding mechanism and anchor points. The outer aims at building a full scale prototype and conduct-
surface of the panel can be designed to host addi- ing physical testing based on radiation sensors and
tional perforations for optimized daylighting consid- local weather data.
erations.
ACKNOWLEDGMENTS Figure 6
We would like to thank The American University in Physical prototype
Cairo for sponsoring this research under faculty re- of the proposed
search support grant agreement SSE-CANG-SA-RSG- responsive facade
FY16-18. system.
REFERENCES
Diller, E and Scofidio, R 2002, Blur: The Making of Nothing,
Harry N. Abrams, Inc., New York
Goulthorpe, M, Burry, M and Dunlop, G 2001 'Aegis Hy-
posurface: The bordering of university and practice',
Proceedings of ACADIA 2001, pp. 334-349
Khan, O 2009 'Elasticity - the case for elastic materials for
CONCLUSIONS AND FUTURE WORK kinetic and responsive architecture', Proceedings of
This paper demonstrated the process of designing a the UbiComp ‘09, Florida, USA
Khoo, CK, Salim, F and Burry, J 2011, 'Designing Archi-
prototype and developing a framework for a soft re- tectural Morphing Skins with Elastic Modular Sys-
sponsive system for a kinetic building facade. The tems', International Journal of Architectural Comput-
prototype uses lightweight materials and mecha- ing, 9(4), pp. 397-419
nisms to generate a building facade skin that is soft, Kretzer, M 2011 'Towards a new softness: The aesthet-
i.e. less dependent on hard and complex mechani- ics of soft dielectric electroactive polymers and their
cal systems, and responsive, i.e. dynamically and si- application in an architectural context', International
Adaptive Architecture Conference, London, UK
multaneously adapting to spatial and environmen- Lalvani, H 1996, 'Origins of Tensegrity: Views of Em-
tal conditions. We built on tensegrity logic, rather merich, Fuller and Snelson', International Journal of
than a direct translation of pure tensegrity structures, Space Structures, 11, pp. 27-55
and incorporated a folding mechanism to develop Lootsma, B and Spuybroek, L 1997, 'Aquatic pavilion and
a lightweight prototype. We used radiation sensors interactive installation', Domus: Architecture, design,
with Arduino to respond to daylighting and solar art, communication, 796, pp. 28-33
Negroponte, N 1975, Soft Architecture Machine, MIT
radiation analysis in Rhino and Grasshopper simu- Press, Cambridge
lations. We tested the prototype physically using Oosterhuis, K 2003, Hyper bodies - Towards an E-motive
shape memory alloys, where electricity is applied to Architecture, Birkhauser, Basel
transform the solid-void ratio of the panel surfaces Skelton, RE, Adhikari, R, Pinaud, J-P, Chan, W and Hel-
based on the surrounding sensor-network data. ton, JW 2001 'An Introduction to the Mechanics of
The scope of this paper is limited to digital sim- Tensegrity Structures', Proceedings of the 40th IEEE
Conference on Decision and Control, Orlando, FL,
ulation of folding and tensegrity using Grasshopper USA, pp. 4254 - 4259
and Kangaroo plug-in and their behavior under the Sterk, TDE 2003 'Using Actuated Tensegrity Structures to
stimulus of preset values, along with some prelimi- Produce a Responsive Architecture', Proceedings of
nary physical testing. This allowed for the genera- ACADIA 2003, Indianapolis, USA, pp. 85-93
tion of multiple iterations of façade patterns and test- Sterk, TDE 2005, 'Building upon Negroponte: A Hy-
ing the resulting patterns against the multiple objec- bridized Model of Control Suitable for Responsive
Architecture', Automation in Construction, 14(2), pp.
tives of radiation and daylighting. Further research 225-232

Sensitive Shelters: Poetics of Interaction
Andrés Martin Passaro1 , Gonçalo Castro Henriques2 ,
Rodrigo Cury Paraizo3
1,2
Universidade Federal do Rio Janeiro, Laboratório Modelos e Fabricação Digi-
tal, Pós-Graduação Urbanismo 3 Universidade Federal do Rio Janeiro, Pós-Graduação
em Urbanismo, Laboratório Análise Urbana e Representação Digital
1,2,3
{andrespassaro|gch|rodrigoparaizo}@fau.ufrj.br
This paper describes and reflects about a workshop activity in the field of Digital
Manufacturing technologies to build responsive shelters that interact with their
users and the environment. It addresses a teaching strategy intended to overcome
tooling or the simple use of instruments and proposes instead to frame the
production of objects using a new language, or a new operative strategy, directly
linked to the production of the objects. It addresses a teaching strategy behind the
workshop two main levels: first, by the development of technical skills by means
of an operative action directly linked to the production of the object, and not
apart from the action of making it (as in learning first and applying later). And
second - and no less important -, it helped foster the maturation of critical
thinking arising from the creation of a dynamic object of architecture - with
moving parts and programmed to respond to its users.
Keywords: Digital Fabrication, Parametric Design, Responsive Architecture,

Sensitive Shelters
INTRODUCTION how to operate software or how to use digital cutting

Digital simulation and fabrication tools can empower tools or 3D printers. To overcome a simple and empty
architects to regain awareness of how things are technical discourse implies the search for new lan-
built, understanding the architectural modes of pro- guages, in order to explore the new operational pos-
duction. Traditional design process using analogue sibilities of the new tools, maybe most notably the
methods seldom the person who designs manufac- embodiment of processes in architectural spaces. A
tures and builds the project is the same. With the Flusserian approach of meddling with the "black box"
digital integration (Kolarevic 2003), processes can is used to subvert its "program" and re-appropriate
have a wider interference in project. Indeed, visual the technology from the artist's point of view (Flusser
programming itself, commonly present in paramet- 2011).
ric design, with its intensive use of diagrams, helps The increased use of parametric design and digi-
shift emphasis from form to process (just as diagrams tal fabrication technologies is not intrinsically bound
in contemporary architecture discourse so often do). to produce any specific result - as with any tech-
This workshop, therefore, was not about teaching nology, it depends on how we choose to use them.

Although new technologies can support the cur- the exploration of parametric solutions that surpass
rent architectural production, that still oscillates be- the traditional sequential logic of Problem> Project>
tween narrative fragmentation and super-modern Building resulting in a single solution. The tradi-
compactness (Moneo 1999); they may also become a tional mode of design, based on paper documenta-
major tool in the service of bland real estate develop- tion produced by someone that is later used by an-
ments. The workshop intended to explore new tech- other one to fabricate the parts and eventually the
niques searching for new paradigms to change (or object is built by a third person, induces communica-
at the very least challenge) the current architectural tion and learning gaps that can successfully be over-
discourse. These tools can modify the modes of rep- come with digital integration (Kolarevic 2013). More-
resentation, the use of the materials, the production over, as Carpo notices (2011, 44-45), digital fabrica-
process and finally the very understanding of archi- tion challenges the Albertian principle of separation
tectural form. The idea to couple visual abstract con- between notation and building, as architects design
cepts (diagrams) with construction, results that these and build at the same time. As Dollens (2002, 106)
processes somehow subvert the Albertian separation notes, in this scenario conceptual design and digi-
of notation and construction that once was essential tal fabrication originate in the same intellectual and
to the profession of Architecture (Carpo 2011). They representational gesture, as exemplified by the archi-
therefore place the current critical systems in a crisis tecture explorations in digital fabrication and algo-
situation and provide a new path about Architectural rithms of Bernard Cache's Objectile.
processes and the language associate with them. Parametric design can generate families of solu-
tions using new logics of modeling and modifying
DIGITAL DESIGN AND FABRICATION PRO- the design sequence (De Landa 2011). Parametriza-
tion fosters new logics of modeling, thus modifying
CEDURES
the sequence of design with the use of mathematical-
When LAMO - Laboratório Modelos e Fabricação
geometric functions replacing notations based on
Digital (Laboratory of Fabrication and Modeling)
the identity between object and description; propos-
acquired the first digital manufacturing machines,
ing instead the development of series of objects
learning parametric design would appear to solve
with parameter-controlled variations between them
any problem regarding the production of innovative
(Carpo 2011). At the same time, tinkering with func-
contemporary architecture. However, after a while
tions and their parameters may be seen as an exter-
we realized - with a feeling of déjà vu - that this
nalized black box, helping to develop the (meta) de-
discussion was very similar to the one that started
sign feedback procedure proposed by Jones (1972),
some 30 years ago when the computer aided de-
as reported by Barki (2006, 95), and thus help shed
sign (CAD) was introduced in professional practice
some light at design heuristics in general. It is also af-
and academia. Now, as then, the simple use of more
filiated to the concept of "reflection-in-action", devel-
technologically advanced tools does not mean that
oped by Schön (1983), as commented by Barki (2006,
the discipline has absorbed the specificity of the new
113), in the sense that those external representations
technologies.
made by the designer, such as drawings and mod-
The ability to automate design procedures to
els, will be reincorporated as images feeding new de-
respond to environmental demands also does not
sign alternatives - only now we can also add functions
mean by itself a better design. Exploring new meth-
and algorithms as proto-imagetic material. Not only
ods to improve quality, social justice or the fore-
they generate several new images, but they also al-
front of thought require, in addition to proficiency,
low some degree of randomness to be carefully intro-
a methodological review of the act of design that
duced in the whole process, increasing the chances
makes the practice. The new digital processes allow

of serendipitous findings. fying processes characteristic of Somol's "diagram-
It should be stressed that the possibility of cre- matic practice" (2010, 90). According to Somol, "the
ating an automated, optimal response of the form role of the architect in this model is dissipated, as he
to external demands does not necessarily imply in or she becomes an organizer and channeller of in-
a better design. Computation can also be used in formation", from vertical forces (gravity, for instance)
passive forms of optimization - like finding a specific to horizontal ones (social, political, cultural and eco-
shape that can profit more from solar orientation, for nomic).
instance. In the end, the exploration of new meth- Starting from these premises, uncertainty arises
ods that promote building quality, social justice or on how to teach parametric design and digital fab-
an avant-garde design requires, apart from immedi- rication in order to promote a significant paradigm
ate knowledge about the use of tools, the ability to shift in the architectural production, and the present
think critically about the whole design process. It research aims to improve the discourse and the
is possible to relate this, along with the relative in- method of design including these new tools and
stability of form that can be found in parametric de- procedures. Using the CAD to emulate a manual
sign and the diagrammatic expression of visual pro- draft is not considered a paradigm shift, as it is also
gramming languages, to the multiplication of signi- no paradigm shift to simply substitute the utility
Figure 1
Tools and devices
from the Workshop
"Sensitive Shelters".

knife for a laser-cutting machine. In order to have a new conceptual method. The choice of the work-
a paradigm shift it is necessary to incorporate a dis- shop format answered our needs to have less hierar-
course that legitimizes the interventions.The poetic chical relationships among participants, an intensive
of architecture is present in the strategy used to de- participation over a short time frame, and to foster
velop project, incorporating in the learning process quick, iterative responses from the students.
the tools and digital processes. The project strategy The workshop became an annual joint activ-
to find design solutions requires the participants to ity among three laboratories, LAMO, LAURD - Lab-
establish an 'image as a synthesis' of project (Cache oratório de Análise Urbana e Representação Digital
and Bergson). That image must be developed from a (Laboratory of Urban Analysis and Digital Represen-
diagram of a mechanism provided at the beginning tation) and NANO - Núcleo de Arte e Novos Organ-
of the workshop. The participants are encouraged to ismos (Nucleus of Art and New Organisms). The first
develop a design idea from these diagrams. Partici- two are part of the Program of Post-Graduation in Ur-
pants should therefore seek, instead of a priori solu- banism of the School of Architecture, and the latter
tions, for solutions that derive from the use of new from the School of Fine Arts, all from the Federal Uni-
tools and processes taught. In this sense a theoreti- versity of Rio de Janeiro.
cal and practical search that is analogue-digital incor- The intention is to incorporate contributions
porates new knowledge in the design process (Dol- from different fields such as architecture, digital arts
lens 2002). In this sense, the culture of project instead and sciences. These different areas are exploring
of generic solution of generic exercises is one of the the new possibilities of digital to integrate in their
strategies used to overcome the pure instrumental- practice. Because they possess different specificities,
ization and to go beyond exercises alien from the de- they find different solutions of how to incorporate
sign context. the digital, therefore making it crucial to share ex-
At the "Universidade Federal Rio de Janeiro", as periences and techniques. By including these differ-
in most Brazilian universities there is little awareness ent areas, the intention is to mingle different modus
of the importance of the integration between digi- operandi on the use of digital processes and instru-
tal design and manufacturing (Henriques and Bueno ments, thus helping redefine the design research.
2009). The training that exists is usually based on Apart from the variations that accompany paramet-
uncritical learning of software without the actual re- ric design, the workshop brought also the multiplic-
interpretation of design processes. Instead new tools ity of states of interactive systems to the design-
are used in the framework of traditional processes, ers' tables. From the architects' point of view, deal-
only aiming to improve the speed and efficiency of ing with interactive structures increases the need for
project development. hands-on experimentation; but it also adds a new
layer of meaning to the design. The structures not
THE WORKSHOP PRODUCTION only change over time, but they have a behavior,
The demand for a discourse or a legitimate narra- and interact with users, forcing the participants to
tive of the production is undoubtedly better than develop the conceptual apparatus and aesthetic vo-
the technical option. Only this understanding would cabulary needed to design shapes, structures, feed-
allow perceiving the change that these new instru- back loops and (meaningful) movements. The under-
ments can bring to the project, expanding the pre- lying (diagrammatic) system must also possess a phe-
vious possibilities. In this sense, the workshop was nomenological presence, capable of triggering sub-
proposed and conducted with the intention to over- jective responses from viewers. Relative to the tradi-
come the stage of pure exploration of digital design tional method such processes and instruments allow
and manufacturing tools, and induce the search for intervening in all project phases in an asynchronous

sequence, unlike some previous (more traditional) is only possible manufacturing and testing solutions
methods in which the characteristics of the medium with physical models. Thus the teaching of these
themselves favored more sequential and unidirec- different areas promotes not only the reflection dur-
tional design processes (Duarte et al. 2011; Oxman ing the learning and use of these new processes and
2006). tools but also requires the definition of a concept.
The workshop targets participants with minor This method originates multiple proposals that are
experience in digital design. Its public mixes un- perfected by trial and error, giving participants a valu-
dergraduate students, Master and PhD candidates able learning feedback. This experience can be used
and even professors, thus creating very interest- in the future for the student to design, cut and man-
ing interactions in each team. It was necessary to ufacture their projects. The participants of the ex-
reintroduce the concept of 'computer aided design', ercise should develop interactive spatial sculptures
along with notions of associative and parametric de- or 'follies' (for lack of a more precise denomination)
sign, digital fabrication and new media like Arduino, from five suggested primary responsive mechanisms,
creating a different - and expanded - mindset for essentially names and diagrams: curtain, octopus,
the participants. Arduino is responsible for mak- shell, Chinese lantern and fuchsia flower ("brinco-de-
ing accessible the production of responsive mech- princesa", in Brazil).
anisms - and therefore the responsive shelters that The workshop was immersive - and intensive; it
ought to be designed within the 2014 edition work- lasted for nine days, starting on a Saturday and fin-
shop. The tutorial sessions included digital model- ishing on the next week's Sunday. The first four days
ing with Rhinoceros and basic visual programming had lectures in the morning sessions and software tu-
with Grasshopper, including various plug-ins to gen- torials in the afternoon ones. Participants were di-
erate multiple design alternatives. The advantage vided in five teams of six people. Each group was
of computer-aided manufacturing, that is, the dig- installed in one of the working tables of LAMO, with
ital integration among these programs in order to enough space for sketchbooks, laptops and the many
digitally control the production of physical objects, parts that constituted each prototype. Four teachers
was quickly perceived by the participants after some were responsible for the workshop, three also acting
training in the operation of the laser cutting machine as instructors during the tutorial phase. Fifteen lab-
and the 3d printer, as well as some basic training in oratory monitors took part in the workshop to help
carpentry techniques, metalwork and some experi- operate the machinery (digital and analogical) and
ence with product assembly. In this context, the act were also responsible for the daily organization of the
of design was experienced as a digital continuum, workshop itself. Undergraduate and graduate stu-
which also influences the project itself and the se- dents showed up and worked as volunteers as well.
quence of operations used. The moment a line in One Lab monitor was designated for each group. The
the design is no longer just a representation, but also more experienced with parametric design were con-
happens to be an actual cut line in a subsequent stantly rotating among the groups helping them as
operation, the integrated operation acquires a new difficulties came along.
meaning. The undergraduate students participation as
However the starting point of the design exercise monitors cannot be underestimated when consider-
is conceptual in nature. Certain primary responsive ing a workshop based on new technologies. Their ex-
mechanisms were suggested and the participants perience arises from the daily operation of the Lab,
should - within a short but intensive period - con- which gives them basic skills to safely operate with
ceive, fabricate, and assemble the solution. The de- different materials and techniques such as laser cut-
sign should include automated moving parts, what ters, 3D printing, CNC routers as well as traditional

modelling hardware and tools. Besides, most of them may indicate that those official financing and sup-
are enthusiastic - and usually self-taught - explorers porting channels have not yet constituted mecha-
of parametric design. nisms to deal with this kind of demand linked to new
The workshop also used several instructors for technologies and, perhaps, the very spirit of fabrica-
the tutorials on parametric design and electronics. tion and the "maker revolution".
Most of these instructors were recently graduated - In short, in the first days, instructors conducted
with a few holding a MA degree. Apart from the basic tutorials on parametric design and digital fab-
software tutorials, they also gave lectures about their rication with the participants, who also received ba-
works and the theoretical approach they were us- sic training in the use of the laser cutter and the
ing. The tutorials of Victor Sardenberg, Orion Cam- 3D printer. We presented five primary devices (cur-
pos and Verônica Natividade discussed parametric is- tain; octopus; shell; Chinese lantern; fuchsia flower),
sues using Rhinoceros + Grasshopper. On the elec- that served as an inspiration for five responsive de-
tronics sessions with Arduino and motion sensors, signs. From the fourth day of the workshop on, at
Lucas de Sordi was in charge of robotic tutorials (Ar- the end of each day, there as a small seminar where
duino and sensors) and Marlus Araújo was responsi- each team would present updated information on
ble for addressing 3D printing and rapid prototyping. their concepts, ideas, mechanisms and shapes. This
The only lecturer with Doctoral Degree was Gonçalo schedule revealed itself paramount to the proper de-
Henriques - who also was a tutor. He shared his velopment of each design, among other things be-
practical experience highlighting the importance of cause daily presentations demanded carefully elabo-
computation, fabrication processes and digital inte- rated concepts, overcoming the illusionary charm of
gration, paving the way for new directions in future techniques and tools. The results embodied both the
workshops. Professor Guto Nóbrega, NANO coordi- knowledge and the trial and error of the workshop,
nator, introduced the issues of hybridization through going beyond the dichotomy shape-tooling.
interactivity with living organisms. Professors Andrés
Passaro and Arthur Lara also presented their lectures. WORKSHOP CHALLENGES: CONCEPTS,
We would like to consider a small digression
ART AND TECHNIQUES
here, in order to illustrate issues typical of new tech-
During the workshop five propositions were devel-
nologies - at least in the digital age. This par-
oped around the theme of "sensitive shelters", that
ticular type of knowledge is more widely spread
is, shelters composed of movable parts that should
among those recently graduated (or even under-
embody a concept in the interaction with the human
graduate students) than tenured professors, for in-
presence, either with a mechanism of actuation or in
stance. There is not, at least in Rio de Janeiro, a con-
the underlying geometrical explorations. Each team
siderable number of academic researchers that mas-
designed and built working mock-up prototypes.
ter digital fabrication and parametric design tech-
The basic idea to go beyond basic and avoid
niques yet. This handicap leads to the situation
strict sequential design thinking (from idea to shape
where instructional sessions have to be conducted
to form, for instance) was the stimulation of inte-
by those recently graduated. While it does not pose
grated action in several moments of the creative pro-
a problem for those directly involved (on the con-
cess. The device - and its supporting concept - should
trary, it is healthy to invert the roles of teachers and
work beyond the rhetorical level, because it had to
students from time to time), it is not easy to have
be prototyped, that is, its parts should work properly,
research agencies to fund transportation and hotel
turning design concepts into matter. The design thus
costs for people that possess a knowledge highly
got scale and movement, and a narrative associated
specialized but not formally acknowledged. That
that should include mechanical, sensorial and algo-

Figure 2
The "Ephyra" design
development (left).
Design
development of the
"Origami" (right).
rithmic concepts at its core. The design system com- and was affected via mouse (simulating data from
prises digital fabrication (CAM) on top of the simula- the environment); later the mouse was suppressed
tion of the mechanisms (CAE) and the design concep- in order to have actual readings from the sensors
tion (CAAD), simultaneously dealing with the parts trigger the responses on the computer screen. Fi-
and the whole. nally, when the models were built, the computer's
Each design was conceived using algorithms role was reduced to convert data from the environ-
written in visual language (Grasshopper software ment in instructions for the actual engines, operating
was used), taking into account the sensor's interfaces on a physical model, becoming thus an intermediate
that would trigger associated re-actions. Light and operator in a responsive product.
presence sensors were used to capture data from the The workshop lectures reinforced the construc-
environment so an Arduino-based interface could tive issues and possibilities of built projects, connect-
transmit this data and processes the date in the CPU. ing architectural theory and practice, showing the
The algorithm here is understood as a "soft" inter- generation, simulation, fabrication and (responsive)
face controlling the relationship between geome- interaction of project from the physical and techno-
tries and their becoming, while transmitting actions logical point of view of the maker. Each team then
to Arduino. From this moment on, Arduino reports tried to enhance their experiences with those solid
instructions as a "hard" interface to activate engines references acquired with the lectures, and the re-
that spin cogs in charge of the movements of the sult was a growing enthusiasm of the participants,
shelter. At first, the geometry resided only in the CPU even among those that did not master the mouse nor

Figure 3
Design
development of the
"Xtensão" (left).
Design of the
"Caterpillar" (right).
the pliers. More than becoming proficient with new therefore should appear as an interaction between
tools, they dealt with new ideas reconfiguring them- the physical and digital worlds - and they even hap-
selves from an empirical process of trial and error. pened to extrapolate the original diagrams at times.
Instructors encouraged symbiotic actions be- G1 - The ÉFIRA ("Ephyra"; Figure 2 Left) is a mech-
tween physical and digital modes of creation and anism with a vertical axis in which dishes on both
production, while advising each workgroup. Each ends rotate in opposite directions. This movement
team proposed new solutions and developments creates a torsion in the exterior skin formed by stripes
and reviewed the results at a very fast pace as de- connecting the dishes. An Arduino presence sensor
signs were rapidly materialized thanks to digital fab- triggers the motion-controlling algorithm that acti-
rication. vates the rotation engine in reverse directions at each
extreme of the axis. The proximity of the user there-
DESIGN DISCUSSION: 5 DIAGRAMS, 5 fore causes the skin of the object to rotate and open
up so he or she can enter.
MECHANISMS
G2 - The ORIGAMI (Figure 2, right) describes a
The core strategy of the workshop derived from the
movement of simultaneous expansion from the cen-
five diagrams and derived mechanisms. The five di-
ter to the four borders in the face of a transparent
agrams presented were also named. As Picon (2004,
cube. A leap motion potentiometer detects human
118) stated, diagrams, in fact, are "inseparable from
presence and makes the folds open up using pulleys
courses of action" throughout "the world appears as
and cables activated by an engine. The system is me-
a field in which forces are manifest rather than static
diated by an Arduino board that manages commu-
geography", and the workshop presented the chal-
nication between sensors and the controlling algo-
lenge of designing the object as a re-action to the
rithm.
environment, as a kind of "live" diagram. The result
G3 - XTENSÃO ("Xtension"; Figure 3, left) is based

on tensegrity structures, that is, a combination of center, becoming arches. Human presence triggers a
bars and cables forming a net of continuous tension. motion sensor that transmits the signal via Arduino
Five pairs of bars and cables are grouped around a cir- to the algorithm, where it operates the four fuse en-
cle; the bars in each module stay up and are arranged gines that bend or stretch the stripes. As people ap-
in a polar array. The movement of contraction and proach the stripes, they progressively bend, relaxing
retraction of the bars towards the center of the circle as people go away.
reconfigures the whole structure. Human movement G5 - The MUSCIPULLUS (Figure 4) emulates a car-
and vegetable-generated impulses trigger responses nivore plant: this shelter opens up as the visitor ar-
from the algorithm through the Arduino board, acti- rives. The shape derives from a set of vertical poles
vating a rotating engine in the center. This engine arranged in a polar array. Each pole is part of a panel,
is connected to five pulleys placed along the circum- and the synchronized movement of the five panels
ference, that contract and retracts the cables, thus re- opens up the surface when visitors arrive and close it
ordering the tension - and shape - of the whole struc- as a trap when visitors reach the inside. The interac-
ture, since bars are pushed towards the center or the tion of sensors and information processing are simi-
border. In the absence of stimulus, the structure re- lar to the previous cases. The algorithm commands
turns to its initial rest state. Arduino to start a central rotational engine, which,
through levers, transmits movement to each of the
Figure 4
five poles, making the carnivore plant open or close
Muscipullus
in interaction with the visitor.
conponents
The emphasis on constructive signal-processing
development.
issues in this section should not lead one to under-
estimate the role of the diagrams in the whole sys-
tem. Eitheras the work was an explanatory synthe-
sis of complex design situations or a visual rhetoric
argumentative device that helped developing the
designs. We should also mention the contribution
of visual programming languages with the use of
Grasshopper to the very possibility of designing and
fabricating the shelters. They make computation
with algorithms more accessible to architects, for in-
stance, but they also make them understand form as
a product of visual thinking and diagrams.
The daily observation of the rhetoric of each
group to describe their designs, even when talking
about their concepts, still reinforced the technolog-
ical aspects. They could only stray away from those
aspects as they mastered the technology. Therefore,
G4 - The CATERPILLAR (Figure 3, right) is constituted as the workshop progressed, instead of teaching new
by longitudinal stripes that rest on the soil. They are tools, we focused on the research for new types of
sensitive to human proximity and bend themselves mimeses, developing a discourse around the con-
like a shell to welcome the visitor. The stripes mod- cepts of bio-shapes and morphogenesis, based on
ify themselves according to the distance of the stimu- proto-mechanics, sensorial algorithms.
lus. They contract or expand from the borders to the

CONCLUSIONS architecture design. Apart from yet another digital
The Sensitive Shelters workshop is just the starting tool, the process induces working with multiple em-
point of a broader discussion. In broad strokes, its pirical propositions, enhancing the usual trial and er-
format tried to recover conceptual devices applied ror of architecture design, something students can
to design, integrating procedures and technologies later employ in their designs. The feedback from the
connected to the act of building. students, showing the importance of the design de-
The intention was to overcome the mere training cisions they had to face, helps corroborate the whole
on the use of new tools; it was not about doing the experience, as demonstrated in this video recorded
same things with new tools, but about researching in the end of the workshop [1].
new possibilities brought by new instruments. From The results, apart from questioning traditional
a Flusserian perspective (Flusser 2011), we could say training as a matter of learning new tools, reinforce
that, in order to subvert the programming of the de- the need to embed a critical approach and the artic-
vice, the programming had to be rewritten, even if ulation of techniques and concepts within these sys-
using the device's own language. tems, bringing new areas of knowledge to the archi-
This discussion ensues that training and tuto- tecture practice. This type of learning goes against
rials, and the new instruments themselves, do not specialized, compartmentalized knowledge: in the
automatically imply meaningful changes in architec- workshop, design solutions were tested through
ture designs. The workshop focused in the perceived their construction, integrating digital and analogical
need to find a novel narrative for those new instru- practices as a single continuum, without the artificial
ments, less literal and with stronger conceptual back- divide that is usually imposed upon those domains.
bones to legitimate or effect an imminent paradigm As a challenge, we can define this workshop as
shift. If the results described sound more as curiosi- a work in progress that creates the conditions for us
ties than revolutions, albeit interesting curiosities, it to explore other ways of architectural production and
is because they indeed do not introduce enough el- design. In LAMO, the laser cutter was first an im-
ements to justify such shift. However, each proposi- proved utility knife, but we soon realized that under-
tion sets solid grounds for their conceptual discourse, stand it that way as a conceptual mistake. The tech-
and overcomes not only the mere "re-drafting" but nology was to be understood as an opportunity to
also the empty yet mesmerizing hedonistic formal think differently, drawing upon the specificities of the
games. instrument to enrich the design process.
Throughout two intense weeks, participants ac- The workshop successfully produced a critical
quired basic proficiency in algorithmic design and discourse in this sense, gathering from the reunion
digital fabrication, as well as conceived, fabricated of several practitioners and thinkers under the same
and assembled a reduced scale prototype of their roof, involved in a practical activity while bringing
chosen design. Prototypes included mechanic parts and refining their own reflections upon this practice.
and electronic sensors in order to allow experimen- Participants were exposed to discussions from para-
tation (and troubleshooting) with actual movement metric design to hybridization of living organisms, in-
and motion and presence detection, refining the de- cluding robotics and 3D printing, at the same time
sign through fabrication. More than promote the use with the notion that these themes are all interrelated,
of digital tools or the reflection upon the design con- and that each of them has possible contributions to
cept, the procedure allowed the participants of the the design of architectural objects.
workshop to understand how both systems can work This window of opportunity is in fact a network
together, fostering new modes of enquiry based on a that is woven little by little, through the acquisition
deeper understanding of the algorithmic logic within of technology and equipment as well as the slow yet

Figure 5
Sensitive Shelters/
abrigos sensíveis
group at LAMO,
Universidade
Federal do Rio de
Janeiro, July 2014.
progressive development of knowledge and a huge the angel can no longer close them. The storm irre-
will to do so. This account reminds us Benjamin's "An- sistibly propels him into the future to which his back
gel of History" (2012 [1940]), in the sense that, as we is turned, while the pile of debris before him grows
look back, we perceive several modes of production skyward. This storm is what we call progress." (Ben-
becoming obsolete. Ours is a less nihilist point of jamin 1969)
view than Benjamin's, our Angel not looking back in
dismay but looking forward with enchantment. Future Work
(Note: cross-reference original text Angel of His- This workshop is part of a series that aims to im-
tory: "A Klee painting named Angelus Novus shows prove upon previous experiences and further refine
an angel looking as though he is about to move away the theoretical and technical framework along with
from something he is fixedly contemplating. His eyes practical constraints of time and materials. It has
are staring, his mouth is open, his wings are spread. already been followed by another workshop, called
This is how one pictures the angel of history. His "Defying Gravity" (Henriques and Passaro 2016). New
face is turned toward the past. Where we perceive a experiences should address more focused concepts,
chain of events, he sees one single catastrophe which allowing faster response time (and more experimen-
keeps piling wreckage upon wreckage and hurls it tation with algorithms and fabrication) from partic-
in front of his feet. The angel would like to stay, ipants. Other themes include temporary interven-
awaken the dead, and make whole what has been tions and tactical urbanism.
smashed. But a storm is blowing from Paradise; it
has got caught in his wings with such violence that

ACKNOWLEDGEMENTS Dollens, D 2002, De lo digital a lo analogico, Gustavo Gili,
The authors would like to thank CAPES, CNPq and Barcelona
FAPERJ for the funding of this project and related Duarte, J, Celani, G and Pupo, R 2011, 'Inserting com-
putational technologies in architectural curricula',
scholarships.
in Ning, G and Xiangyu, W (eds) 2011, Computa-
tional Design Methods and Technologies: Applications
WORKSHOP CREDITS in CAD, CAM and CAE Education, IGI Global, Hershey,
Coordination: Andrés Passaro & Guto Nóbrega. Orga- PA, pp. 21-55
Flusser, V 2011 [1983], Filosofia da caixa preta: ensaios
nization: Clarice Rohde, Elisa Vianna, Gonçalo Castro
para uma futura filosofia da fotografia., Annablume,
Henriques, Laura Basile, Rebeca Duque Estrada. Lec- São Paulo
tures & tutors: Andrés Passaro, Arthur Lara, Lucas de Henriques, GC and Bueno, E 2009, 'Geometrias Com-
Sordi, Gonçalo Castro Henriques, Guto Nóbrega, Mar- plexas e Desenho Paramétrico.', Vitruvius / Drops, 30,
lus Araújo, Oríon Campos, Verônica Natividade, Vic- pp. 30-37
tor Sardenberg. Scientific Committee: Andrés Pas- Henriques, GC and Passaro, A. M. 2016 'Defying grav-
ity: From statics to dynamics, from objects to sys-
saro (FAU-UFRJ), Arthur Lara (FAU-USP), Carlos Fe-
tems', Proceedings of Architecture In-play conference
ferman (FAU-UFRJ), Luiz Felipe Cunha (FAU--UFRJ), 2016 (Forecoming), Lisbon
Gonçalo Castro Henriques (FA-UTL Portugal), Guto Kolarevic, B 2003, Designing and Manufacturing Architec-
Nóbrega (EBA-UFRJ), Naylor Vilas Boas (FAU-UFRJ), ture in the Digital Age, Spoon Press Taylor & Francis
Malu Fragoso (EBA-UFRJ), Rodrigo Cury (FAU-UFRJ), Group, New York
Alexandre Pessoa (UFF), Verônica Natividade (PUC- De Landa, M 2011, 'Virtual Reality', in Menges, A and
Ahlquist, S (eds) 2011, Computational Design Think-
Rio); LAMO Monitors and Photographers: Diogo de la
ing, John Willey & Sons Ltd, Chichester, United King-
Vega, Felipe Madeira, Filipi Dias, João Magnus, Jonas dom, pp. 142-8
Abreu. Moneo, R 1999, 'Los 90 entre la compacidad y la frag-
Workshop Participants: Andrea Baran, Aydam de mentación', Arquitectura Viva, 66, pp. 17-24
Paula, Bruno Vianna, Carina Carmo, Fernanda Petrus, Oxman, R 2006, 'Theory and Design in the First Digital
Gabriela Castro, Guilherme Erthal, Isabella Costa, Ju- Age', Design Studies, 27(3), pp. 20-40
Picon, A 2013, 'A arquitetura e o virtual: Rumo a uma
liana Tobar, Karla Regina Coutinho, Klauss Borges,
nova materialidade', in Sykes, AK (eds) 2013, O
Leroy Otto, Lisa Schweigert, Lucas Fontes, Lucas Pa- campo ampliado da arquitetura: Antologia teórica
cobahyba, Luciana Mayrink, Marcos Pavão, Mariana (1993-2009), Cosac Naify, São Paulo, p. 205–220
Alvares, Mariana Lourenço, Pedro Henrique Oliveira, Somol, RE 2010, 'Dummy text, or the Diagrammtic Ba-
Priscila Luzardo, Ricelli Laplace, Ricardo Lopes, Ro- sis of Contemporary Architecture', in Garcia, M (eds)
drigo Gurgel, Sérgio Porto, Tainá Motta, Thiers Freire, 2010, The Diagrams of Architecture: AD Reader, John
Willey Sons Ltd,, Chichester, United Kingdom, p.
Thomaz Vieira, Tiago Maciel, Yana Inoue, Filipi Dias de
88–91
Oliveira, Marlus Mendonça, Raissa Laban, Diogode la Vider, A 2013, 'O campo ampliado da arquitetura', in
Vega. Sykes, AK (eds) 2013, O campo ampliado da arquite-
tura: Antologia teórica, Cosac Naify, São Paulo, p.
242–251
REFERENCES [1] https://vimeo.com/127316023
Barki, J 2003, O risco e a invenção: um estudo sobre as no-
tações gráficas de conceção no projeto, Ph.D. Thesis,
Universidade Federal do Rio de Janeiro
Benjamin, W 1969, Theses on the Philosophy of History,
Illuminations (trans. Harry Zohn), Schocken Books,
New York
Carpo, M 2011, The Alphabet and the Algorithm (1st ed.),
Cambridge, Massachusetts, The MIT Press, London

Adaptive Pneumatic Shell Structures
Feedback-driven robotic stiffening of inflated extensible membranes and
further rigidification for architectural applications
Paul Poinet1 , Ehsan Baharlou2 , Tobias Schwinn3 , Achim Menges4

1
CITA - Centre for Information Technology and Architecture | ICD - Institute for
Computational Design 2,3,4 ICD - Institute for Computational Design
1
paul.poinet@kadk.dk
2,3,4
{ehsan.baharlou|tobias.schwinn|achim.menges}@icd.uni-stuttgart.de
The paper presents the development of a design framework that aims to reduce
the complexity of designing and fabricating free-form inflatables structures,
which often results in the generation of very complex geometries. In previous
research the form-finding potential of actuated and constrained inflatable
membranes has already been investigated however without a focus on fabrication
(Otto 1979). Consequently, in established design-to-fabrication approaches,
complex geometry is typically post-rationalized into smaller parts and are finally
fabricated through methods, which need to take into account cutting pattern
strategies and material constraints. The design framework developed and
presented in this paper aims to transform a complex design process (that always
requires further post-rationalization) into a more integrated one that
simultaneously unfolds in a physical and digital environment - hence the term
cyber-physical (Menges 2015). At a full scale, a flexible material (extensible
membrane, e.g. latex) is actuated through inflation and modulated through
additive stiffening processes, before being completely rigidified with glass fibers
and working as a thin-shell under compression.
Keywords: pneumatic systems, robotic fabrication, feedback strategy,

cyber-physical, scanning processes
FROM PHYSICAL POST-PROCESSES TO qualities of the final product. Each piece needs to be
CYBER-PHYSICAL FORM GENERATION post-processed uniquely and separately (with differ-
The global design of a free-form inflatable structure ent offsets and resizings) in order to consider differ-
typically needs to be divided into developable strips ent assembly situations (different valences) and var-
(Figure 1 - left) or triangulated patterns in order to ious pressure conditions. This post-process - most
avoid any wrinkles that would diminish the aesthetic of the time partly empirical - often leads to an accu-

Figure 1
A traditional
post-rationalization
of a free-form
inflatable
membrane (left) is
replaced by a
cyber-physical
environment
(right).
mulation of tolerances that are difficult to tackle and lations to fulfill contradictory purposes. The "steering
avoid during the assembly phase. The assembly proof form" is mostly used in a digital context. The pre-
cess also presents some logistical difficulties as it is sented research tries to translate it at full scale in a
necessary to sort all the pieces and find their respec- cyber-physical environment (Figure 1 - right), where
tive locations on the construction site. physical matter and digital analysis interact together
The aim of this research is to create a self- until the inflated membrane satisfies both structural
supporting structure that is not given and further dis- and spatial conditions.
cretized for fabrication purposes but generated from
the material behavior which is modulated by exter- A cyber-physical setup
nal constraints and forces applied iteratively on the To achieve the stated purpose, the global setup (Fig-
inflated membrane. Those local stiffening and mod- ure 2) consists of:
ulations would participate in the intricacy of the final
design that would allow creases and wrinkles.
• An extensible membrane (e.g. latex) clamped
Contrary to established methods where a given
to a steel frame (1x1 meter) that is fixed on a
form is meant to be perfectly approximated (Ant
vertical turntable KPF1-V500.
Farm 1971), those local deformations are not seen as
• An external 6-axis industrial robot arm KUKA
negative outputs but as positive ones, as they fully
KR 125/2 equipped with a carbon fiber ex-
participate in the formation process (Menges 2010).
truder as an end-effector that helps to modify
The form would be modulated and generated until
the stiffness of the inflated membrane.
it finds an equilibrium state between design targets
• A scanning device (Microsoft Kinect V1)
(set by the user at the beginning of the process) and
mounted near the rotating table that will
structural requirements and performances. We can
record and analyze in the current shape
speak here of a "steering of form" (Kilian 2014), a con-
adopted by the membrane.
cept introduced recently that argues for digital simu-
• An air compressor (connected to the steel

Figure 2
The global setup for
further digital
fabrication and
actuation of the
inflated membrane.
frame) that can be controlled digitally from a the fibers in tension.

user interface. During the time of reinforcement, the setup ro-
• A user interface developed within the tates according to the robot movement in order to lay
Grasshopper plug-in for Rhinoceros 3D - com- a carbon fiber thread along a path created between
puter graphics and computer-aided design two hooking points placed along the steel frame (Fig-
(CAD) application software - that integrates ure 5). This particular method has been adapted from
both the simulation and generation of robotic the fabrication process of the structural components
paths, scanned data coming from the Kinect of the ICD/ITKE Research Pavilion 2013-14 (Dörstel-
sensor and structural analysis performed by mann et al. 2014).
the Grasshopper add-in Karamba3d. Once the current shape is entirely reconstructed
by the external scanning device (using the com-
mercial software ReconstructMe and its Software
Iteratively, the membrane is inflated. Each itera-
Development Kit (SDK)) and further digitally post-
tion is followed by a local stiffening and reinforce-
processed, it is possible to run a structural analysis,
ment achieved through the placement of carbon
which considers the physical scanned membrane in
fiber threads helped by the robot arm and its end-
its final state as a self-supporting shell structure. It
effector. The latter is composed by a spool and two
is important to mention that during the fabrication
hooks mounted on a steel bracket that helps to keep

Figure 3
Comparison
between the digital
simulations (left)
and the material
experiments (right).
the membrane surface is under tension - the inverse 1964) and the Binishells (Bini 1978) were the first pro-
structural state of the final shell. If it fulfills both totypes that used shotcrete techniques for convert-
structural performance and design targets asked by ing tension based inflatable structures into compres-
the user, the process stops. If not, a new reinforce- sion based thin-shells.
ment path is created and a new specific pressure is re-
leased. This form generation process continues until MATERIAL EXPERIMENTS AND DIGITAL
all the conditions (spatial and structural) are fulfilled
SIMULATIONS
(Figure 7).
In order to calibrate the system, both digital simula-
Once the form generation process stops, the
tions and physical experiments have been made.
rigidification process starts. The latter consists of
Initial physical experiments (Figure 3 - right)
spraying chopped glass and/or carbon fibers impreg-
only involved material investigations which aimed to
nated with resin on the final physical shape. In the
prove the ability of the latex to shape itself under pre-
context of the present research, the spraying process
determined conditions (local stiffening) and to serve
has been achieved manually. However, the same
as a formwork during inflation. A small clamping de-
robot mounted with a different and suitable effec-
vice has been used to conduct those experiments
tor could realize the same operation. When the resin
(Figure 4).
dries, the inflated membrane starts working as a thin-
Parallel digital simulations (Figure 3 - left) proved
shell under compression.
the potential of representing precisely the behavior
Existing precedents already proved the ability of
of inflated latex membranes under predetermined
inflatable structures to work as compression shells af-
conditions. In order to achieve this, the Grasshopper
ter rigidification processes. The Bubble Houses (Neff
add-in Kangaroo (Piker 2013) has been used to set-

Figure 4
Physical setup for
small scale
experiments.
up differentiated stiffening strategies. First, the de- • Laying tape locally on top of the membrane.
sired surface to inflate is discretized into a square grid • Laying carbon fiber thread locally on top of
with each cell containing specific information related the membrane.
to its local stiffening condition (Otto 1983). A stiffness
gradient is then applied through all the mesh and is Laying carbon fiber threads directly on top of the
modulated with multiple attractor points. This same extensible membrane allows a direct, quick and very
gradient is partitioned through a K-Means clustering efficient differentiation of the material compared to
algorithm which outputs data for further fabrication the gluing processes that need long curing time in
strategy - in particular for the differentiated layering order to reiterate the process. Also, the third option
technique (Figure 3). caused a lot of problems related to the friction be-
In order to perform physical local stiffening, dif- tween the tape and the membrane, which often led
ferent methods have been used, such as: the latter to break. Thus, the last option - laying car-
bon fiber thread locally on top of the membrane -
• Spraying glue on top of the membrane. showed the best potential to differentiate iteratively
• Differentiated layering technique (Figure 3). and locally the extensible membrane. It has been
The different layers have been glued together chosen and used in the following experiments (Fig-
using latex gel. ures 5-6).

Figure 5
Carbon fibers are
robotically placed
on top of the
inflated extensible
membrane.

Figure 6
Fabrication process
of the final
prototype (left) and
the corresponding
rigidified shell with
its scanned data
alignment (right).
Figure 7 3D SCANNING AND FEEDBACK LOOP

Flowchart From the initial experiments that investigated only
describing the material performances, new prototypes have been
overall form fabricated involving 3D scanning, incremental infla-
generation process. tion and robotically controlled stiffening processes
- without implementing any feedback strategy yet.
The step-by-step process unfolds as follows:
• First, an initial inflation activates and tensions

the membrane.
• Then, a predetermined path actuates and
modifies its current state.
• This path is reinforced until the membrane
fully stabilizes under the additional tension
exerted by the carbon fiber threads.
• The process presented here can be repeated
once again along another predetermined
path. During this process, the user could
choose when to increase the pressure and the
duration of inflation.
Further experiments have partially implemented the

feedback-loop strategy within the process. Before
starting the latter, the user can assign a design tar-
get in space (Figure 8) that the membrane needs to
reach in order to achieve the desired spatial require-
ment. After each iteration, the Kinect is automatically
triggered and the turn-table - on which the mem-
brane is standing - rotates 360 degrees. The shape
of the membrane is then digitally reconstructed and
the resulted mesh object can be used and controlled
against two parameters: the initial design target

and structural requirements (self-supporting struc- the membrane (Figure 6). It has been used for the
ture under its own compression loads). If both are in next experiments involving 3D scanning and feed-
accordance with the expectations of the user and the back strategies that enabled both the user and the
machine, the process stops. If not, it continues until machine to reiterate the process after analysis.
all conditions are fulfilled and the membrane finds its From the different methods described above, it
equilibrium state. has been possible to achieve successful prototypes
and a 1x1 meter rigidified thin shell initially con-
strained with carbon fiber threads (Figure 6 - right). Figure 8
These prototypes proved the feasibility of achieving The membrane is
self-supporting structures generated within a cyber- iteratively
physical environment through iterative stiffening, in- constrained in
flation and scanning processes. order to reach a
design target set by
Discussion the user at the
The complexity of physical post-processes in tradi- beginning of the
tional design methods for inflatable structures can process.
be replaced by a cyber-physical setup that enables
the user to iteratively interact and robotically modu-
late the physical membrane until it fulfills design and
structural requirements.
The feedback loop strategy explained in this pa-
per could be improved by letting the machine make
more autonomous decisions when scanning analysis,
fiber laying or inflation processes need to be oper-
ated. The user would still be able to monitor or mod-
Man-machine interaction
ify the actions taken by the machine if the latter does
The role of the user is to guide the robot actions dur-
not operate as one would expect.
ing the fabrication process so the latter can achieve
the design intentions dictated by the former. A de-
cision made by the machine based on an external
Outlook and architectural applications
The experiments featured in this paper do not scale
signal coming from the scanned data generated by
up beyond the boundary reach of the robot-arm.
the Kinect might be contradictory with the user's ex-
One could solve this problem by using mini-robots or
pectations. Therefore, the designer is able to inter-
even drones that would operate in a similar manner
vene in the form generation process and to overwrite
(Figure 9). Successful architectural research projects
the next generated fiber path by an alternative action
have shown potential for fabricating rope bridges us-
(e.g. path correction or new pressure input). Thus,
ing flying machines (Augugliaro et al. 2014). It is most
the user is involved in the form generation process
feasible than such techniques could be investigated
where he negotiates directly with the machine each
further in order to interact directly with dynamic ma-
action for further actuation (Negroponte 1969).
terial behaviors and processes.
Another type of architectural application might
RESULTS be component-based, where each module could be
From the material experiments investigated in this re- fabricated within the design framework described
search, the option using carbon fiber threads showed above, before being aggregated together in order to
the best potential to iteratively differentiate locally produce larger spatial structures.

Figure 9
Expanding the
cyber-physical
setup described in
this paper by using
drones in the form
generation process
of an inflatable
structure.
This paper presented the potential for an au- Augugliaro, FA, Lupashin, SL, Hamer, MH, Male, CM,
tonomous cyber-physical fabrication setup that can Henn, MH, Mueller, MWM, Willmann, JSW, Gramazio,
generate differentiated pneumatic membranes and FG, Kohler, MK and Andrea, RA 2014, 'The Flight As-
sembled Architecture Installation: Cooperative con-
analyze it during the process until the physical shape
struction with flying machines', IEEE Control Systems
satisfies both design and structural conditions. The Magazine, 34(4), pp. 46-64
membrane is further rigidified and finally acts as a Bini, DB and Morelli, GM 1978 'Cassaforme Pneumatiche
thin-shell structure. per la Costruzione di Cupole in Cemento Armato',
L'Industria Italiana del Cemento. Anno XLVIII, Giugno
Acknowledgement 1978
Dörstelmann, MD, Parascho, SP, Prado, MP, Achim
This paper presented the research development of
Menges, AM and Jan Knippers, JK 2014 'Integrative
a master thesis project realized within the Integra- computational design methodologies for modular
tive Technologies and Architectural Design Research architectural fiber composite morphologies', Pro-
(ITECH) master program at the University of Stuttgart. ceedings of ACADIA 2014 Design Agency
Ant Farm, AF 1971, Inflatocookbook, Ant Farm, Sausalito

Kilian, AK 2014, 'Steering of form', in Adriaensses, SA,
Block, PB, Veenendaal, DV and Williams, CW (eds)
2014, Shell Structures for Architecture, Form Finding
and Optimization, Routledge
Menges, AM 2010, 'Form Generation and Materialization
at the Transition from Computer-aided to Compu-
tational Design', DETAIL, Review of Architecture and
Construction Details, 4, pp. 330-335
Menges, AM 2015, 'The New Cyber-Physical Making in
Architecture, Computational Construction', Architec-
tural Design, 85(5), pp. 28-33
Neff, WN 1964, Architecture of Southern California: A Se-
lection of Photographs, Plans, and Scale Details from
the Work of Wallace Neff, Rand McNally, Skokie, Illi-
nois, United States
Negroponte, NN 1969, 'Toward a Theory of Architecture
Machines', Journal of Architectural Education (1947-
1974), 23, pp. 9-12
Otto, FO 1979, Growing and Dividing Pneus, Institut für
leichte Flächentragwerke (IL), Stuttgart
Otto, FO 1983, Air Hall Handbook, Institut für leichte
Flächentragwerke, Universität Stuttgart
Piker, DP 2013, 'Kangaroo: Form finding with computa-
tional physics', Architectural Design, 83, pp. 136-137

MATERIAL STUDIES | Concepts
Model Translations
Studies of translations between physical and digital architectural models
Ashish Mohite1 , Toni Kotnik2

1,2
Aalto University
1,2
{ashish.mohite|toni.kotnik}@aalto.fi
With the rise of the digital in architecture and the availability of digital
fabrication tools, the interest in the material aspect of the model has intensified.
At the same time, the design space for exploration of material behavior and its
design potential has been extended from the physical into the digital. This has
resulted in a cyclic set of translations from the physical realm into the digital by
means of mathematical descriptions and back from the digital realm into the
physical by means of digitally controlled fabrication processes. Despite the
availability of more and more computational power and improvement of precision
in simulation, these translations from the physical into the digital and vice versa
can never be exact (Eco 2006), the translations from the physical model into a
digital model and from the digital into the physical are "spaces of instability"
(Evans 2000). The current paper explores in more detail this space of instability
between physical and digital models, its potential for architectural design, and the
central role of the mathematical description in this reciprocal set of translations.
Keywords: Architectural model, simulation, digital fabrication, material

computation, material behavior
Cardboard physical model turns into a digital mesh or the world, its geometry and structure, its physicality
a NURBS surface that is then 3D printed in a myriad of is continuously augmented by the necessary adapta-
plastic filaments, stuck together or formed out of liq- tions of the 'same' characteristics to different realms.
uid resin, which is exposed to UV light in order to so- Models are used to test concept, geometry,
lidify. And finally the building is built out of concrete structure, they do not merely represent what archi-
and steel. The translation from physical to digital to tecture could be or explain an idea or subjugate to
process to actual edifice means placing avatars of the buildings as by-products (Abruzzo et al. 2007). They
same model in different domains. Inevitably discrep- are often instrumental and generative and contribute
ancies arise. It occurs due to a variety of factors such enormously to the design process. So when a model
as physical conditions and requirements of the space plays an active part, its materiality (chosen material,
in which model is situated, or intended to be situated its properties and effects as well as physical forces in-
as well as material and its properties, actual and sim- fluencing model's behaviour) plays an active part too.
ulated. The way a model behaves and interacts with However, material qualities and their effects are often
MATERIAL STUDIES | Concepts - Volume 1 - eCAADe 34 | 561

an after-thought and are rarely a catalyst for geome- models, in his experiments the model's materiality in-
try. According to Lars Spuybroek (2010) architectural formed the form of the actual geometry. Gaudi, in
models are pure abstraction of forms, independent of his quest to find the structural arch, made a model
material notion. Nevertheless, in past years there has out of hundreds of threads weighted with sandbags;
been a rise of interest in architectural techniques and carefully choreographed interdependency between
production that focus on material performance. Ar- weight, gravity, and tensile strength generated the
chitectural discipline has once again started to form approximate form of arches of Sagrada Familia. How-
a direct relationship with materials (Schröpfer 2011). ever, actual buildings are constructed in completely
In both physical and digital models and their different materials, so, once the form was informed
translations into architectural scale there are inter- by specific make-up of a specific material, the ini-
ruptions in terms of proportions, structure, and for- tial material was abandoned and it's qualities force-
mal geometric performance. This paper seeks to find fully imposed on a different one. Another problem
out how material performance affects geometry and lies in that a chain model can perfectly account for a
structure and what is lost and gained in this process multidimensional gravity field, but not for many lat-
of adaptation from physical understandings of mate- eral forces that occur in any building. There is no
rial properties to digital simulation and back to phys- documented account of exact translation between
ical material geometry through digital fabrication. Gaudi's models and Sagrada Familia's actual geom-
etry, but it is rather safe to assume, that it was not a
Finding form and structure
direct and easy one. Frei Otto was also interested in
"Whether labelled material form-finding, material
arriving at form through structural simulation mod-
self-organisation or material computation, designers
els (Erdinger 2005). "Most of Otto´s analog machines
such as Gaudi, Otto, Isler and others have developed
consist of materials that process forces by transfor-
approaches that instrumentalised the complex re-
mation, which is a special form of analog comput-
lationships between top-down (form) and bottom-
ing" (Spuybroek 2010). Otto, in his form finding exer-
up (material properties) within the design process"
cise of three-dimensional cancellous bone structure
(Nicholas 2012).
and branching column systems, used a wool-water
Let us look at historical examples of the relation-
technique. This involved using tension caused by
ship between a model, its materiality and ensuing
gravity between woollen threads, dipped in water.
form and structure. For centuries builders used to
In both cases the understanding of material prop-
make material models to test structure and for form
erties leads to the development of form. In case of
finding. One of the central structural problems were
Mannheim Multihalle, a wire mesh hanging model
compression systems, such as arches. Before 18th
was made for form-finding by Otto, whereas struc-
century they had to be developed through trial and
tural analysis models were developed separately by
error of successive scale models. In the 18th cen-
engineers using different materials and techniques,
tury due to the interest in this problem of mathe-
including a digital model. As final structure was
maticians, such as Leibnitz, the similarity of compres-
made out of wood, which changes its behaviour de-
sion systems to tension systems became apparent.
pending on scale and exact composition, a linger-
That allowed to test arches through cable models.
ing doubt in Multihalle's stability remained until af-
The first such model was made by Giovanni Poleni in
ter it had been constructed. Engineers designed a
1748 when he was asked to analyse the structural in-
test that consisted of hanging multiple municipal
tegrity of St.Peter's dome in Rome (Addis 2015). In
garbage bins, filled with water, onto the structure.
the 20th century Antonio Gaudi continued the work
Multihalle bowed less than was expected and it is still
with cable models, but approached it from a differ-
standing.
ent angle. He was finding form through material
562 | eCAADe 34 - MATERIAL STUDIES | Concepts - Volume 1

Model as driver for both geometry and structure can with reference to the complexity of a building project
be found in works of Heinz Isler and Sergio Musmeci. and to the way the building integrates into its sur-
Isler was the first one to build structures based on sus- roundings and into society" (Otto 1990).
pension method, where forms for shells are found by
hanging and fixating cloth. During the early part of The question of scale
his career, simulation programs capable of analyzing Scalability of the structural performance still remains
his shells were not existent. Hence, he used physical a problem in both physical and digital models and
models to prove the structural robustness and stabil- this can be attributed to the difficulty of scaling phys-
ity of his forms (Kotnik and Schwartz 2011). He first ical properties in and out. In pure analog models
built models for form-finding and then, several itera- the limitations of the techniques rarely have impli-
tions of the models were made to test lateral stresses, cation on the structural performance but can have
and finally, the model that passed all his tests would real implication on the proportions at architectural
become a building. Musmeci´s work was a quest for scale. Historically, this problem of scaling was one of
unity between structure and form, which led to im- the main obstacles in the activation of form for struc-
mense amount of experiments, designed to find an tural purposes. The results of experimental methods
integral structural form. His main focus was mould- of form-finding could not just simply be scaled up to
able concrete surface structures, he strived to test provide proof of structural soundness under various
plastic potential of concrete while minimising mate- load-conditions. And the underlying mathematical
rial and optimising the form. His approach can be description of the form often was to complex to be
summarized in his own words "the form is the un- handled by hand-calculation (Billington 1980). It was
known, not the inner stresses" (Musmeci 1979). His this situation that in the 1930s motivated the found-
form-finding models were based on soap film being of institutes with research focused on the ques-
haviour and stretched textile membranes. In case of tion of scalability of structural behavior from model
Basento Bridge he built final physical model in con- scale to building scale like for example the Instituto
crete, one tenth the size of the actual bridge in order Técnico de la Construcción y Edificación in Madrid by
to test his system in authentic material and at a rela- Eduardo Torroja in 1934 or the the Model and Con-
tively close to final scale. struction Testing Laboratory by Arturo Danusso at
In terms of expansion of the space of an object, the Politecnico di Milano in 1935, a lab that worked
once it starts manifesting in different media, the most in close collaboration with Pier Luigi Nervi (Olom and
significant occurs when multiple factors come into Chiorino 2010). In the same way, Frei Otto's famous
play. When a designer is concerned with creating a Institut für leichte Flächentragwerke at the Univer-
structure that performs in a specific way in a specific sity of Stuttgart was not only exploring self-forming
context, rigorous modelling, as in case of Musmeci, processes in material systems but also developing ex-
is likely to achieve the goal without much significant pertise in scaling up of the observed phenomena to
(for the goal) deviation between models and build- building scale. An expertise that was crucial in the
ing. However, once the spectre of demands that a realization of the roof of the Olympic Stadium in Mu-
building must fulfil grows, so does the deviation, re- nich in 1971. This project is one of the last that de-
sulting in compromise and augmentation of initial pended still to a large extent on the use of physical
design. "It is extremely difficult to carry out architec- models for the validation of the structural behavior
tural design with the self-formation processes. The of the roof. In the 1970s, more and more computa-
experiment does indeed lead directly to the form, tional methods took over the task of structural vali-
which in itself has already passed through an opti- dation and replaced the labor-intense use of physical
mization model, but a design work can only be seen scale models.

Nowadays, when digital model is a dominant form Galilei, the essential characteristic of the mathemat-
of modelling, the issues of scale have been some- ical description of natural phenomena is the reduc-
what alleviated due to the capacity of digital soft- tion of the phenomenon to measurable sizes such
ware to imitate a multitude of parameters at a mul- as lengths, surfaces, volumes, or temporal distances-
titude of scales simultaneously. Nevertheless, some- that is, the quantification of the phenomena (Kotnik
times the simulation is limited or desired architec- 2011).
tural performance comes into conflict with structural
behaviour. Sometimes the geometry of a digital, Figure 1
weightless model is forced onto reality. Construction
(Figure 1) serves as an instance of that. The desire picture of Metro
to literally translate laser cut scale model into archi- Parasol [1], Jürgen
tectural dimensions resulted in the necessity of ad- Mayer H, 2011.
ditional load bearing structure underneath the sur-
face. In scale models the 3D surface was structural,
in blown-up version it is a skin. In case of (Figure
2), metal panes are a direct reproduction of surface
simulation in CAD model. This representation, when
translated into architectural scale, creates an exact
replica visually. In computer simulation, the pattern
was designed by algorithmic calculations to reduce Figure 2
consumption of memory and time and had structural Dongdaemun
implications, at architectural scale, it increases time design plaza [2],
and consummation of material and has no real struc- Zaha Hadid
tural consequence. So, in terms of structural perfor- Architects, 2014.
mance and cost-efficiency it could be argued that the
translation from model to building resulted in a com-
promise. On the other hand, polygon modelling that
was used to create double curved surface allowed
for geometric precision due to variable size and den-
sity of panels based on surface curvature. As a re-
sult, Dongdaemun is architecturally an example of
gain, where digital modelling made possible the ex- By using this abstracting approach he was able to
act form that designer envisioned. express the assumed causality between cause and
effect in an idealized form by means of a mathe-
Digital simulation and fabrication matical function. The logical combination of math-
The process of adaptation between the digital and ematical descriptions that had been inductively at-
the physical has been a driver for development since tained allowed him, furthermore, to draw conclu-
the Renaissance and the work of Galileo Galilei. sions and make prognoses about the more complex
Driven by the idea of a mathematically ordered inter-relationships in nature. What Galilei started to
world, Galilei replaced the qualitative explanation establish is a process of simulation of reality by means
for the necessity of a specific effect from a given of a formal description that lies at the heart of our
cause, which had been common until then, by the contemporary ability to blur the digital and the phys-
quantitative ascertainment of the phenomenon. For ical (Lorenzo-Eiroa and Sprecher 2013).

"The development of digital design is often pre- ical could be used to discover ingenious solutions
sented as a threat to one of architecture's essen- to geometric and structural problems. "Computer-
tial dimensions: the concrete aspects of construction aided material production seems to abolish the dis-
and building technologies, in a word, its materiality" tance between representation and materiality, pro-
(Frampton 1995). In the past, computer-based de- vided that one defines materiality in other terms than
sign used to appear to neglect the materiality and its traditional tectonics" (Picon 2004).
relationship with architecture. Properties like gravity, Technological advances promise new horizons:
´digital fabrication allows to control manufacturing
Figure 3 through design data. Material is thus enriched by in-
Remote material formation; material becomes " informed" '(Gramazio
deposition [3], and Kohler 2008). If that is true, we should already
Gramazio Kohler be able to design materials, to deconstruct their logic
Research, 2014. and properties and then reconstruct them. It ap-
pears, looking at the work of Gramazio and Kohler
themselves that a different phenomenon could be
taking place. Namely, that information gets 'in-
formed' by material, its properties and behaviour af-
fect design process.
In (Figure 3) Gramazio and Kohler tried to inves-
tigate the translation between the digital simulation
Figure 4 and materiality by designing the path along which
Remote material foam material was projected from a distance. They
deposition attained an architectural form out of this process, but
installation rigid properties of foam and its light weight were not
[4],Gramazio Kohler sufficiently affected by the force of projection. With
Research, 2014. this experiment they interestingly take up the ini-
tial experiment of Galilei in which he deduced the
parabolic flight path of an object based on the de-
ductive interlocking of simple laws of motion. Galilei
used this example of the flight path to demonstrate
the power of his method and the superiority of the
mathematical description over reality. A superiority
active force, densities are immanent to materiality which is questioned by the 'remote material deposi-
of architecture, however, in the earlier days of dig- tion'.
ital design forms floated freely in the digital space, In (Figure 4) they took the fabrication to ar-
without any constraint. The only limitations were chitectural scale, changed the material to claylike
software capacities and imagination. In the current slurry, added more gravity by increasing distance and
age of digital design, forms can still float in the vir- achieved a complex and dynamic deposited form.
tual space, yet they are programmed to respond to Viscosity of claylike material played an important role
simulated constrains of gravity, weight, force, scale, in the experiment. In both experiments the process
and material. By coming close to replicating the of amassing was judiciously designed and was in-
real world constrains 3D software approaches the formed by two different materials, which then were
ability to redefine materiality and not just imitate it. aggregated into two topologically adjacent, yet dif-
The process of adaptation between digital and phys-

Figure 5
Monolith, Andrew
Atwood, 2012.
ferent structures. The discrepancy came from mate- ceptable in the framework. Glitches occurring in the
rial properties. In order to achieve a comparable re- printing process were recorded, studied and brought
sult with a different material they had to change the under control in order to replicate the desired ge-
parameters of system. While both models were geo- ometry and be able to work with it further. In both
metrically similar, architecturally they created a gra- Gamazio's and Atwood's cases materials drove the
dient from an unstable, light, unitary structure to a apparatus even though in one instance the objec-
wall-like, heavy and stable formation. tive was to automate building process and in an-
Another example of utilising specific material other it was a form-finding exercise and an attempt
properties in conjunction with specific technique is to harness the unexpected by-product of rapid pro-
Monolithic process studies project by Andrew At- totyping. The difference between Gramazio's digital
wood. He worked with plastic deposition, which is model and both physical models is architecturally im-
the most common technique in 3D printing. Some- mense even though all three of them perform exactly
times during printing a glitch occurs and printer the same function. Atwood's process focused on the
pauses while plastic continues to pour out, thus cre- geometrical difference between initial digital model
ating thin strands stemming from the body of the and its erroneous 3D printed version, which became
model. That imperfection produced fascinating ge- the goal.
ometry and motivated Atwood to try and replicate
it in a controlled manner. In his studies (Figure 5) Experiments with materiality
Atwood manipulated calibration of material control, According to Sanford Kwinter (1993), form is an in-
fabrication precision and process fidelity through a stance of structural stability in a system as it seeks
formal technique of offsetting and material process homeostasis, thus all form is the result of growth
of layering of plastic filaments on top of each other and resistance, it is the convergence of material and
layer by layer. Atwood designed a custom simu- force. We set to explore the implications of materi-
lation engine for the purpose of modelling the ex- ality on reciprocal translations from digital to phys-
periment. This simulation was able to predict the ical. There is not a fully fledged hypothesis yet, as
material successes and failures within tolerances ac- experimentation is in early stages. What could po-
tentially be the focus is a technique, involving usage

of performance logic of a specific material and trans- its reverse engineering (Figure10, Figure 11), which
fer of that logic onto another material. The tentative shows the potential of control over such geome-
objective of such material studies would be arriving tries. The next phase of the project will be ad-
at a pattern of interdependence between materiality dressing the way materiality affects structural perfor-
and data(simulation) process. Each following exper- mance of a model. Once both experiments geared
iment operates at a small scale, with a yet unknown towards form-finding and the ones concentrating on
potential at architectural scale. All experiments con- structure-finding are carried out and evaluated, we
sist of several stages: digital modelling, import of intent to start applying the results to other materials
the digital model into fabrication simulation, addi- and processes.
tive manufacturing. At each stage specific parame-
ters are changed in order to learn the interdependen- Conclusion
cies. Specifically, the viscosity of the material, depo- This translation from digital simulation to digital fab-
sition layer heights, speed of fabrication, the ratio of rication simulation to digital fabrication and back will
bridge speed to print speed. Material employed is ce- be documented in diagrams, physical models, and
ramic and method of fabrication is layer deposition, digitally fabricated models at various scales using an
using a modified plastic printer as a universal paste array of materials. The essential goal is to study the
extruder. The first phase was to record the deviations translational transformations in order to, on the one
and try and discern the parameters and their combi- hand, gain more control over material behavior and
nations that led to them. The second phase consisted conditions of the physical environment, and, on the
of reverse engineering the first phase and exaggerat- other hand, allow for materiality to exercise creative
ing the determining factors in order to confirm and influence on the design process. This dual action and
establish further the pattern of interdependency. its consequences have the potential of being an aide
and a tool for designers. A final dissertation will orga-
Figure 6 nize and analyze the results of experiments and sim-
Viscosity, Ashish ulations and attempt to build a comprehensive ap-
Mohite. paratus that will enable fluent translations between
various modelling realms and offer more control over
new design processes.
Instead of aiming for increasing complexity and
precision in the description of material behavior,
therefore, the topological nature of the relationship
between the digital and the physical could be used
As this research is just beginning, the set of experi-
to simplify the geometric description and reduce the
ments is limited to studying the correlation between
amount of information necessary for control. This pa-
geometry, a single material (ceramic) and a single
per can be seen as a first attempt towards an under-
fabrication technique (paste extrusion). What they
standing of the digital as a reduced set of essential in-
show, however, is that there is a way to design a spe-
formation that is used to govern the design develop-
cific geometric effect by manipulating not the geom-
ment already at the conceptual level. Material is used
etry itself, but the parameters defining its translation
as an active agent that is not fully controlled but acti-
into physical domain. Such geometric articulation as
vated and is allowed to compute its position in space
achieved in Speed of bridging (Figure 9) or Viscosity
through local interaction. Such diagrammatic use of
(Figure 6) experiments is hard to model digitally and
the digital could enhance design freedom and flexi-
even harder to control.
bility.
Analysis of Speed of bridging experiment led to

Figure 7
Speed of
deposition, Ashish
Mohite.
Figure 8
Scale in fabrication
simulation, Ashish
Mohite.

Figure 9
Speed of bridging,
Ashish Mohite.
Figure 10
Reverse engineered
geometry 01,
Ashish Mohite.
Figure 11
Reverse engineered
geometry 02,
Ashish Mohite.

REFERENCES ture by Materiality, Birkhäuser
Abruzzo, E, Ellingsen, E and Solomon, J 2007, 'Models', Spuybroek, L 2008, The Architecture of continuity, NAi
306090, 11, pp. 102-107 publishers, Rotterdam
Addis, B 2015, Building: 3,000 Years of Design, Engineer- Spuybroek, L 2011, 'The matter of ornament', in Spuy-
ing, and Construction, Phaidon Press broek, L (eds) 2011, The Politics of the Impure, NAi
Atwood, A 2012, 'Monolithic Representations', in Mered- publisher, Rotterdam, p. 242
ith, M and Borden, G (eds) 2012, Matter: Material Pro- Torroja, Eduardo 1958, Eduardo Torroja: The Structures of
cesses in Architectural Production, Routledge, New Ediardo Torroja, Dodge Cooperation
York, p. 208 [1] http://buildipedia.com/images/masterformat/Chan
Billington, D 1980, Heinz Isler as Structural Artist, Art Mu- nels/Go_Green/2011.08.16_metropol_parasol/sit
seum Princeton University e_images_credit_j_mayer_h/parasol_jmh_site_im
Eco, U 1994, How to Travel with a Salmon and Other Essays, age_02.jpg
Martin Secker & Warburg, London [2] http://i1.wp.com/aasarchitecture.com/wp-content
Erdinger, W (eds) 2005, Frei Otto - Complete Works, /uploads/2014/03/Open-Dongdaemun-Design-
Birkhäuser Park-
Evans, R 2000, The Projective Cast: Architecture and Its Plaza-by-Zaha-Hadid-08.jpg
Three Geometries, MIT Press [3] http://gramaziokohler.arch.ethz.ch/web/e/lehre/
Frampton, K 1995, Studies in tectonic culture: The poet- 277.html
ics of construction in nineteenth and twentieth century [4] http://gramaziokohler.arch.ethz.ch/web/e/lehre/
architecture, MIT Press, Cambridge, MA 276.html
Garber, R 2014, BIM Design: Realising the Creative Poten-
tial of Building Information Modelling, Wiley
Gramazio, F and Kohler, M 2012, 'Digital materiality in ar-
chitecture', Space, 537, pp. 100-107
Kotnik, T 2011, 'Experiment as design method: On the
possibility of the integration of the methodology of
the natural sciences in architecture', in Moravánszky,
AK and Kirchengast, AL (eds) 2011, Experiments: Ar-
chitecture Between Sciences and the Arts, Jovis, pp.
24-53
Kotnik, T and Schwartz, J 2011, 'The Architecture of Heinz
Isler', Journal of the IASS, 51(3), pp. 185-190
Kwinter, S 1993, 'Landscape of Change:Boccioni´s ´Stati
dánimoás General Theory of Models', Assemblage,
19, pp. 50-65
Lorenzo-Eiroa, P and Sprecher, A 2013, On the nature of
information in digital architecture, Routledge
Musmeci, S 1979, 'Le tensioni non sono incognite',
Parametro, 80, pp. 36-46
Niccholas, P (eds) 2012, Persistent modelling:Extending
the role of architectural representation, Routledge,
Abington
Olom, C and Chiorino, CH (eds) 2010, Pier Luigi Nervi: Ar-
chitecture as Challenge, Silvana Editoriale
Otto, F 1990, 'Experimente', Bulletin of the Institut für le-
ichte Flächentragwerke, 25, p. 5
Picon, A 2004, 'Architecture and the virtual: towards a
new materiality', Praxis: Journal of Writing+ Building,
6, pp. 114-121
Schröpfer, TH 2010, Material Design: Informing Architec-

Knitflatable Architecture
Pneumatically Activated Preprogrammed Knitted Textiles
Yuliya Baranovskaya1 , Marshall Prado2 , Moritz Dörstelmann3 ,

Achim Menges4
1,2,3,4
Institute of Computational Design (ICD), University of Stuttgart
2,3,4
http://icd.uni-stuttgart.de
1
yuliya.baranovskaya.itech@gmail.com
2,3,4
{marshall.prado|moritz.doerstelmann|achim.menges}@icd.uni-stuttgart.de
Textiles are widely used in architecture for tensile structures, as they are
lightweight and can easily span large distances. These structures typically
require an external framework for a support. Inflatable structures are
self-supporting but are limited to relatively simple forms or require complex and
predetermined cut patterns. The development of an adaptive and programmable
textile system with an integrative method for pneumatic activation would create a
novel self-supporting structure with high degree of design and architectural
potential. This creates a highly integrative hybrid system where the generic
pneumatic membranes are constrained by the differentiated knitted textile skin
that is stretched in several directions under air pressure. This allows for an
innovative, lightweight, easily transportable design, where the preprogrammed
knitting pattern defines the structure, geometry and formation, activated under
pneumatic pressure.
Keywords: programming textiles, binary textiles, analogue computing, air

inflation, grading textile properties
INTRODUCTION ates the potential for manufacturing at an architec-

Textile sheet materials can be divided into two large tural scale. Current industrial knitting technologies
groups based on manufacturing processes. Either are programmable and provide a high level of con-
they are made with a single filament of a continu- trol over the production.
ous length (e.g knitting) or with sets of filaments of a Research in the field of knit-activated textiles
limited length (e.g weaving) (Seiler-Baldinger 1979). is rather extensive in fields other than architecture.
This research focuses on knitting techniques due to There are already some highly developed projects
the several key advantages. Zero waste fabrication which challenge the boundaries of knitting as a per-
and re-usability of filament makes it sustainable tech- formative smart technique. Nike Flyknit shoes are an
nology, fiber can be extended infinitely, which cre- example of a knitting in a single continuous process

for commercial production. The differentiation of the differential forces, whilst allowing for a much higher
shoe textile is made according to the foot perfor- level of the surface differentiation (light transmitting,
mance during running: predefined areas of indura- texture, tactile experience). By understanding the
tions and elasticity as well as details for lace inser- limitations of pneumatic and knitting systems and
tion. The foot activates this textile to adapt and re- combining the advantages of both, the new hybrid
spond in various performative ways. The seamless system is introduced. In this scenario, the knitted ma-
nature of knitting technique resolves potential con- terial is used in tension as a surface and shape defin-
nection complications in a smooth transition of a fil- ing structure, preprogrammed through the manufac-
ament topology. Knitted textiles unlike solid materi- turing process, and actuated by an aggregation of
als are capable of re-adjusting and stretching due to generic pneumatics that are simultaneously inflated.
the loop based structure, which has inherently high
tolerance for assembly. Earlier mentioned continu- CONTINUOUS FILAMENT
ity of the fabrication makes possible to produce el- Although there is a wide diversity of filament materi-
ements with the endless length and introduce knit- als with varying performative characteristics, the re-
ting on an architectural scale as shown in the instal- search presented here focusses on the fiber arrange-
lation commissioned by Nike Flyknit Collective (Sabin ment and distribution, rather than on its material
2013) Textile Hybrid Tower by CITA (Thomsen 2015) properties. The system could be framed within Mas-
or KnitWit textile canopy build at EASA 2015 (Bara- ter Framework for Spatial Design (Vrouve 2015) as
novskaya et al 2015). It is possible to control the ma- surface-active structural system made out of material
terial property distribution through access to tech- of a the rod section profile (filament) with the biaxial
niques and digital technologies, even on the level of material orientation in the outer skin and amorphous
domestic craft tools (Knitectonics 2013). Now the ar- arrangement within the internal structure (latex ex-
chitectural material becomes a library of storing data pandable membranes). Containing knitted elements
and performing data (Thomsen 2016). Continuously are stitched together in order to accommodate mul-
developing textile producing machinery allows for tiple inflatable aggregates. Fiber properties are as-
the manufacturing of highly performative materials sumed to be relatively non-elastic, and 100% poly-
with the integration of various properties within one acrilyc yarn is being used.
piece. There is no longer any need to change ev-
ery component in order to achieve various perfor- Figure 1
mances within the single system (Scott 2013). Hier- Inflatable units
archical relations within the systems are becoming non-restrained (A.),
a solution for multi-performative hybrid structures restrained by solid
(Ahlquist and Menges 2013). boundary (B.) and
Inflatable architecture is known for its generous by soft boundary
scale to material proportion, as well as for deploya- with differentiated
bility. However, in order to achieve architectural de- surface properties
signs, beyond that of simple typologies, a specific cut (C.).
pattern is required as well as an understanding of the
membrane material performance to accommodate
differential stresses which occur during inflation. In
contrast, knitted fabrics can be programmed to adapt
to a variety of geometric applications with heteroge-
neous material arrangement capable of dealing with

Figure 2
Abstraction of the
loop-based knitted
mesh into
quad-based mesh.
SOFT BOUNDARIES ble boundary provides more freedom for the inflata-
Behavior of the individual inflatable unit under an air bles to rearrange in space and affect the global shape.
pressure could be described as an external expan- Both systems have an interdependent reciprocal re-
sion: latex material due to its isotropy and chemi- lationship and jointly determine global morphology.
cal structure stretches in outward direction as long The textile transitions from a flat fabric to a three di-
as there is air pressure applied and as long as mate- mensional volume, and balloons are constrained in
rial has a capacity to stretch. When the same units place rather than being dispersed.
are placed together they share only one tangential
point of contact and under air inflation would expand FABRICATION
outwards becoming larger in volume and remaining Physical prototyping was carried out with the com-
a single point of contact. When constrained within a puter programmed domestic electronic knitting ma-
fixed or flexible boundary, the inflatable aggregates chine Brother KH 930. Internal floppy disc of the ma-
behave differently. A solid boundary (e.g. rigid box) chine was exchanged with Arduino + AYAB Schield
will be directly affecting the balloons: they will take [2], which allowed to gain a high level of control
a shape of a solid boundary and the expandability of over the fabrication and apply custom generated pat-
balloon is limited to the free space between the rigid terns. The mentioned knitting machine is able to pro-
box and the inflatables. Shared contact area would duce a canvas up to 200 stitches in width (100 ± 10
transform from a single point to a planar surface: bal- cm) and unlimited in length. Continuity of the pro-
loons will try to negotiate any free unexplored space cess is the way to an articulated linearity which intro-
within the solid boundary. In the situation with a duces a linear design element.
soft boundary (e.g. textile membrane), the balloons In order to establish a connection between the
and the boundary skin are dynamically affecting each user and machine, a graphic interface was utilized
other. The soft boundary with the differentiated sur- and integrated into a digital workflow. The knit-
face elasticity property would deform under the pres- ting loop based mesh was abstracted into a quad-
sure of expanding balloons and will change in dimen- mesh.Each quad was represented as a correspond-
sions (Figure 1) and shape till the locking point of the ing pixel in a bitmap image (Figure 2). Each pixel
knit structure, which means that all loops reached would carry an information about the location and
their final state of the structural extension. A flexi- the position of the needle on the knitting bed during

Figure 3
A. Binary syntax, B.
Abstracted stitch
bitmap, C. Resulting
knit fabric .
Figure 4
Comparison of
various knitted
surface porosities
and respective
curvature sequence
during inflation.
the process. The fabrication setup utilized two nee- INITIAL PHYSICAL PROTOTYPING
dle beds, which could be used to create a variety of The series of empirical tests were carried out to un-
stitch configurationsand corresponding material be- derstand the relation between surface assigned knit-
haviors. Though several techniques were explored ting patterns and the performance of the geometry
and show potentialfor architectural exploration, the while inflated. The position and the ratio of the A
process explained in this paper utilizes this setup to and B stitches in the pattern would create areas in the
create variation in the stitch size. The pattern was inflated cushion that expand or keep initial stiffness.
programmed as a binary message containing two Therefore, the inflated geometry would perform a
types of information - A andB. Stitch A, the white bending action under the expansion of the fabric ar-
pixel, is a standard knit stitch, while stitch B, the black eas formed out of B stitches. That means, that lin-
pixel, is structurally the same as stitch A (knit stitch) early distributed expandable pattern along the can-
but twice as large in dimensions. Those two would vas will create linear expansion along the side of the
become the main variables for the pattern genera- inflated fabrics, which will lead to the single plane
tion. Combined together into a textile, the stitches equal bending of the geometry. Respectively, the
would create areas of expandable or constricted sur- more converging distribution of the B stitch pattern
face (Figure 3). would perform local sharper kink curvatures of the
geometry (Figure 4).

The set up for the conducted initial physical exper- prototyping mode were integrated into a computa-
iments contained a linear piece for testing that was tional model development. It was taken into consid-
constructed out of a single knitted textile surface eration the following parameters such as the dimen-
with the differentiated density of stitches along the sions of the tested elements and the location of ex-
central axis. The surface was folded and stitched into pandable textile properties within the piece. In the
the tube geometry that was filled with 8 inflatable first set of digital exercises the aim was to represent
aggregates that were placed linearly inside the cush- investigated material system properties digitally and
ion and inflated simultaneously. Changes in the be- to achieve the similar behavior as was observed in
haviour of the knitted inflated element depend on a physical experiments (Figure 4). This implied the
the porosity ratio of the knitted surface: 1:8 (less furling behaviour of the linear tubular element under
amount of porous stitches) shows much smaller cur- the air inflation force with the particular set of sur-
vature rather than the one with 1:4 ratio of surface face properties conditions (Figure 5). The following
porosity (Figure 4). Promising results of the initial set of surface conditions could be described as a va-
physical experiment engaged to the further digital riety of simple geometrical distributions of the A and
explorations of the investigated material system. B stitch areas within a linear canvas: A - 0% porosity,
(420 A stitches), B - 8% porosity, (33 B stitches to 387
COMPUTATIONAL MODELLING A stitches). C - 24% porosity (99 B 321 A), D - porosity
With the purpose to control the resulting bending 47% (198 B / 222 A), E - porosity 24% (102 B / 318 A), F
curvature of the inflated hybrid geometry - a pattern - porosity 47% (198 B / 222A), G - porosity 49% (204 B
generator was developed and computational simula- / 216 A) , H - porosity 36% (152 B / 268 A) , I - poros-
tion model. All the observations obtained in an initial ity 56% (236 B 184 A). Interesting rules were drawn
from this digital exercise. The surface that was made
Figure 5
Bottom Up
approach in
representing hybrid
system: from
pattern to the
shape.

out of 100% stitches of same elasticity was not per- patterns from the target curvature of the geometry.
forming any geometrical three dimensional furling. An alternative approach was also tested where the in-
The element remained straight. By gradually rais- put for the top down computational model was an
ing the percent of B stitches represented in the mid- axial curve of the target geometry. The target is a
dle of the surface by black pixels, the entire element base element for tubular geometry to be built along
started to perform planar bending in accordance to the curve axis. Depending on user input values for
the percent of present stitches B in the surface (Fig- dimension and resolution, the pipe is divided into
ure 5, A - D). Even though the mentioned examples quad based mesh. All quads carry the same informa-
have an equal percentage of surface porosity (both tion about a spring strength, length relaxation factor
47%), the more condensed distribution of expand- etc, meaning no geometric bending would happen
able stitches in example D, are resulting in more ex- during inflation. In the physical world this would be
treme curtailing rather than in F, when A and B (dense a textile tube knitted out of only A stitches. In or-
and porous stitches) are equally distributed stripes der to bring in a surface differentiation of the tube
along the surface. For stimulating the geometry to quad-base membrane - the target input curvature
bend off the plane the condensation of B stitches was was analyzed and based on resulting values the quad
placed asymmetrically to the surface axis (Figure 5H). mesh faces are extracted into a separated cluster of
The described digital experiment was conducted B stitches. The computational simulation tool was
in bottom up approach, where the user designs the feed two data flows with two different setting for the
knitting pattern and utilizes computational tools for mesh, which results in a bending behavior of the sim-
simulating the emerging shape. In order to be able ulated elements. After the geometry is simulated -
to use the discovered material system and the pat- the unrolled 2D non-distorted bitmap pattern is gen-
tern related geometrical behaviors for the design, the erated and later saved as jpeg. file to be transferred to
computational model was developed further to the the numerically controlled knitting machine for man-
next step to be able to generate pixelated surface ufacturing (Figure 6).
Figure 6
Top Down
approach in
representing hybrid
system: from target
curvature to 2D
pattern.

Due to the programmability of the described knitting matic network of tubing is inserted into the fabric for
technique fabrication details needed for the combi- simultaneous inflation and the cushion is sewn into
nation of the knitted textile and the pneumatic struc- the enclosed volume with the pneumatic aggregates
tural system could be integrated. Multiple openings embedded inside. When inflated, the flexible and
were introduced into the knitting pattern in order to soft hybrid textile system transforms into the rigid 3D,
interface the pneumatic network (stitch C, single lace semi-transparent articulated geometry with the co-
opening, Figure 7). The computational model also herent differentiation between expanded reticulated
able to calculate the intersection of several elements and the dense constricted surfaces (Figure 8).
and creates a denser area (stitches A) where those
could be connected. RESULTS AND CONCLUSION
Figure 7 For the proof of concept a 1:1 hybrid prototype was
Details for built. It was constructed out of 3 linear elements that
pneumatic network formed three-dimensional arch-like structure. Each
integration, lace element had a unique predefined pattern of the tex-
openings, Stitch C. tile skin according to the target bending of the geom-
etry. Due to the width limitations of the machine and
the diameter of the pneumatic units, knit dimensions
were 147 stitches wide and up to 1500 stitches long.
Those knitted fabrics were shaped into enclosed lin-
ear cushions and inflated (Figure 10). Digitally pre-
programmed patterns proved their reliability. The
physical knitted modules of the prototype performed
a relatively precise three-dimensional bending com-
Due to the complexity of the modeling process, paring to the computational simulations. The abil-
the inflation was simplified and represented digitally ity to differentiate shapes and reach a precise level
as a single internal volume rather than set of indi- of surface property control leads us to new stan-
vidual reconfigurable pneumatics. The simulation of dards of production. Two-dimensional, computer
the pneumatic system assumes that the cushion is controlled fabrication allows to produce large three-
air tight and inflated with the single air volume. As dimensional pieces. The design target user input al-
the current stage of the research focused on the de- lows for the direct calculation of the code data for the
velopment of the knitting processes, it was not im- machine and to be ready to start fabrication immedi-
portant to refine the simulation of individual pneu- ately. The high level of surface articulation admits the
matic elements inside of textile formwork. It is clearly control of the global geometry's behavior while in-
understandable that the digitally modeled and the flated, as well as the surface articulation for the trans-
build environment on the current stage of the re- parency and the light transmission. The highly intri-
search project remains uncallibrated. Nevertheless cate nature of the knit could contribute to design so-
the digital tool was precise enough to generate dig- lutions for lightweight architecture.
ital messages for communication with the domestic The flexible yet tensile nature of the knitted fab-
knitting machine to be able to knit and build topo- rics in combination with pneumatics has a high ar-
logically similar geometries. chitectural potential. Both materials can easily be
After the pattern is generated, the machine is folded and packed into a small volume and trans-
ready to fabricate a custom differentiated textile sur- ported, which is a great advantage for a temporary
face (Figure 11). When the textile is finished, a pneu- or deployable structure. Flexibility and the conti-

Figure 8
Physical
experimentation
with distribution of
expandable and
tight stitches in the
knit surface.
Figure 9
Design of 1:1 scale
prototype. A.
Perspective. B.Top.
C. Unrolled 2D pixel
map for fabrication.

Figure 10
Above: 1:1 Scale
prototype (Photo: Y
Baranovskaya.
Model: O Kalina) .
Figure 11
1:1 Scale fabricated
textiles.

nuity of the filament flow allows to merge elements tiple stage activation for more dynamic architectures.
with the single fiber. This opens up opportuni-
ties to use the system for compliant kinematic de- REFERENCES
signs, where usually a sophisticated joint system is re- Ahlquist, S and Menges, A 2013 'Frameworks for Com-
quired. The synthesis of the air and a textile filament putational Design of Textile: Micro-Architectures
is a combination of materials, with the zero waste fab- and Material Behavior in Forming Complex Force-
rication logic, which makes it very sustainable. Aactive Structures', ACADIA 2013, Cambridge, On-
The programmability of knitted textiles makes it tario, Canada, University of Waterloo, University at
Buffalo, SUNY, University of Nottingham, pp. 281-
a promising technique for fabricating differentiated
291
surfaces that perform various property of a shape Chaturvedi, S, Colmenares, E and Mundim, TS 2011
generation under the external/internal force applica- 'Knitectonics', Proceedings ACADIA 2011, Cal-
tion. A traditional technique for making textiles that gary,Canada, The University of Calgary, pp. 186-195
is accompanied with evolved modern tools could be Sabin, JE 2013 'myThread Pavilion: Generative Fabrica-
another alternative for moldless production: 2D pro- tion in Knitting Processes', ACADIA 2013, Cambridge,
Ontario, Canada, University of Waterloo, University
duced differentiated textile membrane activated by
at Buffalo, SUNY, University of Nottingham, pp. 347-
expandable latex sheet under air pressure will be a 354
good reusable formwork for light weight composite Scott, J 2013 'Hierarchy in Knitted Forms: Environmen-
material. This could also further be used to reinforced tally Responsive Textiles for Architecture', ACADIA
the air supported knit structure into a self-supporting 2013, Cambridge, Ontario, Canada, University of Wa-
fibrous structure. terloo, University at Buffalo, SUNY, University of Not-
tingham, pp. 361-366
However the fabric structures are more suitable
Seiler-Baldinger, AM 1979, Textile: A Classification of Tech-
for flexible temporary architecture solutions, there is niques, Crawford House Press
a possibility to transform them into permanent by Thomsen, MR, Tamke, M, Gengnagel, C, Faircloth, B and
stiffening them. Impregnation the yarn with epoxy Scheuer, F (eds) 2015, Modelling Behaviour, Springer
polymer would solidify a knitted mesh into a rigid International Publishing, Switzerland
surface. A better structural performance would be Vrouwe, I and Pak, B 2015 'Framing Parametric and Gen-
erative Structures: A Novel Framework for Analysis
achieved if continuous non-knitted roving are inter-
and Education', eCAADE 32 2015, Vienna, Austria / Vi-
woven into the knit mesh in the direction of the load. enna University of Technology, pp. 365-371
Further investigations on the following topic can [1] http://www.nike.com/us/en_us/c/innovation/flykn
focus on the more integrated interrelation between it
inflatable aggregates and textile in digital and phys- [2] http://ayab-knitting.com/
ical environment. Digital environment could be im-
proved drastically by introducing a separate layer
for simulation the individual inflatable aggregates
within the flexible membrane. The process from
3D membrane patch to the unrolled flattened stage
could correlate with the more specific membrane
properties such as strain deformation factor in course
and wales direction which will imply the distortion of
the pixelated image in according to those. The con-
cept of using aggregates within textile cushion could
lead to deeper investigations on aggregate shape de-
velopment and activation, as well as introduce a mul-

Natural Complexity
Embedded Fabric Materiality in Fibre-Reinforced Polymer Fabrication
Arielle Blonder1 , Yasha Jacob Grobman2

1,2
Technion Institute of Technology
1
arielleb@tx.technion.ac.il 2 yasha@technion.ac.il
Fibre composites are synthetic in their composition, but structured similarly to

biological materials, as fibre and matrix. While the fibre constituent in
Fibre-Reinforced Polymer (FRP) is mostly used under the form of fabrics, its
standard fabrication processes do not rely on its inherent textile attributes.
Embedding the fabric qualities in the fabrication of architectural FRP can
potentially introduce concepts and properties of biological materials into
engineered fibre composites; it can promote and enable the generation of an
architectural complexity of a biological nature. The paper presents Fabric
Materiality as a framework for a new design and fabrication process and
demonstrates through a case study its integration in the fabrication of
architectural FRP elements to achieve a complex structure with bio-inspired
properties.
Keywords: Textile, Materiality, Self-organisation, Resilience, Composites, FRP
INTRODUCTION (Satterfield and Weinstock 2015). Within the archi-

Man has always been impressed and amazed by na- tectural discourse, the natural interrelation of matter,
ture's beauty and complexity. Advances of technolo- form and performance sets the ground complex sys-
gies, sciences and theoretical paradigms, along with tems of material and form (Hensel et al. 2010).
the increasing fragility of our ecosystem, have trans- The paper presents the results of a research that
formed our attitude to nature in the past decades. develops alternative approach to architectural FRP. It
From astonishment and inspiration we have shifted presents Fabric Materiality as a framework for a new
to the sophisticated study of complex natural sys- design and fabrication process that introduces con-
tems, under the general approach of Biomimet- cepts and properties of biological materials into en-
ics (Mazzoleni and Price 2013), and the blurring of gineered fibre composites, enhancing the material's
boundaries between the synthetic and the natural, in natural capacities to generate complexity.
practices such as Bio-design. In both academia and The following paper starts with the definition of
practice and across disciplines, from material science the concept of Fabric Materiality, with its three main
to architecture, spreads the conviction that nature characteristics of fabric manipulation, self- organisa-
holds a simple answer to complex issues and holds tion and resilience. It then follows with the case study
the keys for advancement of technology and design of The LifeObject, demonstrating the integration of

Fabric Materiality in the design and fabrication pro- performance materials, the use of FRP has spread ex-
cess of a FRP installation structure, through the three tensively over the past decades to almost every in-
characteristics, achieving material and formal com- dustrial field, from aeronautics to automotive, naval,
plexity. infrastructure and design. In architecture, there is a
renewed interest in the material in recent years, in-
NATURALLY COMPLEX fluenced by the introduction of digital tools for de-
sign and fabrication and by the quest for complex
Naturally Complex Material
free-form morphologies (Kolarevic 2005) . Biological
The inconceivable variety and complexity of natu-
paradigms and characteristics of natural systems can
ral living forms, shapes and structures is actually
be of high relevance in the design and engineering of
the product of several low-atomic building blocks;
architectural FRP, and can be relatively easily applied
the constituents of biological materials are very few,
(Knippers and Speck 2012).
mostly combined as composite fibre structures (Chen
et al. 2012). Only four main fibrous structural mate-
rials constitute the core of living materials: cellulose
Naturally Complex Form
On-going developments in technologies of visuali-
for plants, collagen for animals, chitin for insects and
sation and parametric design tools in the past two
crustaceans and fibroin silks. With a material den-
decades have fuelled a strong drive for the gen-
sity and relative stiffness that is typically much lower
eration of complex surfaces of free-form nature.
than engineered materials, their structural capacity
The challenge in bringing virtual complex-curved
is achieved by their geometrical and hierarchical fi-
geometries into materialisation at the architectural
bre architecture. By different spatial arrangements,
scale is supplemented by the parallel architectural
elements of very different properties are constructed
practices aiming for variation, differentiation and or-
using the same fibre material; collagen composes the
namentation of the surface. Such tectonic aspira-
stiff bone as well as the flexible tendon and the soft
tions are somewhat difficult to achieve in a top-
blood vessel, in the form of a fibre composite material
down fabrication system, relying on sheet materials
(Hensel et al. 2010).
or moulds; on-going developments and research in
Fibre based composite materials have been used
the field of full-scale 3D printing attest for this need
by man since antiquity, when wood and straw were
and interest. Tackling this issue from a different an-
coupled with mud, to produce the brick. From the
gle, bottom up processes of self-organisation allow
1950's and until today, advanced fibres such as glass,
for a complex outcome by simple means. In form-
carbon and aramid, are combined with polymers, to
finding design methods, either physical or digital,
make an extremely strong and lightweight material
form is not defined by the designer but rather gener-
commonly referred as composites, or FRP (fibre re-
ated by the definition of boundary conditions (Burry
inforced polymers). Structured by fibre and matrix,
1993); nature's patterns of tight relationship between
similarly to living materials, FRPs introduce biologi-
form, force and mass lead to optimisation and sta-
cally inspired design paradigms; like in nature, a wide
bility features. Controlling self-organising materials,
range of functional properties can be achieved by the
under tension or compression, can easily generate
careful design of the two components, varying spa-
forms of high complexity and intricacy.
tial and hierarchical organisation. It suggests engi-
Textile is a material system with a capacity for
neered continuous variation, local deposition of ma-
self-organisation in three-dimensional space; the
terial in accordance with structure and performance,
fibre-based structure of the material and its resilient
as opposed to conventional and uniform mechani-
character enables the generation of complex forms
cally engineered quality.
by simple means. The spatial arrangement of fibres
As part of the ever-increasing quest for high-

by mechanical interlocking only allows for the dy- 2012), Fabric Materiality is an approach to tightly re-
namic reaction to extrinsic forces, such as gravity, lated design and fabrication, originating from mate-
and to the induction of low-stress forces on the ma- rial properties. In this research the general concept
terial by its manipulation. Garment-making has de- of Fabric materiality is developed with regards to fi-
veloped, through millennia, the art of creating 3D bre composites.
complex surfaces and volumes, by simple manipula- Three main characteristics of textiles were identi-
tions and low-stresses through pattern-making and fied as defining the essential properties of Fabric Ma-
needlework. Being an engineered fibre-based mate- teriality: fabric manipulations, self-organisation and
rial, its properties and behaviour are defined by its resilience.
fibre architecture; as in natural materials, variations Fabric manipulation refers to the multitude of
in fibre type, density or spatial configuration (such as parameters that affect the physical properties of the
knit, weave, braid and more) define its performance. fabric, from the fibre itself to its spatial structuring by
weaving, felting, knitting etc. Being and engineered
Naturally Complex FRP material, its build-up is controlled by a great number
While the fibre constituent in FRP is mostly used un- of parameters; density, pattern, tension, fibre con-
der the form of fabrics, its standard fabrication pro- struction and type of machinery are just a few exam-
cesses do not rely on its inherent textile attributes. ples of parameters that affect the characteristics and
Standard FRP fabrication processes press fabrics onto behaviour of the resulting surface, enabling its ma-
rigid moulds, utilising the fabric's ability to adhere nipulation.
to the given rigid form in an optimal way; mechan- Self organisation refers to manipulations on an
ical pressure over the mould overrules the fabric's re- higher level of hierarchy, dealing with the fabric it-
silient character and its capacity for self-organisation. self. It refers to the ability of fabrics to embrace com-
The resulting morphology reflects only this of the plex form as a result of their manipulation. The ability
rigid mould, with no presence of any typical textile to apply low-energy stresses upon fabrics and gener-
form (Mallick 2008). ate complex three-dimensional structures is a unique
While FRP is structured similarly to biological property of the material, which has been the found-
composite materials, standard architectural FRP ap- ing principle of fashion along history. This ability to
plications do not relate to biological design princi- easily manipulate the material, by needlework (Wolff
ples. Experimental research, mainly conducted by 1996), heat and pressure (Wada et al. 2012) or by
ICD Stuttgart (Institute for Computational Design) pattern, is another level of the infinite variation of
with the ITKE (Institute of Building Structures and forms in garment making. It relies on the capacity
Structural Design), focuses on the integration of prin- of the material for self-organisation, the spontaneous
ciples of biological composites into the design and organisation in space as a three dimensional struc-
fabrication of architectural FRP (Parascho et al. 2015). ture, as a result of the internal local interactions be-
This research tackles the material on the fibre level, tween the fibres, and the external forces applied on it.
mainly dealing with direct robotic fibre placement This emergent behaviour of the material holds within
it the capacity for natural optimisation, relying on the
FABRIC MATERIALITY multitude of elements on the lower level of hierarchy
The term of Fabric Materiality is coined to represent of the material, namely the fibres.
the unique properties of textile materials and their This multitude of simple and weak elements that
processing techniques, together with the inherent interact and construct a greater whole is a key factor
design approaches these introduce. Treating mate- of the fabric's resilient properties. It gives the mate-
riality as part of a larger design paradigm (Oxman rial its flexibility and ability to recover to an initial or

improved state after an event of stress, demonstrat- towards one that is synthetic, yet follows that struc-
ing soft stability and robustness. ture of biological materials, as fibre and matrix. Com-
posed out of over 1500 tubular elements, the tubu-
Fabric Materiality in FRP lar elements were made of fibre composites, embed-
Fabric Materiality could be embedded in the process ding Fabric Materiality in their fabrication and assem-
of fabrication of FRP, enhancing textile attributes and bly process, to enable the assimilation of biological
biologically inspired design methods. Integrating principles in the overall design. The focus of this pa-
textile-related techniques of form-making and mate- per is on the fabrication process of the components
rial construction from the world of garment making and the way by which it contributes to the resulting
as well as from the architectural form finding disci- structure.
pline suggests the freedom from moulds, and pro- Several quantitative and qualitative principles
poses ways for local differentiation for performance, were extracted from the nest, setting the require-
optimization and ornamentation (Blonder and Grob- ments from the individual composite components.
man 2015). Relying on the natural properties of fi- The structure is composed out of a multitude of light
bre architecture, both at the material level and on elements that are similar yet varied; the components
the structural level, enables simple the generation of therefore should be light and their fabrication pro-
complexity. cess should afford easy variation. Moreover, differ-
Trying to embrace Fabric Materiality and take ad- entiation of the components should be obtained by
vantage of its unique qualities, potentially affects all the variation of material properties. The overall form
level of the process, from design to fabrication and is achieved by the combination of random elements
construction. When integrated, the characteristics of with no additional glue or joints, but by bending
fabric materiality are reflected in the resulting struc- forces; the components should therefore be flexible
ture across levels, from its design to its performance to a point that enables bending while maintaining
and structural behaviour. Embedding Fabric Materi- some resistance that generates internal stresses. Fur-
ality in the fabrication and design process of archi- thermore, details in the components should be inte-
tectural FRP to achieve a complex structure with bio- gral to the fabrication process.
logical attributes was the purpose of the case study, Embedding Fabric Materiality in the process en-
demonstrated hereby through the three features of ables to meet the above-mentioned design and fabri-
Fabric Materiality. cation requirements in a coherent way, which would
have been difficult or even impossible by traditional
LIFEOBJECT, A CASE STUDY fabrication processes of FRP. The basic process con-
LifeObject is an architectural installation for the Is- sisted of the fabrication of fabric sleeves, their im-
raeli pavilion of the Venice architectural biennale pregnation with resin, curing in ambient tempera-
2016 ; it serves as a case study for the demonstra- ture and short post-curing in low-temperature oven.
tion of embedded Fabric Materiality in the fabrica- The following section will describe the imple-
tion and design process of FRP, and its ability to in- mentation of the three key features of Fabric Mate-
herently enable complexity and variation of structure riality in the case study: fabric manipulation, self-
and design, as a parallel to natural composite sys- organisation and resilience.
tems. "LifeObject" is a structure that originates from
the analysis of a bird's nest, aiming to transpose its re- Fabric Manipulation
silient properties into an architectural installation. As The components of the LifeObject were made as con-
part of the overall biology-oriented design approach tinuous sleeves of two types of fibres, fiberglass and
to the structure, the choice of material was oriented aramid. Knitting or braiding processes were chosen

Figure 1 for the material construction, according to the na-
Para-aramid sleeve ture of the fibres, their fibre breakage and internal
knit in circular structure (flat or twisted). Fibreglass' fibre breakage
knitting machine. is high and present difficulties for knitting. As fibre-
glass braided sleeves are readily available in the mar-
ket, these were used in different diameters and den-
sities. The para-aramid fibres are soft and can easily
be knit, with minimal adaptation of settings of the
knitting machine to the fibre's strength. Two grades
of fibres were knit, (3600Denier, 200Denier), both in
flat and circular knitting machines (Figure 1). Circu-
lar knitting machines are mainly used by the indus-
Figure 2 try for the fabrication of stockings and socks. Assur-
Variants in knitting ing speed of production and homogeneity of surface,
patterns, circular they are limited in their diameter and the grade of fi-
machine. bre they allow. It was therefore used for the lighter
fibre grade only.
Several knitting parameters were varied, affect-
ing the mechanical properties of the resulting fab-
ric sleeve and its behaviour under forming process.
Different knitting parameters were explored, in the
search for sleeves that would have homogenous sur-
face quality, the ability to deform and take shape un-
der minimal tension while holding in bending forces
Figure 3 without collapse. Major variations included fibre
Knitting variants, composition (fibre grade, number of filaments, ad-
formed and cured. dition of other fibres), number of needles, tension,
and knitting patterns (single, tuck and miss for the
flat machine, single, various ribs and piqués for the
circular). Out of the complex web of interrelated pa-
rameters, all affecting the final outcome, a few simple
examples can be given:
Additional Material. The addition of a different fibre
in the knit, such as nylon monofilament, transformed
the fabric to a thicker and more rigid one. While im-
proving significantly the performance of the lighter-
gage knit, the nylon filament introduced in the heav-
ier gage knitting made it harder for the sleeve to em-
brace form under tension, collapsing into a flat shape.
This option was therefore eventually discarded.

forces, with pattern lines acting as lines of reinforce- Figure 4
ment or as hinges. Some elements were designed Knitting patterns,
with alternating patterns within them, locally opti- input through the
mising to adhere to the forming elements (Figure 5). knitting machine's
Knitting technology allows for the simple varia- internal software.
tion of parameters within the continuous material, by
changing densities, patterns or even materials with
the addition of threads. It enables the easy genera-
tion of variation within the element, differentiating it
in shape and performance.
Figure 5
Changing pattern
within the element
for local
optimization.
Density. The amount of tension applied on the fab-

ric during its knitting process affects the density of
the resulting fabric. While a looser knit allows for
easier deformations and better embrace of the nat-
ural form, it encapsulates within it a lesser amount of
resin. While such a component might perform better
for forces applied on its axis, it is mechanically weaker
and will more rapidly collapse when bending forces
are applied. The knitting tension was defined opti-
mising both targets, in accordance with the knitting
pattern,
Knitting Pattern. The knitting pattern acts as inter-
nal material structure, affecting its overall mechanical
behaviour. Different knitting patterns were tested for
the circular machine, as variations of piqué and rib
structures by an alternating Lycra filament that was
introduced in the knit (Figure 2 and 3). Apart from the
aesthetic value of the pattern, it affected the stiffness
of the tube's wall and its behaviour under bending

Figure 6 Self-Organisation
Resin impregnation. Biological material is synthesized with low-energy in-
vestment, and the generation of variations does not
require additional resources. As standard composite
fabrication process do not enable the easy variation
of form, requiring the costly changing of the mould,
Fabric Materiality was integrated in the forming pro-
cess, through self-organisation of the fabric material.
The knitted sleeves were impregnated with resin (Fig-
ure 6), and hung in tension for their curing, with no
moulds (Figure 7).
Simple means such sections of standard plumb-
ing tubes and rings were used as end-parts for the
tensioning of the component during curing, and to
create 'stoppers' along the tube, for the interlocking
of the components in the structural assembly. The
placement of the rings inside the sleeve and the ma-
nipulations on it (such as torsions, or the tension ap-
Figure 7 plied on the extremities) controlled the resulting el-
Forming by tension. ement in terms of overall shape and articulation of
details and stoppers, as boundary conditions. This,
in addition to the fabric manipulations mentioned
above, (such as density and pattern), together with
geometry parameters (sleeve diameter, length).
Using self-organisation as a fabrication process
enabled endless variations in length and placing of
stoppers within the elements, while keeping similar
characteristics of global appearance and behaviour
of elements. Similarly to biological material, a mini-
mal energy is invested in the generation of variants
(Figure 8). The resulting structure is comprised of
1500 varied elements that are extremely light and its
density is of 10kg/m3 only.

in The LifeObject both on the component level and Figure 8
on the level of the overall structure (Figure 9). Variation of
The internal knit and braided structure of the components.
components gives it its flexibility and strength while
being extremely light. The overall structure is also
the outcome of the interconnected components, rel-
atively weak by themselves, but globally stable and
strong as a whole. Relying on a multitude of weak
elements, which are of similar type but varied, the
structure demonstrates properties of resilience. It en-
ables local errors, withstands local failures without
global damage and springs back to a stable state af-
ter stress is applied. Self supportive and with no ad-
ditional joints or gluing agent, the structure can with-
stand several parallel lateral loads of 50 kgs.
Figure 9
LifeObject overall
structure.
Resilience
The resilient nature of fabrics is the combined prod-
uct of its internal material structure across levels of
Figure 10
material hierarchy and its material properties. At the
Finalised structure.
fabric level, the textile sleeve is composed out of a
Photo: Dacian
multitude of elements, the fibres, which are intercon-
Groza.
nected by friction in their spatial arrangement, while
maintaining some freedom and independent move-
ment. Going down in scale into the fibre itself, it is
also an assembly of filaments, interconnected by tor-
sion and friction while keeping relative freedom of
movement. At both levels, the single element - be
it the filament or the fibre - is light and weak, and
strength is achieved by multitude and spatial config-
uration. These principles of interconnected elements
with relative independence of freedom are reflected

The term of Fabric materiality is coined to repre- Blonder, A and Grobman, YJ 2015, 'Design and fabrica-
sent the unique properties of textile materials and tion with fibre-reinforced polymers in architecture:
their processing techniques, together with the inher- a case for complex geometry', Architectural Science
Review, 59(14), pp. 257-268
ent design approaches these introduce. It is char-
Burry, M 1993, Expiatory Church of the Sagrada Família:
acterised by the capacity of the material system to Antoni Gaudí, Phaidon Press, London
self organise, its ability to generate complex form Chen, PY, McKittrick, J and Meyers, MA 2012, 'Biolog-
by low-energy manipulations, and its resilient prop- ical materials: Functional adaptations and bioin-
erties. The notion of Fabric Materiality as a design spired designs', Progress in Materials Science, 57(8),
and fabrication process was tested in The LifeOb- pp. 1492-1704
Hensel, M, Menges, A and Weinstock, M 2010, Emer-
ject case study, demonstrating its ability to gener-
gent Technologies and Design: Towards a Biological
ate complex form and material, which are biologi- Paradigm for Architecture, Routledge
cally inspired. Fabric is manipulated by the variation Knippers, J and Speck, T 2012, 'Design and construction
of several knitting parameters, and form is achieved principles in nature and architecture', Bioinspiration
by self-organisation on the component level. By \& Biomimetics, 7(1), p. 015002
simple means, a high level of complexity and intri- Kolarevic, B (eds) 2005, Architecture in the Digital Age: De-
sign and Manufacturing, Taylor \& Francis, New York
cacy is achieved, demonstrating biological proper-
Mallick, PK 2008, Fiber-reinforced composites: materials,
ties. The manual fabrication of elements without manufacturing, and design, CRC Press, Boca Raton,
moulds together with their fabric manipulation en- Fla
ables easy variation between the similar elements, in- Mazzoleni, I 2013, Architecture Follows Nature-Biomimetic
troducing singularity within the types. The naturally Principles for Innovative Design, CRC Press, Boca Ra-
optimised form with embedded stoppers, achieved ton
Oxman, R 2012, 'Informed tectonics in material-based
by self-organisation, generates differentiation within
design', Design Studies, 33(5), pp. 427-455
the element. Being light, relatively weak and requir- Parascho, S, Knippers, J, Dörstelmann, M, Prado, M and
ing minimal fabrication efforts, the structure is com- Menges, A 2015, 'Modular Fibrous Morphologies:
prised of a multitude of elements, based on a logic Computational Design, Simulation and Fabrication
of redundancy rather than optimised minimalism. It of Differentiated Fibre Composite Building Compo-
achieves form and stability by its hierarchical material nents', in Block, P, Knippers, J, Mitra, NJ and Wang, W
(eds) 2015, Advances in Architectural Geometry 2014,
construction, from the fibre to the overall structure.
Springer International Publishing, pp. 29-45
Based on simple fabrication processes, an ex- Satterfield, B and Weinstock, M 2015, Hypernatural: Ar-
tremely light and stable structure is constructed. chitecture's New Relationship with Nature, Princeton
Comprised of 1500 varied elements, its density is of Architectural Press, New York, New York
10kg/m3 only (lighter than feather). It is self support- Wada, YI, Rice, MK and Barton, J 2012, Shibori: The Inven-
ive, with no additional glue or joints, and is capable tive Art of Japanese Shaped Resist Dyeing, Kodansha
America, Inc, New York
of withstanding several parallel lateral loads of 50kg.
Wolff, C 1996, The Art of Manipulating Fabric, Krause Publ,
Thanks to the inherent resemblances between Radnor, Pa
Fabric Materiality and characteristics of biological
materials, the integration of Fabric Materiality in the
fabrication process of FRP can lead to a material
and structural complexity of a biological nature. It
suggests an alternative to the standard architectural
means for achieving complexity, relying on material-
ity.

"Free Skin" Collaboration
Negotiating complex design criteria across different scales with an
interdisciplinary student team
Timo Carl1 , Frank Stepper2

1,2
University of Kassel
1
www.uni-kassel.de/fb06/fachgebiete/architektur/experimentelles-entwerfen-
und-konstruieren
1,2
{tcarl|stepper}@asl.uni-kassel.de
The complex nature of architecture requires often planning teams with specialists
from multiple disciplines. Architectural education however, addresses this
interdisciplinary modus operandi rarely. This paper presents the design and
production process of a real world solar façade installation realized at the
University of Kassel to illustrating the potentials of such an approach.
Interdisciplinary teamwork allowed students not only to solve complex problems,
but also to produce knowledge and to advance into design research. Student
exploration resulted in a unique fabrication technique, combining tensile fabric
and resin to facilitate the fabrication of multifunctional, monocoque shells;
combining all necessary technical components in a single building element. This
paper discusses the success of student collaboration and teaching strategies for
key parts of the design process at different scales. Moreover, it highlights the
importance of physical form-finding models and an analogue - digital workflow
for collaborative communication. The Free Skin project offers both insight into
applied use of interdisciplinary teamwork, and a proposal for incorporating such
collaboration into architectural education.
Keywords: interdisciplinary collaboration, design-build, form-finding, reactive

design, shell structures
ENGAGING COMPLEXITY ments concerning material, structure, design, and

Complexity is defined as "[...] the state of being function are all part of a coherent concept. An en-
formed of many parts" (Oxford Advanced Learner's deavor which often cannot be solved by one disci-
Dictionary 2016). Architectural design is by defini- pline alone anymore, but involves inter-, multi-, trans-
tion a complex subject matter as multiple require- , or anti-disciplinary (i.e. designers and scientists)

teams. The "comprehensive" nature of architecture terdisciplinary research project at the Department of
requires in practice often planning teams, which con- Experimental Design and Construction at University
sist of multiple specialists. The New European Central of Kassel, investigating architectural strategies for the
Bank project by Coop-Himmelb(l)au for instance in- activation of urban spaces through sensor activated
cluded about 40 different planning disciplines, all co- real time interventions. For the design and fabrica-
ordinated by the architects (Mönninger 2010). Multi- tion of a functional real scale prototype, we are build-
disciplinary work among professionals is the modus ing upon experiences of the previously realized Ur-
operandi in practice. In the educational context how- ban Glow installation (Carl and Juric 2015). A research
ever, neither multidisciplinary teamwork (collabora- class with senior undergrad students featuring mate-
tion among disciplines) nor collaboration between rials and structures in architecture formed the frame-
disciplines (interdisciplinary work) are yet widely work for design and production. The studio acted as
spread. Therefore, the authors present their investi- real world laboratory for design and experimentation
gations into collaborative work to promote a poten- and the production of prototypes as testing ground
tial framework for interdisciplinary teams in the ed- for interdisciplinary collaboration. The Free Skin stu-
ucational context. Due to its innovative nature and dent team included architects, electrical engineers,
the integration of science and art concepts, the com- and an information scientist. The team exhibited a
plexity of this project was rather high. Collaboration functional demonstrator in an outdoor museum set-
was the key for successfully realization, as the integrating. It is helpful to set the stage by outlining first the
tion of solar energy and reactive behavior required project brief of the Free Skin project as a conceptually
new skills and methods in design. We argue in this open framework encouraging complexity and inter-
paper, that interdisciplinary design teams are more disciplinary teamwork. We will then discuss educa-
successful in solving real-world and complex archi- tional strategies and the influence of digital and ana-
tectural problems, because "The integration of the spe- logue tools for the creative process and team com-
cialist´s diversified approaches, methods and strategies munication. Finally, we reflect on the success of stu-
ideally enables a multifaceted view of a given problem dent collaboration and evaluate teaching strategies
and therefore enhances the creative potentials of the for key parts of the design process at different scales.
team members" (Sakatani 2005). This paper serves
to show that an inter-disciplinary team aided better- FREE SKIN
informed design decision across different scales for The pedagogical context focused not only on build-
developing and indeed building this Free Skin instal- ing performance issues, but rather on their relation to
lation. Furthermore, we discuss how students en- theory, form, technology and build object. The aim
gaged in experimentation and questioning of con- was to expand the skill-set of students while main-
temporary architecture as team and reflect on this taining the focus on design and built form. One
based on the design and fabrication process. We the one end the brief encouraged students to ex-
emphasize the role of interdisciplinary teamwork as plore second-degree auxiliary structure as described
a critical endeavor, not only to master complexity, by Michael Hensel for their use as sun shading sys-
but also to stimulate novel inquiries into form and tems (Hensel 2013). Such systems can be retrofitted
structure: A construction method for double curved to existing buildings and constitute a supplementary
composite shells, not yet widely explored was devel- layer of material systems (Hensel). On the other end,
oped for this project. The students managed to in- second-degree auxiliary facades correspond to the
tegrate thereby multiple functional layers such as so- architectural idea of the Free Skin as postulated by
lar cells, sensors and fiber optics into a single build- Silvia Lavin: "The idea of the Free Skin gives new in-
ing component. Free Skin is part of an ongoing intelligence, instrumentalities and plasticity to surfaces"

(Lavin 2009). The performance of the realized instal- both projects do not transgress a certain scale, are
lation unfolds during the day from the interplay of designed for indoor use and do not embed multi-
the architectural structure with the dynamic effects ple functional layers in one component. Inspired by
of light and shadow and at night from intricate mate- Heinz Issler´s fabric formed ice sculptures, students
rial effects and reactive behavior. combined tensile membranes with glass fiber rein-
forced plastic (GFRP´s). By eliminating the need for
Figure 1
tension cables or anchoring, students also investi-
Showing material
gated a material system, which facilitated the inte-
effects for night and
gration multiple components. Students used ten-
daylight.
sioned membranes for form-finding strategies, but
also as textile formwork to unify solar cells and other
components into one multi-layer surface. Thus, the
fabrication process could be easily applied in an ed-
ucational context and individual free-form molds be-
It is a free-form lightweight structure, reducing heat came expandable, saving substantial resources and
loads for south facing facades by means of solar shad- money. The resulting GFRP shells are weatherproof,
ing, while actively generating energy by implement- resisting environmental forces of sun, wind and rain.
ing solar cells. The generated energy was then used
to explore material effects by powering the imple- STRATEGIES OF REIFICATION
mented High Power LEDs and fiber optic cables. The We as educators are interested in a design culture of
brief of the Free Skin project was complex in its mul- experimentation, investigating alternative notions of
tiplicity of requirements but also conceptually open form and space not only on paper, but also in real
enough to act as an excellent design objective, as scale. This emphasis on design led to a complex
"Creativity can only occur when a task involves com- project, making it necessary to expand the pool of
plex and ill-defined problems that require the develop- knowledge beyond computational tools, methods or
ment of novel and useful solutions" (Sakatani 2005). skill sets typically found in architectural design stu-
We wanted to create a situation where interdisci- dios. The architecture studio is an environment of
plinary teamwork would assist students in exploring synthetic activity, negotiation formal, aesthetic and
the breadth of issues of such a Free Skin. To support functional aspects of design. Technology and design
this, we made a deliberate decision to engage stu- are in the Free Skin project inexorably linked. Our
dents to use textile membranes for their design ex- goal was to help architecture students to focus more
plorations. Thus, we could fall back on a rich body of on design potentials instead of struggling to mas-
previous work and rely on freely accessible off-the- ter technical necessities. However, collaboration al-
shelf software. Moreover, textile membranes facili- lowed not only for expanding the focus on function,
tated not only the production of complex geomet- but also on design. By crossing disciplinary bound-
ric formations but their use resulted also from a de- aries, an interdisciplinary team advances in other ar-
sire to explore material qualities and to exploit them eas of knowledge and can make more informed de-
for their particular effects. Previous work such as the sign decisions. "The integration of the specialists' diver-
Toroidal textile hybrid installation by Sean Ahlquist sified approaches, methods and strategies ideally en-
(Ahlquist and Menges 2012) or the Entry Paradise ables a multifaceted view of a given problem and there-
Pavilion by Lava (Lava 2007) demonstrate the po- for enhances the creative potentials of the team mem-
tential for geometric differentiation and the capac- bers" (Steinheider and Legrady 2004). An interdisci-
ity for color- and transparency modulation. However plinary Team can pick methods most suitable for de-

Figure 2
Showing
form-finding
modells, GFRP
lamination tests
and construction
process.
sign, construction and fabrication processes. Specif- achieve this goal the scope of the project had to be
ically for this reason, it was necessary to integrate well defined (project brief ); team setup had to allow
the other disciplines into the design process by en- for cross- pollination of ideas (integration of engi-
couraging interdisciplinary teamwork. For the suc- neering students) ; representation and communica-
cessful realization of this reactive installation, all team tion needed to be understandable for all team mem-
members involved had to define problems, gather bers (physical models). As there were only two stu-
information, and then progressively refine and ex- dents from electrical engineering participating in the
tend their initial ideas towards successful implemen- project, we formed three teams at the beginning of
tation. "A multifaceted view is only possible, if there is the project. The teams consisted at first only of ar-
a shared understanding of the problem and the vari- chitecture students and self-organized the division of
ous approaches to solve it" (Steinheider and Legrady). work with guidance form the instructors. The elec-
Strategies and methods to support this process are trical engineering students provided technical exper-
common in architectural practice, but not in edu- tise for the three teams, but conducted also their own
cation. Therefore, it was necessary to develop a research into lighting options. They presented the
common ground for communication and to define results at regular intervals to the teams, and started
tools and methods for the team. The psychologists to actively influence design decisions; the team owes
Brigitte Steinheider and George Legrady call this pro- for instance the utilization of fiber optic cables to the
cess knowledge sharing: "Teams must develop con- design interests of the electrical engineers. The me-
sensus about a project´s goals, have a clear under- dia artist Olaf Val acted as adviser to pollinate the
standing of team members´ diverse professional areas teams during informal pinups with ideas for reactive
of expertise and acquire a meta knowledge connecting design options. All students participated in short in-
the areas of expertise" (Steinheider and Legrady).To troductory workshops covering the basics for ana-

logue form-finding, solar analysis and optimization, Moreover, physical scale models were paramount for
spring particle system modelling, and introductory the success of collaborative work and played a key
programming for Arduino and Processing. Before stu- role for finding a common language for the different
dents moved to fabrication and production only one disciplines, as "prototypes [...] are the core means by
design option was chosen by the whole team for ex- which the designer builds the connection between fields
ecution. The two electrical engineers worked from of knowledge and progresses towards a product" (Stap-
then on in close collaboration with their peers to de- pers 2007). This interchange was critical in mastering
sign, build and fabricate the Free Skin installation. The not only the complexity of the task, but also for the
educators organized funding, overall organization, creative development of the project, and of a com-
schedules, company contacts for the project. mon team language. Students not only used proto-
types for testing technical feasibilities and design po-
MACRO SCALE: SIMPLICITY OF FORMA- tentials, but also as representational tools to convey
and to promote their ideas to the team members of
TION
the other disciplines. All students were able to collab-
During the first phase, the overall design ambition
orate closely across disciplinary boundaries through
was investigated by the use of analogue form-finding
the intuitive nature of physical models. Working with
strategies and through learning by doing, as not
physical models facilitated quick results and allowed
enough computational knowledge was available to
for rapid feedback, ensuring that all team members
the team in the beginning. Physical form-finding
could participate in the design decisions.
experiments have a long tradition in architecture
and engineering ranging from Gaudi´s invention of
an analogue computer for the design of the Colo-
nia Güell, to the scientific research of Frei Otto and
Heinz Issler. Especially Issler´s ice experiments pre-
cede several contemporary tendencies in architec-
tural discourse for the design of efficient and perfor-
mative architecture. Working with elastic textile ma-
terials students could generate minimal surfaces by
the simplicity of form finding. Based on the observa-
tions of the Belgian physicist Josef Plateu (1895) min-
imal surfaces fulfill the condition of a disappearing
mean curvature, which implies that in each point of
the area the sum of positive and negative curvature
is zero (Finsterwalder 2015). The resulting double
curved geometry stabilizes the membranes and fa-
vors a minimal use of material. Students experienced
an instant feedback of their design decisions by ob-
serving the interplay of tension forces, curvature and
boundary conditions. The material itself "computes"
form (Landa 2010). Pedagogically this is interesting,
because students not only learn to understand the
principles that generate form, but also to appreci-
ate architectural models as design method, which
places behavior over representation (Menges 2012).

tational setup. While spring particle systems, based Figure 3
upon the principles of Hooke´s law, simulate mem- Showing physical
brane behaviors, not all parameters of material stiff- models for
ness and plasticity do equate to real world param- investigating
eters. However, the results of the physical model material effects and
testing can be used to calibrate the correct material reactive behaviour.
settings, establishing a feedback loop between dig-
ital and analogue models. The digital model facil-
itated performative optimizations for solar shading,
programmatic distribution (amount of light required)
and social functions (views and privacy) and these
parameters became the drivers for the local differ-
entiation of the geometry. Linking the spring par-
ticle system with Autodesk Ecotect Analysis through
the Gecko plugin (Grabner and Frick 2013) the stu-
dents interactively manipulated the parameters of
the overall system with instant visual feedback for so-
lar optimization. This workflow supported collabora-
tive design decisions for calculate energy production
and for determining solar cells layouts and distribu-
tion. This workflow supported collaborative design
decisions for calculating energy production and for
determining solar cells layouts and distribution.
Figure 4
Showing visual
feedback analysis
for solar radiation.
MESO SCALE: COMPUTATIONAL DESIGN

AND LOOSE FIT
Analogue form-finding procedures facilitate struc-
tural formation and expedite design decisions re-
garding shape and material qualities like translu-
cency and interplay of light and shadow. Whereas a
computational approach is more useful for geomet-
ric optimization, the generation of multiple design
variations, and the production of fabrication data. As Students established a parametric framework for the
the team had no prior knowledge in simulation tools, design of the overall façade system, and had to
easily available visual algorithms such as Kangaroo3D solve the penalization of this tension active system
(Piker 2013) helped to transfer the underlying gener- (Menges 2012). Each tensile unit is thereby defined
ative principles of the analogue models to a compu- through a boundary frame. This frame could be eas-

ily adapted to specified design criteria and translated molds for the glass fibers, fiber optics, solar cells,
into precise geometric entities for production. In sensors and other elements. With the application
contrast to the linear components of the boundary- of resin, the discrete components could be unified
frame, the minimal surface geometries of the mem- into a single monocoque shell structure, combin-
branes could not be "unrolled" as they are double ing tensile and compressive behaviors. The linear
curved and constructed of a pre-stressed material. components of the cell frame formed a bending ac-
Nevertheless the tubular approach of the cell setup tive boundary, gaining stiffness through the tension
facilitated a "loose fit" accurate enough for the defi- forces of the elastic bi-axial fabric, similar to a spoke
nition of inner and outer radii for production. The pa- wheel. This material efficient production process
rameters for this "loose fit" was determined through a could be easily applied in an educational context. We
series of analogue scale models, carefully testing and as educators think, that it is important for students
documenting the stretching behaviors for different to understand not only the benefits of digital design
cut patterns and textiles. tools, but also the limitations of available computa-
tional simulations. We encouraged students rather
Figure 5 to develop their own solutions, instead of learning
Showing complex software. The team solved a number of
Grasshopper problems collaboratively by utilizing physical mod-
defintion defining els. This included specific investigations for cable lay-
cell openings and outs, integration of technical components into dou-
orientation. ble curved surface geometry, as well as the integra-
tion of reactive components like sensors and Arduino
boards. The use of new modelling and simulation
tools supported the overall workflow, but was not
paramount for the overall success of interdisciplinary
collaboration. Nevertheless, the Free Skin project de-
manded participation of an interdisciplinary team of
specialists to resolve investigations into reactive be-
havior at micro scale.
MICRO SCALE: CONSTRUCTION DETAILS

AND THE IMPORTANCE OF FAILURE
Constructions details as well as Arduino program-
ming, sensors and optical fibers were all tested in
real world settings. Team Size and increasing com-
plexity led to the development of sub teams solv-
ing problems in close collaboration. Students used,
for instance, different 3D printing techniques for cus-
tomized casings to integrate electrical components,
resulting in close cooperation between architects
The high degree of three-dimensionality in the mem- and electrical engineers for design and fabrication.
brane design was deliberately designed to mini- This led to discussions for a particular design deci-
mize undesired wrinkling trough pre-stressed tension, for instance the integration of cables, involving
sion. The membrane, surfaces were then used as not only the technical material aspects of this, but

Figure 6
Left: showing
electrical
components and
customized casings,
Right: showing
material stress tests
and connecting
node details.
also the phenomenological and conceptual implica- team member from the information science depart-
tions as it applied to the overall project. ment. As designers, we accept challenges without an
This dynamic served to demonstrate to students a priori solution and that it is the process of experi-
that the actions of one discipline within the team af- menting that generates knowledge and solves prob-
fected the actions of another as they coordinated the lems. Another important aspect of this approach is
production process and the integration of complex to understand that "any investigation which attempts
building components. Students focused not only on to challenge conventional notions of architectural form
their respective field to solve problems, but collabo- must also accept the importance of failure within the
rated with other disciplines for design decisions and process" (Enright 2009).
defined goals together. At the core of the reactive This is however fundamentally different than at-
concept was the idea, that solar generated energy tempting to teach technology through proven meth-
could be stored during the day and then used to ods and presented especially for the engineering
generate intricate lighting effects. As the Free Skin students a new way of learning. In the course of
installation was part of an outdoor Museum event, the project occurred several investigations, which
the team designed and programmed a slowly puls- started as failure, including 3D printed construction
ing light effect, which spectators could influence by knots, solar cell configuration and weather proofing,
varying their distance to the object. Most artists en- as well as electromagnetic sensor shielding. Never-
gaging with reactive installations value Arduino for theless, most failures led eventually to novel knowl-
its simple interface. However, the design decision edge production within the team and constituted an
to use 36 RGBW LED´s required programming the important learning model for all disciplines.
micro controllers in C language to compute signals
from the sensors and to handle the amount of out- DISCUSSION
put data. This posed a problem for the electrical en- Students managed to establish an interdisciplinary
gineering students, as the project was far more com- workflow across different scales and were able to ac-
plicated than their usual tasks and resulted in a new tivate realms of learning beyond their usual com-

Figure 7
Detail of final 1:1
prototype showing
integration of light
fibers and electrical
components .
fort zones. The inherent complexity of the Free Skin components to control interactive behavior and for
project was instrumental in challenging the students testing material effects on physical scale models.
to go beyond their knowledge of built form; ad- Consequently, the team negotiated differences in
vancing the skills of all participants. Categories typ- design cultures and knowledge through design ex-
ically associated with knowledge sharing resulted perimentation and made informed design decisions,
from this collaborative teamwork, including shared solving complex problems. In the end, an interdisci-
understanding of objectives and problems, shared plinary student team investigated a novel construc-
terminology, and adopting the other´s perspectives. tion process, combining tensile structures and GFRPs;
By hindsight, the schedule for the Free Skin installa- though significant post-production in sewing, edge
tion was rather tight, as it was taught in a research detailing and assembly needs still to be optimized.
seminar setting over the course of two semesters. The Free Skin installation will be placed on the roof of
This resulted for some team members committed to the department for Energy and Information Sciences
other projects in a lack of time, as they did not re- for further experimentation.
alize the complexity of integrating multiple require-
ments. Nevertheless, students were highly dedicated ACKNOWLEDGEMENTS
and applied multiple methods for design and fabrica- This project was made possible through the gen-
tion, including analogue form-finding experiments, erous support of the CDW Foundation, the Pfeiffer
off-the-shelf software for spring particle system mod- Foundation and 3M(TM). The seminar was co-taught
elling and solar optimization, physical computing by Vanja Juric, who deserves equal credit for the

work achieved. Special thanks goes to Olaf Val who [1] http://www.l-a-v-a.net/projects/entry-paradise-
generously assisted the course, advising on reactive pavilion/
design strategies and to Line Umbach for her dedi- [2] http://www.cluster.eu/2010/10/08/materialism-fo
r-architects-a-conversation-with-manuel-delan
cation and endurance. We would like to thank all the
da/
students who participated in the Free Skin project
2015-2016 for their hard work: Line Umbach, Tim
Stöhr, Eric Wiederhold, Lena Henriette Neuber, Anne
Liebringshausen, Oliver Raderschall, Philipp Kern,
Jonathan Schmidt, Christian Slama, Cagdas Alakus
Department of Experimental Design and Con-
struction (eek) - University of Kassel - www.uni-
kassel.de/fb06/fachgebiete/architektur/experimentelles-
entwerfen-und-konstruieren
REFERENCES
Ahlquist, S and Menges, A 2012, 'Physical Drivers: Syn-
thesis of Evolutionary Developments and Force
Driven Design', AD, 83(216), pp. 61-67
Carl, T and Juric, J 2015, 'Urban Glow- Interaktive Stad-
trauminterventionen', Archimaera, 6, pp. 135 -138
Enright, J 2009, 'Applications in Cross-Curriculum Teach-
ing The Synthesis of the Design Studio and Building
Technology Seminar', ARCC JOURNAL, 6(1), pp. 14 -
22
Grabner, T and Ursula, F 2013, 'Architectural Design
Through Environmental Feedback', Architectural De-
sign, 83(2), p. 142–143
Hensel, M 2013, 'Performance-Oriented Architecture: Re-
thinking Architectural Design and the Built Environ-
ment, Wiley, Chichester, UK
Lavin, S 2012, 'Performing the contemporary, or: to-
wards an ever newer architecture', in Grobman, Y
and Neuman, E (eds) 2012, Performalism. Form and
performance in digital architecture, Routledge, pp.
62 - 64
Menges, A and Ahlquist, S 2012, 'Physical Drivers: Syn-
thesis of Evolutionary Developments and Force
Driven Design', Architectural Design, 216, pp. 61-67
Mönninger, M (eds) 2010, Coop Himmelb(l)au Complete
Works 1968-2010, Taschen, Cologne
Sakatani, K 2004, 'Harmony Quest: An Interdisciplinary
Arts-based Project Incorporating Virtual Reality', Vi-
sual Arts Research, 31(1), p. 53–62
Steinheider, B and Legrady, G 2004, 'Interdsiciplinary
Collaboration in Digtial Media Arts: A Psychological
Perspective on the Production Process', Back Matter,
63(63), pp. 315 - 320

MATERIAL STUDIES | Applications
Bio-Enabled Façade Systems
Managing Complexity of Life through Emergent Technologies
Martina Decker1 , George Hahn2 , Libertad M. Harris3

1
College of Architecture and Design | New Jersey Institute of Technology 2 Department
of Computer Sciences | New Jersey Institute of Technology 3 School of Architecture
| New Jersey Institute of Technology
1
www.materialdynamicslab.com
1,3
{decker|lm229}@njit.edu 2 george@genericmaker.com
The abundance of vertical surfaces in urban environments and their use for green
installations have been of great interest in the architectural community. This
paper discusses how vertical gardens and urban farming can be spurred on and
enabled by two emerging fields of investigation: Material Science and the
Internet of Things (IoT). This paper discusses how two emerging strategies, smart
materials and novel sensing and actuation systems linked to the IoT, can facilitate
the creation of a stable environment or zone along architectural facades for the
creation of new ecosystems. Furthermore, this paper will contemplate future
trajectories for the integration of living matter into buildings and propose an
open source database that will facilitate the exchange of collected data and
knowledge to spur on further developments.
Keywords: Microalgae, Nanotechnology, Internet of Things, Bio-Enabled

Design, Microcontrollers, Sensors
INTRODUCTION: BIOMIMICRY AND BIO- an increasingly fragile natural environment, design-

ENABLED SYSTEMS ers and scientists alike are looking towards biological
Biomimicry in architectural design has enjoyed a lot systems and processes once more for inspiration.
of attention in recent years and is now going beyond Biomimetic artificial leaves that strive to match
the adaptation of geometric and formal similarities the performance of photosynthesis in plants are be-
(Brownell et al. 2015). The performative adaptation ing developed. The solar-to-fuel converting devices
of biological systems is inspiring smart building sys- are using solar powered catalytic processes that are
tems (Park et al. 2013) that are emulating biologi- inspired by the complexity of photosynthesis that
cal processes and are incorporating them into active has evolved over millions of years. Though the de-
and reactive building components (Decker 2015). velopments are very promising, scientists are con-
To counteract the negative effects of unsustainable stantly trying to improve the efficiencies of these de-
building practices, and to alleviate their impact on vices while using earth-abundant materials to con-
MATERIAL STUDIES | Applications - Volume 1 - eCAADe 34 | 603

Figure 1
Hydrogel at various
Stages of Water
Absorption, tested
at the Material
Dynamics Lab.
vert widely available resources such as CO2 and wa- to urban agriculture and food production (Specht et
ter into renewable fuels (Joya et al. 2013). al. 2014).
Rather than emulating biological processes,
other investigations are driven by bio-enabled strate- FROM HORIZONTAL TO VERTICAL INSTAL-
gies that directly integrate biologically derived mat-
LATIONS
ter or even living organisms. Examples, demonstra-
The same positive effects on cities that are attributed
tive of this strategy on the scale of products and de-
to green roofs, can also be true for green walls. One
vices, include another artificial leaf design that incor-
of the major challenges that green walls face in com-
porates biologically derived chlorophyll into a soft
parison to the horizontally planted structures is that
material matrix (Koo et al. 2011) or the Phillips' light
moisture retention is notoriously inadequate due to
fixture that is making use of bioluminescent bacteria
gravity induced runoff of water and nutrients. Ad-
that emit a soft glow in dark conditions [1].
vancements in the material science that have been
In Architecture we can build on a rich history
greatly spurred on by nanotechnology are offering
of both strategies for centuries. Biomimicry for es-
solutions for new green roof assemblies that can
thetic and decorative features have been mentioned
greatly influence the development of plant life.
in Vitruvius' deliberation on the origin of biomimetic
ornament design and the inspiring features of the Figure 2
acanthus plant (Vitruvius and Morgan 1960). First ex- Hydrogel
amples that are indeed integrating plant life into ar- Evaporation Study.
chitectural environments may be found referenced Credit: Material
as early as in the example of the Hanging Gardens Dynamics Lab.
of Babylon. Even though little references remain of
the description of the Hanging Gardens they have
been said to consist of a series of green roofs and ter-
races, resembling lush mountain surroundings (Clay-
Hydrogels that can hold between 400 to 1400
ton 2013).
times their own weight in water (Abedi-Koupai et al.
Contemporary green roofs are being praised for
2008) (see Figure 1) have been proven to support
their positive effects on urban ecologies. They are
plant life on green roofs especially during the early
being recognized to improve air quality (Speak et al.
growth stages (Savi et al. 2014) . In a series of stud-
2012) and boost the thermal performance of roof as-
ies that are currently conducted at the Material Dy-
semblies (Niachou et al. 2001) , and can alleviate the
namics Lab [2] at the New Jersey Institute of Technol-
urban heat island effect (Santamouris 2014). They
ogy, the application of these smart materials is being
provide green space for human use and support bio-
studied in order to sustain vegetation in vertical ar-
diversity and wild live in city settings (Brenneisen
rangements. While early test results seem promising,
2006). They also have a great potential to contribute
the hydrogels that expand in the presence of water
604 | eCAADe 34 - MATERIAL STUDIES | Applications - Volume 1

Figure 3
Microalgae
Samples. Credit:
Material Dynamics
Lab.
(Figure 1) have a bigger potential as a supporting fea- calls for careful consideration of all constituencies.
ture for vertical vegetation arrangements. Since nat- In architecture, we can build on a rich history of cli-
ural precipitation is rarely driven towards the vertical mate conscious building practices and special atten-
structures more elaborate irrigation systems are still tion has to be given to how buildings can mediate be-
necessary, even though the emergent material can tween the fluctuating conditions of the exterior sur-
assist in reducing the loss of moisture due to evapo- roundings and an ideally stable interior climate. The
ration. (see Figure 2) cyclical movement of the sun and insolation in par-
Successful green wall installations such as the ticular greatly influences many design decisions such
vertical plantings of Patrick Blanc's Mur Végétal [3], as location, orientation, or the shape of the building.
ensure the growing of ornamental plants through a With these passive techniques the building itself is in-
quite extensive irrigation system that has to be uti- strumental in shaping its own microclimates.
lized to maintain a livable environment for plant life The advent of HVAC (Heating, Ventilation and
(Gandy 2010). Blanc's work, akin to hydroponic sys- Air-Conditioning) systems has led to manufactured
tems, features continuously circulating water and nu- interior climates and environments that can be
trients through a vertically installed felt structure that mostly independent from the location of the build-
is substituting the earth and soil that plants usually ing, its climatic zone, or the time of day. One of the
rely on. first built examples that features a successful cooling
Water retention is only one of the challenges system was the New York Stock exchange (Nagen-
that we face when integrating living matter such as gast 1999), designed by Alfred Wolff in 1901. This
plant life into the architectural context. The difficul- building aspired to create an interior environment
ties lie in the placement of the organisms under new that is "superior in atmospheric conditions" in com-
and sometimes entirely artificially controlled envi- parison to anything that existed at the time (Wolff
ronmental conditions. Organisms that have evolved 1901). Since this early air-conditioning engineering
over millions of years, integrated in very specific milestone, the access to cheap energy from fossil
ecosystems, often resist being subjected to reloca- fuels and the further development of sophisticated
tion. In order to establish the conditions that are nec- HVAC systems play a significant role in the energy
essary to sustain life in a façade system the environ- consumption in buildings, which makes for roughly
ment has to be controlled quite precisely. 40% of all the energy that is being produced in the
United States (United States Energy Administration,
MICROCLIMATES AND CLIMATE CONTROL DOE 2014).
To establish a boundary layer ecosystem between the Through next generation façade technologies
surrounding climate and the human interior climate, that have the capability to control shading systems

Figure 4
Scenedesmus in
Photobioreactor.
Credit: Material
Dynamics Lab.
and ventilation, we have been able to gain even more tuned supply of water, nutrients, light, and CO2 is
control over the interior environment without hav- of the utmost importance. Crop yields can also be
ing to rely as heavily on contemporary HVAC systems greatly influenced by temperature fluctuations or hy-
that are known to be energy intensive. Many of the drodynamic forces within the reactor.
early dynamic façade elements that can mediate be-
tween interior and exterior conditions are currently Figure 5
driven by pre-set schedules (Krietemeyer et al. 2015). Scenedesmus in
New technologies are being developed at the Uni- Bioreactor on Day
versity of Syracuse that hold the promise to not only 10 of a 14 Day Grow
stabilize the interior climate, but can also cater to the Cycle. Credit:
specific needs of the individual users (Krietemeyer et Material Dynamics
al. 2015). Besides the mitigation of heat gain, day- Lab.
light and views, the system would also have to ac-
commodate the needs of the integrated living matter
in a bio-enabled façade system. Green wall ecosys-
tems require for example specific temperatures, light
exposure, and supply of other resources such as nu-
trients, carbon dioxide, or oxygen. Human expecta- Emergent investigations are envisioning the benefits
tions on the other hand also have to be met on the of integrating microalgal photobioreactors directly
interior. The input parameters of both user groups into the constructed environment (Prokssh 2013) in
have to be in constant negotiation with climatic con- order to capitalize on the symbiotic potential be-
ditions that fluctuate seasonally and diurnally and tween the photosynthetic organisms and the human
have been increasingly hard to predict due to climate users. In closed loop systems, the carbon dioxide
change. contained in the building's exhaust can be utilized to
supply the bioreactor which in turn can provide the
MICROALGAE IN THE ARCHITECTURAL building with oxygen. Algae that is being said to sig-
nificantly contribute to earth's oxygen supply has the
CONTEXT
potential to bring us one step closer to carbon neu-
Photobioreactors (Singh 2012) that are used in the
tral architectures. A building's greywater can serve as
farming of microalgae (Figure 3) for example create a
the water and nutrient supply for algae in the biore-
very specific ecosystem that are slowly finding their
actor. Instead of being considered waste, the build-
way into the urban setting (Hemmrich et al. 2014).
ing's greywater can be utilized to purify water on site.
When the photosynthetic organisms are being culti-
Furthermore, much like the green walls, the algae re-
vated in architectural context a constant and finely
actor panels can serve as a thermal buffer when ap-

plied in a double skin façade system. Besides the po- The fabricated prototypes consist of a test environ-
tential to mitigate the negative effects of unsustain- ment that allows for the monitoring of translucency
able building practices, the bio-enabled façade sys- as well as transparency (see Figure 6). It integrated
tems are being scrutinized for their capacity to aid in two bioreactors that are designed to swivel by 90
the production of renewable energy, or as new future degrees to rotate the reactor into position for maxi-
food and feed supplies. mum light irrigation of the algae medium or to allow
the human users access to views and direct sunlight
CASE STUDY PROTOTYPE at times. Scenedesmus, a genus of green algae was
In a case study conducted at the Material Dynamics tested in this reactor, that requires roughly a 14-day
Lab at the New Jersey Institute of Technology the po- cycle before the algae solution is mature enough for
tential integration of algae in a façade system is be- harvest. In the beginning stages of the grow cycle,
ing scrutinized. The design experiments specifically the algae solution appears nearly clear before turn-
look at the use of flat panel reactors applied in a dou- ing dark green near harvest (see Figure 4) Depend-
ble skin façade system. Integrating this very specific ing on the grow cycle, daylight penetration as well as
ecosystem in the building context has a great im- varying quality of views have been the main concern
pact on the overall translucency and transparency of of this first study.
a building skin. To monitor the environmental conditions, this
design experiment that was conducted under labora-
Figure 6 tory conditions has integrated a series of sensors and
Algae Bioreactor in actuators into the prototype. Four lux sensors are uti-
Test Environment at lized in this design experiment. One is placed in the
the Material interior of the test environment, two sensors are di-
Dynamics Lab rectly located on the reactor panel sensing the light
levels through the algae medium, and a control sen-
sor is located on the outside of the reactor measur-
ing the incoming light before it enters the prototype.
The RGB sensors are monitoring the interior light dis-
tribution of the environment.
The experiment explores a platform that allows
for the integration of the prototype into the Internet
of Things (IoT). This setup is well suited for the nec-
essary sensor monitoring and data logging as well as
device automation and orchestration, to facilitate the
ecosystem controls necessary to maintain the health
of the algae colony. A major benefit of this setup is
that it allows researchers to access the environment
Figure 7
remotely.
Photobioreactor
The platform includes a Raspberry Pi SD-Card im-
Test Setup. Credit:
age that serves as the hub of the system, which boots
Material Dynamics
to a fully configured Linux system. This system pro-
Lab.
vides essential services such as data storage, graph-
ing, and analysis along with an easy to use visual pro-
graming environment. The provided SD-Card image

includes a number of useful preconfigured services: a bandwidth overhead. The various sensors, that in-
web-based visual programming tool (Node-Red [4]), clude lux, RGB, temperature and humidity sensors,
(see Figure 8) a web-based graphing tool (Grafana publish their data to various topics, such as "bioreac-
[5]), a timeseries database (InfluxDB [6]), a communi- tor/1/interior/left/lux".
cation server (Mosquitto [7]), and an auto-discovery Finally, the Avahi auto-discovery service allows
service (Avahi [8]) the sensors to locate the Raspberry Pi computer on
the local network and connect to it. This tool is in-
valuable for a timely setup, as no hardcoded IP ad- Figure 8
dresses are involved. The sensors as well as actua- Node-RED
tors can be simply attached and will start communi- Interface.
cating. All user-facing systems run in a web browser
and can be potentially used on any local computer
or mobile smart device. The sensors and actuators
in the presented prototype use microcontroller sys-
tems that are capable of internet connectivity such as
the ESP8266 [10] system-on-chip. In a future devel-
opment of this prototype setup, a necessary update
will add security and authentication to enable inter-
net access of the graphing and programming tools.
PRELIMINARY TEST RESULTS

Node-Red (see Figure 8) can be described as "a visual Figure 9
tool for wiring the Internet of Things" [4]. It provides a Light Condition in
drag and drop interface to configure the behavior of the Test
the prototype by facilitating the control of actuators Environment
operating the swivel function of the bioreactor. The depending on
various sensors that monitor the environment are Growth Cycle
placed onto the Node-Red canvas, allowing it to di- documented at the
rect its data through various operations like smooth- Material Dynamics
ing or sending it to be stored in a database. The uti- Lab.
lized InfluxDB timeseries database that receives the
environmental data is designed from the ground up
to store and query temporal information. The web
based graphing tool Grafana connects to InfluxDB to
display the live data inside any web browser.
Furthermore, the communication protocol
MQTT [9], which is a publish-subscribe communi-
The data that was collected during this early labora-
cation protocol originally developed at IBM, was uti-
tory testing stage (See Figure 10) was mainly focused
lized. MQTT is particularly well suited for the commu-
on the varying lighting conditions in relationship to
nication between the sensors, actuators, and control
the geometric configuration of the reactors. The pre-
software, since it was designed with low-power, low-
liminary findings show that upon rotating the flat
bandwidth devices in mind and hence has minimal
panels away from the ideal relationship to the light

Figure 10
Bioreactor Test
Data: Credit:
Material Dynamics
Lab.
source an immediate drop of the light levels within DISCUSSION AND FUTURE OUTLOOK
the reactor is observed due to the specific geome- Architectural simulation software could have
try of the reactor frame. Furthermore, the light inten- brought on similar results if the bioreactor was an
sity inside the reactor was not even due to the frame inert and static material. This experimental set-up
details of the flat panel reactor as well as some algal is designed to achieve long term testing of the inte-
residue on the reactor's interior. The conditions on gration of living organisms. The long term studies
the interior of the test environment respond to the that are being currently conducted at the Material
panel rotation at first with a slight drop in the lux lev- Dynamics Lab is observing the growth of microalgae
els due to the opaque framing detail of the reactor, in the test environment, simulating a series of con-
which is followed by a rapid increase once the panel ditions that might arise in the architectural context.
is rotated by more than 40 degrees. Each test includes a full 14-day growth cycle from
The testing cycle was repeated during the vari- incubation to harvest. Various settings that are sim-
ous stages of the algae grow cycle as shown in Fig- ulating southern, northern, and western exposures
ure 10. The collected data clearly showed that the are going to be tested. This will help explore the algal
density of the algae colony had a great impact on the growth rate under the varying conditions and docu-
light levels inside the testing environment. ment the impact on interior spaces simultaneously.
The integration of the architectural prototypes with

the IoT is of great importance to the success of this Figure 11
study. It not only enables an effortless way to collect Microalgae
data it also allows for remote management and mon- Incubators at the
itoring of the lifeform. Even though microalgae are Material Dynamics
considered to be among the fastest growing, most Lab. From Left to
efficient and adaptive organisms on the planet, the Right :
attempts of integrating them into buildings holds its Nannochloropsis,
challenges. Through the remote control and access Spirulina,
to the testing data, immediate action can be taken Scenedesmus,
if the test setup is compromised. This is particularly Callithamnion.
useful since any malfunction of the reactor can lead
to an immediate algae die-off. CONCLUSION
The first round of tests has been conducted with Whether we are integrating smart materials such as
scenedesmus, while spirulina, callithamnion, and hydrogel or living organisms such as algae into the
nannochloropsis will be observed at a later time. The architectural context, we need to become accustom
various species of algae are very different in appear- to testing and verification in our explorations. Smart
ance (Figure 11) and hence will have a very unique materials and living organisms both react to exter-
impact of the interior lighting levels. Furthermore, nal stimuli that arise in the architectural context and
we will look at in situ test setups that will move the are greatly influenced by changing seasons or diurnal
study out of the laboratory testing stage. fluctuations.
In the design toolkit for adaptive and reactive While addressing energy efficiency in the con-
façade systems, great importance has been given to structed environment the design process has been
the maintenance of homogenous interior environ- greatly enhanced by modeling and simulation tools.
mental conditions. Most recently studies that strive However, evidence suggests that one of the major
to include comfort and control of the individual on problems that the building industry encounters to-
the interior have emerged (Krietemeyer et al. 2012) day is that our building stock often does not per-
that will allow users to influence the visual and ther- form as it was expected or designed to (Demanuele
mal quality of the building. By integrating living or- 2010; The Carbon Trust 2011). The so-called perfor-
ganisms into the building envelope the complexity of mance gap between predicted energy efficiency in
the interaction protocols that will negotiate between buildings and actually achieved outcomes has been
all the constituencies will increase significantly. The of increased concern (Menezes et al. 2012; de Wilde
needs of the integrated lifeform will have to be better 2014). With the rise of available automated meter
understood before such an undertaking can be suc- readings in architecture to verify the energy con-
cessfully tackled. sumption of buildings the performance gap is be-
coming more and more evident. But the lack of track-
ing the performance of individual building compo-
nents is contributing to the problem (Menezes et al.
2012).
The integration of individual building compo-
nents with the IoT can be an invaluable tool for ar-
chitects and researchers that will enable the creation
of new knowledge in the field of architecture.
The research that is being conducted at the Ma-

Figure 12
Spirulina - End of
Growth Cycle
documented.
Credit: Material
Dynamics Lab.
terial Dynamics Lab is integrating prototypes into United States Energy Administration, DOE 2014 'Annual
the IoT as a first step in this development process Energy Outlook', Annual Energy Review
that will enable us to address the performance gap Brenneisen, S 2006, 'Space for urban wildlife: designing
green roofs as habitats in Switzerland', Urban Habi-
in complex building systems. Through publishing
tats, 4(1), pp. 27-36
the real time test results in the next testing phase Brownell, BE and Swackhamer, M 2015, Hypernatural :
this approach may become an invaluable tool for re- architecture's new relationship with nature, Princeton
searchers and practitioners and aid in the creation of Architectural Press, New York
new knowledge for the field of architecture. Clayton, PA and Martin, J 1990, Price. The Seven Wonders
of the Ancient World, New York: Routledge
Decker, M 2015 'Soft Robotics and Emergent Materials in
ACKNOWLEDGEMENTS Architecture', eCAADe 2015, Vienna, pp. 409-416
The research presented in this paper has been sup- Demanuele, C, Tweddell, T and Davies, M 2010 'Bridg-
ported by a NJIT URI grant as well as the NJIT Cen- ing the gap between predicted and actual energy
ter of Building Knowledge. The authors would like to performance in schools', World renewable energy
congress XI
thank Dr. Wen Zhang and Likun Hua from the Depart-
Gandy, M 2010, 'The ecological facades of Patrick Blanc',
ment of Civil and Environmental Engineering at NJIT, Architectural Design, 80(3), pp. 28-33
as well as Dr. Eric Fortune from the Department of Hemmrich, E, Blaschitz, M, Scholz, O and Wurm, J 2015,
Biology at NJIT, for their support, insights and advise Algenhaus: \"uber das erste Geb\"aude mit Bioreaktor-
throughout the research. We would also like to thank fassade, Niggli
Andrew Biron for his dedication to this project during Joya, KS, Joya, YF, Ocakoglu, K and van de Krol, R 2013,
'Water-Splitting Catalysis and Solar Fuel Devices: Ar-
the prototyping stage.
tificial Leaves on the Move', Angewandte Chemie In-
ternational Edition, 52(40), pp. 10426-10437
REFERENCES Koo, HJ, Chang, ST, Slocik, JM, Naik, RR and Velev, OD
Abedi-Koupai, J, Sohrab, F and Swarbrick, G 2008, 'Eval- 2011, 'Aqueous soft matter based photovoltaic de-
uation of hydrogel application on soil water reten- vices', Journal of Materials Chemistry, 21(1), pp. 72-79
tion characteristics', Journal of plant nutrition, 31(2), Krietemeyer, B and Godlewski, J 2012, 'The interior ex-
pp. 317-331 perience of daylighting technologies: Histories and
potential futures', Interiors: Design, Architecture, Cul-
ture , 3(1-2), pp. 59-84

Krietemeyer, B and Rogler, K 2015 'Real-Time Multi-Zone Vitruvius, P and Morgan, MH 1960, Vitruvius: The Ten
Building Performance Impacts of Occupant Interac- Books on Architecture, Dover Publications, New York
tion with Dynamic Façade Systems', eCAADe 2015, Vi- de Wilde, P 2014, 'The gap between predicted and mea-
enna, pp. 669-678 sured energy performance of buildings: A frame-
Menezes, AC, Cripps, A, Bouchlaghem, D and Buswell, work for investigation', Automation In Construction,
R 2012, 'Predicted vs. actual energy performance 41, pp. 40-49
of non-domestic buildings: Using post-occupancy [1] https://www.90yearsofdesign.philips.com/
evaluation data to reduce the performance gap', Ap- [2] http://www.materialdynamicslab.com/
plied Energy, 97, pp. 355-364 [3] http://www.verticalgardenpatrickblanc.com/
Nagengast, B 1999, 'Early twentieth century air- [4] http://nodered.org/
conditioning engineering: The first century of air [5] https://grafana.net/
conditioning', ASHRAE journal, 41(3), p. 55 [6] https://influxdata.com/time-series-platform/inf
Niachou, A, Papakonstantinou, K, Santamouris, M, Tsan- luxdb/
grassoulis, A and Mihalakakou, G 2001, 'Analysis of [7] http://mosquitto.org/
the green roof thermal properties and investiga- [8] http://avahi.org/
tion of its energy performance', Energy and buildings, [9] http://mqtt.org/
33(7), pp. 719-729 [10] https://espressif.com/en/products/hardware/esp8
Park, D and Bechthold, M 2013, 'Designing Biologically- 266ex/overview
inspired Smart Building Systems: Processes and
Guidelines', International Journal of Architectural
Computing, 11(4), pp. 437-464
Prokssh, G 2013, 'Growing sustainability - integrating al-
gae cultivation into the built environment.', EAR, 33,
pp. 147-162
Santamouris, M 2014, 'Cooling the cities--a review of re-
flective and green roof mitigation technologies to
fight heat island and improve comfort in urban en-
vironments', Solar Energy, 103, pp. 682-703
Savi, T, Marin, M, Boldrin, D, Incerti, G, Andri, S and Nar-
dini, A 2014, 'Green roofs for a drier world: effects
of hydrogel amendment on substrate and plant wa-
ter status', Science of the Total Environment, 490, pp.
467-476
Singh, R and Sharma, S 2012, 'Development of suitable
photobioreactor for algae production--A review', Re-
newable and Sustainable Energy Reviews, 16(4), pp.
2347-2353
Speak, A, Rothwell, J, Lindley, S and Smith, C 2012, 'Ur-
ban particulate pollution reduction by four species
of green roof vegetation in a UK city', Atmospheric
Environment, 61, pp. 283-293
Specht, K, Siebert, R, Hartmann, I, Freisinger, UB, Sawicka,
M, Werner, A, Thomaier, S, Henckel, D, Walk, H and
Dierich, A 2014, 'Urban agriculture of the future: an
overview of sustainability aspects of food produc-
tion in and on buildings', Agriculture and Human Val-
ues, 31(1), pp. 33-51
The Carbon Trust, CT 2011 'Closing the Gap: Lessons
Learned on Realising the Potential of Low Carbon
Building Design', The Carbon Trust, London

From Interactivity Towards Ambience Through a Bottle-
brick
Kateřina Nováková1 , Henri Achten2
1
Faculty of Architecture Czech Technical University in Prague 2 FA CTU Prague
1,2
{bruhokat|achten}@fa.cvut.cz
According to the dictionary ambient architecture should be kind of object or

space that relies to its surrounding or spontaneously reacts on the presence of
human. Ambient architecture can also be musically expressed [1] or painted [2].
We developed special architectural building units that offer space for
incorporation of intelligence and media for human interaction and for
ambience.We are introducing an object called PET(ch)air made of PET(b)rick
[3], a hollow transparent bottle-brick. The first intention was to generate new
building unit from recycled PET material. Now that we observe its qualities, we
can see it is well prepared for ambient intelligence application, especially in
combination with light. For the purpose of a brick we are transforming old
recycled plastic into new bottle-bricks. Using the bottle-brick as building unit we
build interior objects that are ready to turn spaces into ambient rooms, places
that can be customized by their visitors or spontaneously react on them. Together
with this, we opened a design studio, where students were asked to develop
ambient interior pieces for a special event using the method of learning by doing.
Keywords: Interactivity, ambient architecture, waste reuse , bottle-brick,

PET(b)rick, PET(ch)air
INTRODUCTION surrounds the user regularly and thus may be cus-

Ambient intelligence is a term introduced by Phillips tomized to his needs. The advantage of intelligent
in 1999 as a vision for the year 2020. [4] The sce- equipment of home is that it saves time, which is
nario of ambient intelligence is invisible technolog- usually spent on maintenance or that it extends hu-
ical support of human everyday life. In many cases it man memory with reminders. Regardless the pur-
was associated with the home situation and dealing pose of its application, this technology must inter-
with procedures such as organization of time, tem- act with humans via various media: screens, reactive
perature fitness, and light comfort. Examples may be surfaces, kinects, sensors, and so on, which become
automatic light behaviour according to the daylight, also ambient. Definition of ambient media is used
and home facility adaptation on the inhabitants via in advertising industry and was described by Arthur
voice or gestures. Ambient intelligence has been de- Lugmayr in 2007 as the most relevant combination
scribed as clever technology (Cook et al. 2009) that of message and media available in order to adver-

tise products [6], (Lugmayr 2007). Combining ambi- Short report on bottle-bricks
ent media and intelligent technology needs real envi- The idea to develop a building brick with material
ronments to be embodied in. In addition to the idea used for bottles is at least one decade old. There are
of Internet of things, architecture and design offer many reasons why to attempt this: the main material
relevant platform for implementation of ambience. polyethylene tereftalate (PET) is produced in abun-
Ambient intelligence researchers are aiming for a sit- dance, can be recycled, and is quite strong; many bot-
uation, where the home is a pleasant and beauti- tles end up after consumption thrown away in na-
ful space without visible hills of technologies: wires, ture; areas which face an immediate need for build-
chargers, black screens, computer boxes and HIFIs. ing materials after natural disasters also have a need
[8,9] Also the publication called Invisible future (Aarts for fresh drinking water, and so on. Based on this it
et al. 2001) mentions, that technology will be gener- makes sense to develop a brick that can at first hold
ating a new social landscape, and new type of archi- drinking water and then be used as a building brick.
tecture. Several people have attempted to create a building
brick out of PET material. We found an idea mock-
A building brick for ambient structures up made by Tim Dubitzky, who tried to interconnect
In this paper we report on a building brick that we square plastic bottles, generating a small wall. [5]
used to create ambient structures and its use in a stu- Two working prototypes of blow moulded PET bricks
dio task. Initially the building brick was developed were found in the world very far from each other. (see
as a bottle, which holds consumable drink-water - Figure 1) In 2007 united_bottle [7] was developed by
after consumption it could be used as a stackable a research group from Switzerland. This bottle was
construction element in non-load bearing structures intended to be filled with water and transported into
such as space dividing walls (see below for more de- war/catastrophic regions, where the disposed bot-
tailed description). After we developed the brick, it tles would serve as filling into broken walls. In 2010
immediately occurred to us that we could put intelli- united_bottle was followed by Polli-brick, invented in
gence inside the brick. The advantage of such an in- Taiwan by a company called MINIWIZ (Xu and Wang
vention is that the brick is transparent, light and hol- 2013). This bottle was not used as a beverage con-
low and still withstands stress. Therefore electronic tainer, it only served as a facade panel. It was inter-
devices can be easily included in the interior of the connected by a plate.
bottle. For initial test-use, we designed a small seat
that was able to interact through lights. Our idea was Figure 1
to let people play with the objects that anyway sur- Bottle-bricks from
round them with the use of their own electronic de- the left: polli-brick,
vices and together generate an interior furniture that united_bottle,
changes whenever somebody starts to interact. This PET(b)rick
could be done through a web application, which is
entered by a qr code. The interaction had a form of a
game. We developed objects, that are apart of their At Czech Technical University, Faculty of Architecture,
initial function prepared to become also media and we developed a brick - named PET(b)rick - made from
education platform for interactivity. We are targeting recycled polyethylene tereftalate. Out of the list of
objects that hide naturally interactivity/ambience in known bottle-bricks PET(b)rick is the only one which
themselves. can be bound in six directions and thus it is possible
to assemble it into never-ending structures. The de-
velopment of the brick had to answer several bound-
ary conditions. The most important ones were that

it could be produced in a regular production line for The basic quality of light is that it attracts attention
PET bottles, and that it can be stacked to create archi- and users start playing with the colours manually im-
tectural elements such as walls. For the production mediately. It was easy to operate rotating a simple
we used blow-moulding technology that is the same round button or choosing various modes of change.
as for regular PET bottle production. In the future, its Led lights being ruled by an application via Internet
price will be the price of a bottle. Because of its par- had the advantage that lights could be changed re-
ticular shape, the PET(b)rick can be bound in all three motely -even from far away- meaning that the system
cardinal directions - therefore it allows the construc- was independent of the position of the user. Visitors
tion of planes and volumes. The PET(b)rick bottle is were able to play more sophisticate game. More than
7x14x9 cm big and has volume of 0.6 l that can be one cube could be controlled at once. The constraint
filled with various media and technologies. was that there had to be a smart device with qr code
reader in order to master the change of colour. Also
From PET(b)rick to PET(ch)air WIFI had to be distributed by access point device.
The advantage of a structure made of hollow trans-
parent units is that these units can be filled either From PET(ch)air to PET(b)ar
with material or media or technology. The shape An experimental studio topic was prepared on the
is original [3] and allows for connecting in all 6 di- base of this research: "an ambient bar." Students
rections. Any rectangular shape can be built from were asked to use the developed technology and in-
this bottle-brick, the final product must be bound to- vent new principles of ambience in architecture. The
gether and the vertical force must be distributed. Be- aim of the project was to design and build a num-
ing internally lit and equipped with batteries, proces- ber of stage sets and objects for the so-called "stu-
sors, and WIFI antenna it can join Internet of Things, dent's night" at the Faculty of Machine Engineering
waiting for instructions or respond to the presence of of Czech Technical University. This event takes place
people. on several locations of the faculty building, and in-
cluded objects such as a bar, stage, seating area,
Figure 2 corridor decoration, lighting, and so on.The whole
Interactive and project would have to run in a very short time. From
ambient PET(ch)air the start to design and full realization of all the ob-
exhibited at the jects for the "student's night" event was only one
EXPO in Milano. month. Through the knowledge gained by making
The PET(ch)air (see Figure 2) was a small cubic struc- the PET(ch)air we could introduce the task of the am-
ture bound together with nylon threads and covered bience to a group of students who had never worked
with a Perspex grid (see Figure 2). We constructed a with this material before. We prepared a cooper-
number of these chairs and experimented with var- ative seminar for students of humanistic discipline
ious technologies to incorporate in the PET(b)ricks of New Media at SNM Charles University in Prague
that make up the chairs. Interaction with PET(ch)airs together with students from Faculty of architecture
was tested with the function of light control at the at CTU Prague. Participants of this pilot project got
EXPO in Milano (May-October 2015) and TEDx Prague the PET(b)ricks and the hardware (plus sensors on re-
(date) where peoples' enquiry and interest was ob- quest) in order to be able to directly test their ideas
served. Visitors were allowed to change colours in reality. The result of their cooperative work was
of the objects manually or via their smartphones immediately used within a public event, where they
through Internet application. Both systems had their could observe the installation in real time interaction
advantages and both failures, but once an object with lay-users. With this experiment we hope to ob-
shines, people look for the possibility to operate it.

tain answer to several questions: does perceiving of Table 1
ambience differ in both branches? What are the de- Ideas from non
mands? How much are both fields complementary? technical students.
And will the output fulfil the request of ambience?
Analysis phase. The experiment with the use of
PET(b)rick plus lights and sensors started by intro-
duction about interactive bars, that we found as ex-
amples on the Internet. We tried to introduce di-
verse types of interactive technologies, digital as well
as mechanical. Spintouch Interactive Solutions bar Table 2
was found as an example of interactivity. [10] The in- Ideas from students
teractivity was performed by a multi-touch surface, of architecture.
which reacted on any kind of weight laid on the table.
Students were inspired by this system, and wanted
to use principle of stress although they had no pos-
sibility to use any screen. In 2015 MIT introduced
TRANSFORM. A transformable mobile furniture was
developed by Professor Hiroshi Ishii and the Tangi-
ble Media Group from the MIT Media Lab. [11] Stu-
dents could see the transformation of data into me-
chanical movements in this example. Ambience was
described on an example of Phillips research as well
as their precedent Hiroshi Ishiis' AmbientROOM (Ishii
1998).
Ideas from non-technical students (see Table 1)
were based on analysis phase. Students from the
Charles University had different attitude than archi- The process of designing. We used the method of
tecture students. They did not derive their ideas learning by doing and the agile designing, which
from material possibilities, but came up with rather meant that short phases of designing switched with
megalomaniac ideas of transmitting of big data (like hands-on work with PET(b)ricks. Students were given
videos) into installations. Their ideas were rather de- the material and conditions and we observed them
tached from real possibilities we had. We can also designing. It followed the principle of Experimen-
see, that their ideas were mostly interactive and play- tal studio (Nováková et al. 2010), apart of that each
ful rather than ambient, since the installations were phase lasted maximally for one week:
only reacting on visitors conscious activity.
Ideas of ambience from technical students (see 1. Analyses
Table 2) were rather simple and connected to the real 2. Design
objects. The ambient or interactive activity was at- 3. "Selling the idea to the investor" - presenta-
tached to the designed interior installations and de- tion
rived from their form. All overall activity was per- 4. Building prototype
formed by the change of light intensity, flashes or col- 5. Public exhibition
ors of light. Students could also try out their ideas
immediately. Building prototypes (Figure 3) formed their designs.
Students were designing while constructing proto-

types, because the material they were using was un- were different due to the last minute changes of the
known. actual design at the site (see Figure 4).
Figure 3 Discussion. In order to see, what knowledge stu-
The process of dents gained within this studio, we created a Ques-
hands-on tionnaire, where we asked, what was the biggest con-
designing. tribution of such a studio and we asked students di-
rectly, what they think they learned. From the to-
tal number of 42 students we received 34 responses.
In this quantitative part of qualitative research we
Figure 4 searched for following keywords: First time/mew, Ex-
Design and real periment/invention, Build own design, real event, in-
installation of one tensive, material / reuse, group-work, interactivity.
of the groups called No ambience appeared among the answers. (see Fig-
DJs. ure 5)
More than 1/3 of respondents appreciated the
rhythm of the studio: that it was short and inten-
sive and the same number of students mentioned
the positive effect of group-work. The possibility
of building own design was also highly appraised
Figure 5 by 30% of respondents. 25% of students men-
Results of the tioned they learned about interactivity and material
questionnaire: 32 reuse. Although we explained the ambience as nat-
respondents were ural reaction of the architectural objects on people,
asked what they students kept calling their installations interactive,
learned. Repeated even though some of them were really rather ambi-
keywords were ent, giving the users spontaneous information about
collected and something.According to the questionnaire, the de-
quantified. sign phase of the projects was underestimated - stu-
dents experienced it as too short. Students had to
embody lights and sensors into the structures from
the beginning of the construction process. The ob-
jects had to be deconstructable and reassemblable,
which had the consequences on the connection of
The process of designing lasted for 10 days and then the lights inside of the designed object. Because of
ready designs were presented. All of them were ac- the technical challenges concerning electronics, con-
cepted and students started constructing their instal- sultation with programmers and people from Faculty
lations. The event took place at the faculty of Me- of Electrical Engineering was inevitable and highly
chanical Engineering, but the designs were built at valued. The students unanimously appreciated the
the Faculty of Architecture, which meant they had to teamwork and learning how to organize themselves.
be deconstructed and transported. The actual event Many of them started to appreciate the concept of
lasted only for eight hours and all of the installations ambience and interaction as this was the first time
were fully functioning there. The real objects looked they were confronted with it. Students also men-
always quite like the designs, usually the light effects tioned that the time for designing was too short.We

also pointed out comments, which were beyond the came very clear that the help of students from the
standard answers. Faculty of Electrical Engineering was crucial. With-
• "I felt responsible for my installation and had out their assistance students of architecture were nei-
in mind it must be secure and withstand the ther able to generate concept nor able to manually
users activity throughout the evening." put it together. There is a thin border between in-
• "For me, doing experiments is important for teractivity and ambience. Students were not able to
my learning process, maybe for a learning pro- understand it within one month of the studio, and
cess generally. I changed my attitude in my they called their installations interactive. The con-
regular studio too." cepts of ambience were rather complex in the begin-
• "Designer must be ready for change." ning and radically simplified when it came to their
• "An person becomes an architect, when real implementation. We could see that many stu-
he/she realises first design." dents have never been thinking about implementing
• "I saw how important it is to see the dimen- neither ambience nor interactivity into their architec-
sions of my design in reality, it was bigger than tural designs and they were also building a real object
I thought." they designed for the first time. The studio had two
"wow effects":
By developing PET(b)rick, a building unit of a new
• I can build my own design, it is working!
type which is flexible but withstands stress, we have
• Implementing ambience/interactivity is cool,
prepared a platform for easy implementation of inter-
but not easy.
activity and ambience. We tested interactivity of our
chairs at several public events and we could observe In further work we will surely continue to explore am-
visitors' reactions in confrontation with objects they bience and interaction in architecture. In order to
could operate. The software/hardware combination avoid overly simplistic action-response schemes, it
allowed for playing with the colour of light, the actual is of vital importance to work together with people
form of interactivity embodied in lightweight mov- from other disciplines such as Electrical Engineering.
able objects such as chairs. We designed, built, and In this sense we found that the rather limited scope
stress-tested a simple transparent cube, where the of the studio also gave the students valuable experi-
PET(b)ricks have proven their qualities as bricks as ence for future practice.
well as technology containers.
Acknowledgements
Conclusions We want to thank all students who participated in the
From the experience with the ambient bar, we can design project, the dean of the Faculty of Mechanical
state that the most significant advantage of doing re- Engineering, prof. Valášek, and mrs. Špačková and
search with real objects is that everything can be di- mrs. Götzová of the PR department of Mechanical En-
rectly tested and we did so. We found out, that us- gineering for inviting us to také part in the Student's
ing smart devices in order to operate simple func- Night and for financial support; the dean of the Fac-
tions is a constraint for the user. The next step was ulty of Architecture, prof. Labus, for allowing assem-
application of ambient intelligence into bigger struc- bly at the Faculty grounds, the assistants who helped
tures built from PET(b)ricks We improved the design in the studio (Lucia, Vasilija, and Šimon), and the peo-
from boxes into more sophisticate objects where the ple from Faculty of Electrical Engineering and Build-
technology was hidden and the objects became am- ing Engineering (Kuba, Jedle, and David).
bient intelligent, in a very simple and limited way.
Students of architecture and from Charles University
were able to work with the PET(b)ricks, but it also be-

REFERENCES [11] http://tangible.media.mit.edu/project/transform
Aarts, E, Harwig, HH and Schuurmans, M 2001 'Ambient /
Intelligence', The Invisible Future, New York, pp. pp.
235 - 250
Cook, DJ, Augusto, JC and Jkkula Vikramaditya, VR 2009
'Ambient intelligence: Technologies, applications
and opportunities, Pervasive and Mobile Comput-
ing 5, 2009.', Pervasive and Mobile Computing 5, p.
277 – 298
Dennig, PJ (eds) 2001, The Invisible Future, McGraw-Hill
Ishii, HH 1998 'AmbientROOM: Integrating ambient me-
dia with architectural Space.', CHI 1998, pp. 173-174
Lugmayr, A 2007 'Ambient Media, Vol. 33', Novatica, pp.
pp. 35-39
Nigel, S 2003 'Ambient Intelligence', IEEE intelligent sys-
tems, p. 2 – 3
Nováková, K, Achten, HH and Matejovská, D 2010 'A De-
sign Studio Pedagogy for Experiments with Unusual
Material, Collaboration and Web Communication,
2010, roč. 8, s. 557–572, ISSN 1478-0771.', IJAC - In-
ternational Journal of Architectural Computing, Multi-
Science Publishing Co Ltd
Weiser, M (eds) 1993, Ubiquitous computing, IEEE
Xu, JP and Wang, JP 2013 'Ground improvement
with interlock-Able Plastick bottles', International
Symposium on Design and Practice of Geosynthetic-
Reinforced Soil Structures, Bologna, Italy, pp. 24-33
[1] https://soundcloud.com/ambient_architecture.
[2] http://risdmuseum.org/art_design/exhibitions/63
_
[3] https://isdv.upv.cz/webapp/WEBAPP.vzs.det?xprim
=90026377770001&lan=cs&s_majs=&s_puvo=
&s_naze=
[4] http://www.research.philips.com/technologies/am
bintel.html
[5] http://www.core77.com/posts/23643/building-a-bo
ttle-brick-at-glasslab-a-case-study-by-tim-du
bitsky-23643.
[6] http://www.computer.org/csdl/mags/ex/2003/04/x4
002.pdf.
[7] http://www.united_bottle.org
[8] https://www.youtube.com/watch?v=VjF-kapcFsU&ebc
=ANyPxKof2b5OmqiTyBl0m0_ZvbchAy_oX5yY6cSg8DHO
YIMbV_CRMFJB9qroXEz_nrxbEFoC0QARYkBP5I0ddfWFS
0AyCSykmQ.
[9] http://www.youtube.com/watch?v=lCARHatJQJA&feat
ure=iv&src_vid=lvtfD_rJ2hE&annotation_id=anno
tation_4171896729.
[10] http://www.spintouch.com/interactive-solutions/
hardware/interactive-bar/

Bending-active Structures
A Case study for an Office Chaise Lounge
Eleanna Panagoulia1 , Simon Schleicher2

1
Foster + Partners 2 College of Environmental Design (CED), University of Cali-
fornia, Berkeley
1
eleannapan@gmail.com 2 simon_s@berkeley.edu
This paper seeks to explore the process of elastic bending in furniture design and
presents a case study that demonstrates the creative and structural potential of
bending-active structures as possible improvement to the current state of the art.
This case study brings together design procedures, borrowed from declarative
design in software engineering, architectural design, and material science in
order to envision new applications for bending-active structures. It investigates
how bending can be used strategically for the design of furniture scale objects
and, particularly, an office chaise lounge for one person. Active-bending
implementation is the key for creating structures that achieve new milestones
beyond the perceived limits of material and process. Moreover, the project stands
as a great opportunity for the development of a pipeline for fabrication that
automates the translation of a given high-level description of a design, to the
production of the data required for fabrication via a particular material system.
Keywords: Bending-active structures, Matter compiler, Optimization
INTRODUCTION straints. Designers are able to integrate those factors

Architectural practice has been significantly influ- already in the early stages of the design process by
enced by the development of cutting-edge design simulating forces, structural performance, and mate-
tools that have set the scene for new ways of under- rial behavior accurately. This paper will focus on ac-
standing complex forms, analyzing their constraints, tive, elastic bending in plate structures and will ex-
and optimizing their performance accordingly. Lat- plore the formal and structural possibilities that are
est improvements in software and the increasing ac- available through bending, both as a form-finding
cessibility of design tools have allowed designers to method, as well as a strategy for achieving challeng-
shift from a purely experimental state of extensive ing structural performance. Elastic bending in plate
trial and error, towards a more scientific integration structures has already been implemented at various
of design constraints to enrich the design process. scales. In this paper we will examine the principles
Current design tools allow for additional parame- of elastic bending and its benefits at furniture scale
ter integration beyond formal description, such as design, by analyzing a case study for an office chaise
physics, material properties, or environmental con- lounge. In order to provide an informative frame-

work for our research on elastic bending in furniture enabling it to be bent in multiple directions. Then,
design, we will analyze the groundwork of bending the plywood is let to cool down in order for the defor-
in furniture design through a series of examples that mation to become permanent. Unlike natural wood,
operate as precedent studies on the field. the resulting material does not have the same ma-
terial properties at the final state, as it is less likely
BENDING IN FURNITURE DESIGN to split when exposed to moisture. The shipbuild-
A major innovation that designated substantially the ing industry was an early adopter of this technique
design of mid 20th century and beyond, was the infor forming the curved members used to build ship's
troduction of bending in furniture design. The first hulls. This paper will portray three examples of heat-
method that was implemented in furniture and par- formed furniture design, two of which are considered
ticularly chair design, was bending by means of heat- to be primary pioneers in the mid 19th and 20th cen-
forming. Heat-forming enabled the creation of nat- tury and a third example that represents its contem-
ural and organic shapes that were refreshingly dif- porary evolution nowadays.
ferent in their aesthetic to the previously used mod-
ernistic standardized forms. The strategic use of Examples of heat-formed furniture
bending made it possible to fabricate furniture with
Figure 1
less pieces, which had great impact in preserving the
Matthias Pliessnig,
continuity and fluidity of the overall form. However,
Amada bench [3].
bending was not only implemented to increase the
aesthetic value of the design. Moreover, many de-
signers recognized that bending as a novel fabrica-
tion technique was a real paradigm shift that had the
power to revolutionize furniture industry by signifi-
cantly reducing production costs.
Heat - Forming Bending Plywood Charles and Ray Eames, Lounge Molded Chair
One of the most suitable materials for heat bending (LCW). The first example is the Chaise Lounge and
applications was plywood. Plywood veneer offered the Molded Plywood Chair by Charles and Ray Eames.
designers the flexibility not only to explore more The Eames have pioneered in inventing strategies
complex forms but also allowed for manual post- to craft complex forms. They experimented with
processing and premium quality finish. Another ad- wood molding techniques and developed strategies
vantage of using plywood bending was the possi- to achieve curved complex forms out of planar, thin
bility to accommodate the comfort factor, as these sheets of plywood. One of their main design drivers
shapes were friendlier to the human body geometry was to develop a compound-curved chair that ei-
and usually had nice tactile qualities. The most com- ther consisted out of very few or even a single shell
mon heat-forming process was steam bending of a single-shell. To accomplish this goal, the Eames in-
dried piece of lumber or a multilayer arrangement of vented their own machine to press large-scale ply-
plywood veneers that are bound together with glue wood sheets, which enabled for example the mold-
with their grains perpendicular to each other. The ing of unprecedented large aircraft parts (Eames
shape is given when the structure is bent under pres- 2012). Prior to furniture, the Eames took advantage
sure with heat. Here, hot steam was a particularly of bending characteristics (elasticity, resilience), to
suitable medium since it was able to penetrate deep manufacture stretchers and crutches for the World
into the plywood to momentarily soften its fibers and War II. Later on, they implemented a similar strategy

in furniture design and their pioneer methods had gonomics to suit the human body, as well as evident
a profound effect in the way future designers per- beauty and elegance.
ceived form and matter. However, due to the lim-
its of the plywood itself, they were unable to make ELASTIC BENDING AND BENDING-ACTIVE
a single-piece shell out of molded plywood as origi-
STRUCTURES
nally intended. Only years later they found a way to
Figure 2 produce the single-shell chair they were striving for,
Digital simulation by using a new material that has never been used in
of bending process furniture before: plastic.
in Kangaroo Physics Matthias Pliessnig, Amada Bench. The second ex-
and SOFiSTiK ample is the steam bent benches of Matthias Pliess-
(Schleicher et al. nig (Figure 1). Pliessnig reframes heat bending
2015). through the lens of contemporary digital craft and
he is interested mainly in human proportion integra-
tion. His design is characterized by minimal shapes While the key concept behind the previous examples
inspired by wave flow and fluid dynamics in gen- is based on bending as means of plastically deform-
eral. Pliessnig's design combines craftsmanship and ing a structure, the following case study will intro-
comfort, structural stability and strength with sculp- duce and investigate the alternative of elastic bend-
tural finesse [3]. As Pliessnig became familiar with ing as active and holistic forming process for shape
surface modeling tools, such as Rhinoceros, he dis- exploration. Moving away from steam bending appli-
covered similarities in the way the geometry is de- cations, we will now investigate elastic active bend-
scribed within the digital environment and the way ing as a holistic process that addresses form explo-
a boat scaffold is constructed. Same as is in boat ration, as well as structural performance integration.
shell constructions, there are leading ribs that des- The term "bending-active" is introduced by Knippers
ignate the shape as cross section elements and sec- et al. to describe curved beam and surface struc-
ondary curves that bend along the ribs, forming a tures that base their geometry on elastic deforma-
curved grid. Pliessnig implemented the exact same tion of initially straight or planar elements (Knippers
logic in his bench design. The workflow that he de- et al. 2011). Active bending can be defined as a
veloped starts from a digital model and then contin- form-finding process that derives from the elastic de-
ues with constant refinement of the heat bent pieces formation of a rod or plate structure. It is a process
until a perfectly smooth surface is achieved. He starts that creates curved geometry out of planar, straight
by drawing a free-form shape and then contouring members or surfaces (Figure 2). One of the key as-
it in temporary wooden ribs that are CNC cut and pects of bending-active structures is that they devi-
put along the target outline curve. Then he bends ate from the existing structural typologies, such as
temporary wooden slats, which he calls "Sketching space frame structures, planar or curved trusses etc.
Stringers" [3], because they operate as guides to visu- As Lienhard et al. explains, bending-active structures
alize the overall shape. After replacement of all tem- are distinguished as an approach rather than a dis-
porary pieces, he steam bents final slats until they tinct structural typology (Lienhard et al. 2014). This
form a continuous surface. Lastly, he strengthens is because they allow for heterogeneous configura-
the structure with epoxy and pins at the intersections tions and nonlinearities in their load bearing behav-
and sands the surface in order to become as smooth ior. The latest advancements in Finite Element Anal-
as possible. Pliessnig's benches clearly address the ysis software (FEA) (Lienhard et al. 2014) allow en-
notions of variability and fluidity combined with er- gineers to analyze and simulate structures beyond

the existing typologies with a high level of accuracy The Graz chair by CODA office (Figure 3) is an out-
in the load calculation. Bending-active structures of- standing example of elastic bending plywood and
fer a wide field of experimentation on non-standard has operated as a precedent for our case study. The
load configurations, as they derive their complex original design is a revision of the Medea chair by Vit-
geometry from elastic deformation processes. Fi- torio Nobili that suggests improvements in the fabri-
nally, in comparison to the previous existing struc- cation methodology and the quality of the plywood
tural typologies, in the resulting lightweight struc- layers. The design suggests a flat, rectangular piece
tures, bending is not avoided but instrumentalized of plywood that carries some CNC smart slits. When
to create complex curved geometries on the basis the naked edges of the slits are overlapped and con-
of standard, semi-finished building products. A par- nected, the shape of the chair is formed. The ground-
ticular advantage of this approach to construction breaking idea of this design is that it achieves dou-
(design approach to bending-active structures based ble curvature without requiring expensive, special-
on elastically deformed members) is the opportunity ized equipment or craftmanship as the previous ex-
to continue using conventional flat-based manufac- amples. It also allows for customization and adapt-
turing processes but in an unconventional way. As ability in a wide range of target shapes. These princi-
usual, parts can be produced on CNC routers, laser pals influenced our design and motivated us to push
cutters, and waterjets in a cheap, quick, and reliant the boundaries further, explore more complicated
manner. As is newly introduced with my research, shapes and achieve stability by a single operation
the materials then get bent and coupled together method (in our case, elastic bending of flat surfaces).
to form highly complex origami-like structures. This
fabrication method provides a very time and material Bending-active as a form finding process
efficient alternative to traditional construction pro- Form-finding is generally understood as the pro-
cesses since it is neither depended on the fabrication cess of developing the geometric form of a struc-
of expensive molds nor on the auxiliary support of ture based on mechanical behavior (Lienhard et al.
complicated formwork. Before moving on to the case 2014). This means that the resulting form is desig-
study that this paper mainly focuses on, we will exam- nated by a particular set of loads that operate on
ine an example of elastic bending plywood in chair it within a constrained set of parameters (material,
design. scale). In bending-active structures, since the form
derives from an elastic deformation of a planar shape,
Elastic bending of plywwod, CODA Graz the resulting geometry is hard to predict with high
Chair accuracy without a real time bending simulation.
Computational tools such Kangaroo Physics [7], or
FEA software allow the user to apply loads on a struc- Figure 3
ture and simulate the deformation in real time and Graz chair, CODA
therefore; observe the result as well as the interme- office [6].
diate steps of the simulation. Having real time feed-
back from the process, the user is able to better un-
derstand the relationships between the variables of
the system and adjust them until the desired result
is achieved. Although this process offers the user
freedom to explore new forms, it is worth mention-
ing that form-finding result does not necessarily cor-
respond with the most efficient load-bearing solu-

tion. Hence, the main challenge is to optimize the set the building of a system with contributing variables,
of variables to a functioning relation that addresses such as forces, material properties, human propor-
both aesthetic qualities, as well as sufficient load- tions etc. where all the processes are part of a single
bearing capacity and stability. Furthermore, to suc- workflow that transfers data from one phase to the
cessfully create a bending-active system it is impor- subsequent. The benefits of a continuous workflow
tant to incorporate material properties, as well as to lie on the fact that the designer operates at a high
experiment with physical modeling. We will examine level set of intentions for the design, such as overall
the contribution of physical scaled models in the fol- size, number of users, seating posture etc. and the
lowing sections of the paper. system translates them in a set of instructions for fab-
rication.
HOLISTIC DESIGN PROCESS
Based on what was mentioned above, this section Accommodating comfort
will describe the design process that was imple- The second strategy that defined the design process
mented in the chaise lounge case study. The goal was that of body conscious design (Cranz 2000). One
of this case study is to achieve an optimum balance of the objectives of this case study was to improve
between comfort, stability, easier material handling, the basic configuration of the conventional seating
weight and cost. To achieve that we will start from posture. To achieve that, we analyzed seating pos-
identifying two strategies, in which the design pro- tures of existing furniture and ended up in choosing
cess was based on. the chaise lounge posture as one of the most com-
fortable and with the best stress and load distribu-
Matter Compiler tion for the body (Figure 4). Lounging, which can be
The first strategy of the design process suggests the defined as halfway sitting and lying down, takes load
construction of a continuous workflow that is called off the spine neck and head. After analyzing in detail
"Matter Compiler". This approach borrows some of designers' chaise lounges as precedent models and
the principals of declarative design [1] in software as role models for the correct angles that fit the hu-
engineering. It promotes a pipeline of linked pro- man body, we ended up isolating the main variables
cesses that are automated in order to reduce large that are associated with the comfort factor. These
manual effort and enhance the search for better and variables are, the points of contact with the human
more efficient designs. The matter compiler suggests body (arms, hands and legs), the alignment and sup-
port requirements of those points and the angles that
Figure 4
Angle study of
precedent designs,
Final pieces of the
prototype.

Figure 5
Prototype chaise
lounge built out of
1:16" thickness
HDPE sheets.
are formed between the points of contact. lar shapes out of planar surfaces and examine their
structural performance. The method we resulted in
CASE STUDY: BENDING CHAISE LOUNGE is called pinching and it is described as the process
This section will delve into a more detailed analysis of strategically removing an internal part of a planar
of the processes that are engaged in the case study, surface and connecting the naked edges of the cut
from form-finding methodologies to fabrication and out shape, allowing the surface to bend accordingly
assembly. in three dimensions (Figure 6). The main motivation
for the pinching method lies in the simplicity of pro-
Form-Finding method: Pinching ducing complexly curved elements. This process ap-
pears to have two benefits: first, it produces intri-
cate shapes out of a single planar surface; second, it Figure 6
creates three-dimensional shapes that perform struc- Pinching as a
turally in the bended state. form-finding
technique.
Matter Compiler: Steps and Processes
Figure 7
Matter Compiler
steps.
This case study approaches bending-active struc-

tures topic under the lens of minimal material em- As stated above, the pipeline used for this project
ployment, in order to prototype a chaise lounge out suggests the use of a "Matter Compiler" that trans-
of the minimum amount of bended surfaces with the lates high-level descriptions of a set of intentions or
highest structural performance. A geometrical inspi- the design to low-level instructions for digital fabrica-
ration that gave direction to the form-finding pro- tion and construction documentation. This method
cess was the exploration of conical bending. Inspired is constrained by a particular material system, a spe-
by Plücker's conoid as reference geometry enabled cific method of joining/assembly and a specific pro-
us to formulate a methodology for achieving simi- cess of fabrication. Our compiler suggests the follow-

ing main steps (Figure 7): tance from each other that overlap completely and
are connected with blind rivets. Having said that, it
• Form-finding and Geometric rationalization is apparent that given a modification in the length
• Joints and seams handing of the edge, due to form adjustment, the tabs and
• Layout automation - Fabrication the hole pattern adjust accordingly. For the riveting
• Assembly Instructions process we used steel rivets and washers, as well as
a rivet gun. The fourth and final step of our compiler
The following paragraph describes the workflow in is the layout automation that provides the data for
terms of types of processes, relationships and in- fabrication. Since the main goal of this project was
cluded variables. The first step involves the Form to achieve best structural performance, while using
Finding process that was rendered through the the minimum amount of joints and pieces, the lay-
pinching methodology that was described above. out automation process is a straightforward 2D nest-
This involves a parametric model of the planar suring process of few planar outlined shapes in 4ft. by
face with the cut out shapes. The first simulations 8ft. planar sheets.
were done in the Grasshopper plug-in Kangaroo
Figure 8 Physics [7] and later in the Finite Element Analysis
Digital simulation in software SOFiSTiK [5] (Figure 8). The Finite Element
FEA software Analysis produced more accurate results, as we were
SOFiSTiK. able to calculate the stress and loads along the sur-
face and observe the stress distribution. Along with
the digital simulations, a large part of the form ex-
ploration was carried out with physical models. The
physical experiments were scaled models of either
paper or the actual material in smaller thicknesses
that were created for actual material testing and for
cross-referencing the results of the digital simulation.
The second step involves geometric rationalization of
the outcome shape, in order for it to adapt to the hu- Materialization
man body proportions and ergonomics. We created In this section we will analyze all the contributing
different families of cut out shapes along the planar processes from design to construction of the chaise
surface that correspond to different angles when the lounge. We will divide the processes in two cat-
surface bents in three dimensions, hence this leads egories, high-level and low-level intentions. High-
in achieving postural variation. Based on the anal- level intention processes are those, which affect pri-
ysis that was carried through in precedent chaise mary properties of the design, such as comfort, aes-
lounges, we depict target angles and proportions thetic, size, proportions, stiffness, stability etc. Low-
that lead to a comfortable seating posture. The third level intention processes are those, which are respon-
step consists of the handling of joints and seams. In sible for the production of the prototype, such as, de-
our case, the joining of the naked edges is by the tailed shape refinement, tab creation for assembly,
technique of riveting and particularly, blind riveting. hole pattern for the rivet joints and data for fabrica-
The same method is used to join multiple surfaces to- tion. Low-level intention processes translate the dig-
gether. Rivets were preferred because they can sup- ital, optimized model into a set of components that
port tension loads, as well as shear loads. Riveting is are constructible by the chosen fabrication equip-
achieved by the creation of parametric tabs along the ment within a set of feasibility constraints.
given edge with a number of holes at a certain dis-

Form development and Optimization. Investigat- tain dimension ranges that would cause material fail-
ing further the technique of pinching, which was de- ure. The material system chosen in this project is 4ft.
scribed above as a form-finding strategy, the further by 8ft. high-density polyethylene (HDPE) sheets of
development of the design was to define the cor- 1/16" thickness with a Young's Modulus of 1200N /
relation between angle of cut out shape and angle mm2 . The material itself, when in the flat state, did
of bended shape. In that way we would be able not demonstrate characteristics of stiffness, however
to test the ergonomics, angles, lengths and propor- when bent, it provided enough stability to hold the
tions and make the necessary adjustments. The ini- weight of one person. The simulation aided us to best
tial test models included symmetrical shapes (circle, utilize material resilience and strength when bent,
ellipsoids), in order to assess the relationship of the by providing numerical feedback regarding the loads
2D shapes with the 3 dimensional bent forms (Fig- and stresses along the bent shape. As mentioned
ure 6). The relationship between the elements was before, along with the digital simulation, we carried
linear for most cases, meaning the wider the angle out multiple physical experiments. Experimenting
of the cutout, the smaller the bending radius was, with physical models speeded up the optimization
causing higher deformation to the surface. Once a process, because it allowed an instant understand-
principal model was developed, the variation of an- ing of the load and stress distribution, as well as the
gles and proportions allowed for fine calibration of association of the variables with resulting geometry.
the geometry. The modifications concentrated on However, scaled models tended to appear stiffer than
adjusting the angles of the cut out shapes, so that the they actually operated in 1:1 scale and the same type
bended outcome shape is comfortable and adapted of material demonstrated different properties in dif-
to the human body. Hence, the outlines (exterior, in- ferent scales. This condition had to be studied care-
terior) of the planar surfaces were optimized to re- fully in the digital simulation in order to make sure
spond both in terms of functionality; the object has that the material used in the prototype demonstrates
to perform as a chaise lounge, as well as in terms the desired properties.
of structural stability; the object has to support the Low-Level processes. The Low-level process in-
weight of a person without material breach. The final cludes four processes. The first process includes the
shape of the chaise lounge was determined by sim- refinement of the bended shape by trimming and ad-
ulating the planar surface with cutout shapes in the justing the outline of the bended shape in order for it
right locations and sizes, with a given load set. The to meet the desired shape expectations, such as fillet-
bending simulation was processed in the Finite Ele- ing sharp corners and removing unwanted material
ment Analysis software, SOFiSTiK, following the elas- from areas that do not affect the structural stability of
tic cable approach. This approach involves defining the chaise. Since shape refinement can only happen
the fixed and sliding points of the geometry, as well when the surface is bent and the chosen method of
as contraction cables that join the naked edges of the fabrication operates in 2-dimensional shapes only, an
planar surface causing elastic deformation. The sim- additional process is required. This process involves
ulation terminates, when the cables' length reaches the re-flattening of the bended surface, in order to
the value of zero. get the exact outline of a planar surface for fabrica-
Material depended geometry. Along with the form tion. This process is necessary because the bended
development strategy to address comfort, the pro- shape cannot be predicted before simulation; there-
cess of high-level description of the design involves fore the refinement takes place after the bending
material property integration, such as the minimum process. However, since our fabrication method ap-
bending radii and the Young's Modulus of the mate- plies only to flat surfaces, re-flattening the bended
rial, in order to predict and exclude in advance cer- surface after bending is required. The second process

suggests the creation of tabs along the edges of the fore the bending so that they are within the range of
interior cutout shape that are important for connect- the bending radius of the HDPE polymer sheet. The
ing the edges together. In this project we explore the fourth and last process is nesting the pieces in 4ft. by
limitations of this technique in terms of dimension 8ft. HDPE sheets for fabrication. The overall shape
constraints, as well as the maximum material thick- of the chaise lounge is composed by 2 bended sur-
ness where rivet joining is still effective. The tabs had faces connected in 3 critical areas, in order to provide
to provide sufficient surface area for the rivets to be strength to the structure, as well as a third smaller
placed in a way allows equal stress distribution along surface that creates tension in the whole structure,
the tab surface to avoid tearing. Also, i is important preventing it from self-load deformation. The fab-
to mention that the calculation has to be as exact as rication method that is most suitable for this mate-
possible so that when all the tabs of the structure, rial, in terms of cost, speed (1 hour setting up and
the hole pattern will much so that the rivets can pass cut time) and accuracy is cutting in the "Zund" blade
through the holes. Any discrepancy in the simulation cutter (tool: blade Z10, setting: multi-pass, 3 passes).
or the generation of the tabs would cause difficulties (Figure 5, Figure 9)
in the assembly process. The third process involves
removal of material in areas of high stress concentra- CONCLUSION
tion. Based on the minimum-bending radius of the The objective of this paper was to provide general
material and the stress information from the simula- insight into form-finding and structural analysis of
tion we were able to allocate areas of high stress dis- bending-active structures. The work aimed to ex-
tribution that would cause the material to tear. To plore the potentials of bending-active structures in
avoid that, we strategically removed those areas be- furniture design as an approach in generating new
Figure 9
Prototype chaise
lounge built to of
1:16" thickness
HDPE sheets, proof
of concept.

forms and structural strategies. Briefly, the parts of team work of Cindy Hartono, Fei Du, Shima Saheb-
the case study that were considered successful were nassagh, Eleanna Panagoulia under the supervision
the accurate translation of software results into fab- of Prof. Kyle Steinfeld Prof. Simon Schleicher, Prof.
rication, as well as the accurate simulation of angles, Jonathan Bachrach and Luis Jaggy and generously
forces and material property integration that resulted supported by Autodesk. The authors want to partic-
in a comfortable chaise lounge that can bear the load ularly thank all member of the team: Cindy Hartono,
of one person. It is worth mentioning that the over- Fei Du, Shima Sahebnassagh, Eleanna Panagoulia.
all cost of the chaise was low (around 100 USD) and it
was ight weight in relation to its size and load bearing REFERENCES
capacity (around 20 kg). A future development of the Cranz, G 2000, The Chair, Rethinking culture, body, and de-
project would be a larger design space exploration sign, w. w. Norton & Company, New York, London
that would respond to a wider range of demands, Eames, D 2012, Eames: Beautiful Details, AMMO Books,
in terms of functionality and use, such as fit multi- Los Angeles, CA
ple people, accommodate extreme load cases with Knippers, J, Cremers, J, Gabler, M and Lienhard, J 2011,
Construction Manual for Polymer Membranes: Ma-
local material reinforcement. Also, in terms of join-
terial, Semi-Finished products, Form-finding Design,
ery and assembly, the possibility of disassembling Birkhauser Achitecture, Basel
and re-assembling the chaise lounge, by introducing Lienhard, J 2014, Bending-active Structure, Form finding
temporary, yet strong connections would be desir- strategies for elastic deformation in static and kinetic
able. Another aspect that is definitely worth looking systems and the structural potentials therein, Ph.D.
into would be the use of a more sustainable material. Thesis, Institute of Building Structures and Struc-
tural Design University of Stuttgart
However the challenge would be to allocate a ma-
Schleicher, S, Rastetter, A, Schönbrunner, A, Haber-
terial with similar structural properties. The poten- bosch, N and Knippers, J 2015 'Form-Finding and
tial of plywood as we saw in the Graz chair example Design Potentials of Bending-Active Plate Structures
was considered, but was excluded eventually due to in Modelling Behaviour', Proceedings of Design Mod-
the large bending radii that it requires. Moreover, we elling Symposium 2015
were aiming for an easier handling in terms of weight [1] http://www.eecs.berkeley.edu/˜jrb/Projects/decl
arative-design/declarative-design.htm
and fabrication speed. Before ending up choosing
[2] http://global.rakuten.com/en/store/chaoscollect
high-density polyethylene (HDPE), we ran some tests ion/item/eames_lcw_walnut_p/
with polystyrene as well as with low density polyethy- [3] http://www.designboom.com/design/matthias-
lene. The results were disappointing in terms of pliessnig-amada-bench/
structural stability and stiffness, as Polystyrene was [4] http://www.eamesoffice.com/the-work/airplane-fu
too brittle and low density polyethylene (LDPE) was selage/
[5] http://www.sofistik.com/en/
not stiff enough. A potential answer might be lying in
[6] http://coda-office.com/work/Graz
recycled High-density polyethylene, in case it main- [7] http://kangaroo3d.com
tains its original properties after the recycling pro-
cess, however this is a topic that needs to be looked
into and experimented upon.
ACKNOWLEDGMENTS
This work was conducted as part of Studio One
2014/2015 and the Bending and Folding Seminar at
UC Berkeley's College of Environmental Design at the
Department of Architecture. The case study was a

rolyPOLY
A Hybrid Prototype for Digital Techniques and Analog Craft in Architecture
Andrew Wit1 , Simon Kim2

1
Temple University 2 University of Pennsylvania
1
andrew.wit@temple.edu 2 simon.kim@i-k-studio.com
The rapid emergence of computational design tools, advanced material systems

and robotic fabrication within the disciplines of architecture and construction has
granted designers immense freedom in form and assembly, while retaining
pronounced control over output quality throughout the entirety of the design and
fabrication process. Simultaneously, the complexity inherent within these tools
and processes can lead to a loss of craft though the production of methodologies,
forms and artifacts left with extremely recognizable residues from tooling
processes utilized during their production. This paper investigates the fecund
intersection of digital technologies and handcraft through core-less carbon fiber
reinforced polymer (CFRP) winding as a means of creating a new typology of
digital craft blurring the line between human and machine. Through the lens of
an innovative wound CFRP shelter rolyPOLY completed during the winter of
2015, this paper will show the exigencies and affordances between the realms of
digital and analog methodologies of CFRP winding on large-scale structures.
Keywords: additive manufacturing, composites, form finding, craft, analog /

digital
INTRODUCTION weight ratio, malleability during fabrication, capacity

The recent introduction of carbon fiber reinforced to create complex yet precise forms while minimiz-
polymer (CRFP) based simulation tools (Waimer et ing waste through reconfigurable formworks, the ca-
al. 2013) and robotic fabrication methodologies (Gra- pability to maneuver between structural and superfi-
mazio and Kohler 2014) into the disciplines of archi- cial surfaces by means of subtle manipulation of ma-
tecture and construction has enabled designers to terial orientation and depth, as well as the ability to
not only reimagine potential formal and spatial qual- be easily fabricated on-site through either robotic or
ities of both built artifacts and environments, but has hand winding (Doerstelmann et al. 2015; Reichert et
also allowed for the complete re-envisioning of our al. 2014). These properties allow for a high-level of
current systems of fabrication and assembly through precision and adaptability that traditional means of
the creation of a single ubiquitous craft-based pro- construction cannot afford.
cess. CFRP's have properties such as high strength to

Context and fabrication of rolyPOLY, this paper will demon-
Current methodologies of CFRP placement within in- strate the potentials inherent within craft driven
dustrial processes place emphasis not on the visual, CFRP winding.
tactile, and craft based potentials inherent within the
material, but rather solely focus on producing op- rolyPOLY
timized and repeatable systems. Unlike disciplines Following the completion of several small-scale pro-
such as aeronautics or automotive where products totypes and projects focused around robotically
are mass-produced to be identical in dimension, wound pre-impregnated CFRP tow for the "One Day
structural properties, and appearance, architecture House Initiative" (Wit et al. 2015); rolyPOLY was initi-
requires the freedom and flexibility to freely alter ated as the research initiatives first intervention at an
configurations on a project-by-project basis. The inhabitable scale (Figure 1). Created as a traveling ex-
creation of robotic craft driven architectural artifacts hibition currently on display at the Tyler School of Art
though is complicated by the inherent complex- and slated for the "Beyond the Horizon" exhibition at
ity within these new tools and processes (Kudless the Philadelphia Museum of Art in the spring of 2017,
2013). For this reason this project re-evaluates cur- the artifact functions as a refuge for a single occu-
rent methodologies of CFRP placement, investigat- pant with the ability to be inhabited while remain-
ing the benefits native to handcraft in relation to ing structural in all-possible orientations. Through
those of numerically controlled machines while be- the operation of human assisted tumbling, a series of
ing tested and evaluated through the completed ar- unique spaces, visual / tactile textures and opacities
tifact's optical, structural, functional and tactile mer- are crafted and experienced by the inhabitant. Func-
its. tioning as a prototypical module, rolyPOLY is a single
The integration of handcraft into the digital component within a larger aggregating installation
workflow also has the potential to reinstate meth- to be produced at an architectural scale.
ods for rapid yet backwards-compatible revisions
to winding rules, on-the-fly improvisation and intu- Figure 1
itive alterations that would currently be difficult to rolyPOLY on display
achieve during the robotic CFRP winding process. On at the Tyler School
the other hand, aspects inherent through digital and of Art.
robotic production such as strength through speed,
accuracy and repeatability can be reintroduced into
the winding process creating a more robust hybrid
fabrication system where there is a continuous di-
alog between human and machine. For the work
documented in this paper, rolyPOLY acts as a proto-
type or bespoke object in which to study ephemer-
ality and perceptual effects such as texture, pattern
interference, and variable levels of opacity or inher- Continuing from the successful experiments in com-
ent patterning such as moiré. The discoveries gained putation and robotically programmed CFRP per-
through the prototype by human craftsmanship are formed at Ball State University in the spring of 2015,
then calibrated and formulated into rapid fabrication (Wit et al. 2016), rolyPOLY further examines the digital
methodologies via machine processes. workflow of numeric fiber winding through the rein-
Through a further examination of the materials, troduction of the traditional craft of hand winding as
processes and techniques utilized during the design a direct translation of computational data. Through

the utilization of a rigorous back and forth dialog ness, limited working time, and finally large amounts
between hand and machine, traditional elements of of wasted resin.
craft are reinserted into the winding process allowing
Figure 2
for the formation of new hybrid typologies of wound
CFRP winding
CFRP.
detail.
Based on a 10 sided primitive with two open
faces, each 7' x 4.5' x 4.5' module was fabricated
to easily aggregate with adjacent modules creat-
ing larger, more complex, self-supporting structures.
Wound from a single continuous strand of CFRP, each
finished module contains over 100,000 linear feet of
carbon fiber tow with a minimal self-weight of less
than 20 pounds. The high strength to weight ration
was achieved by the design and implementation of
variable winding procedures that allowed for a vary-
ing of skin thicknesses ranging between 1/16" and For these reasons, a pre-impregnated resin system
1/4" while simultaneously encouraging unique pat- was chosen. rolyPOLY utilized a specifically created,
terns and textures to emerge. Initially requiring a full high-impact resin system for several reasons includ-
24 hours of hand winding per module, subsequent ing: Material consistency, overall stability, shelf life,
projects have substantially reduced winding times low temperature curing, structural attributes and fi-
making the process more feasible for large-scale ar- nally because of its minimal toxicity and off gassing
chitectural applications. throughout the winding and baking process. A
12,000-strand count (12K) CFRP tow of roughly 1/8"
width and with a ±27.51% resin content was chosen
MATERIALS
for the application.
From the onset, the project aimed to further the dis-
In choosing the CFRP system, minimal toxicity
cussion on CFRP's position within the discipline of
and low baking temperature became the deciding
architecture and construction as a means of giving
factors, as the material would be both robotically and
designers more formal freedom, minimizing waste
hand wound over extended periods of time. This par-
throughout construction, as well as for the minimiz-
ticularly chosen resin system had the ability to be
ing of individual building component. To test mate-
cured in a temperature range between 220o F mini-
rial viability, initial prototyping examined the poten-
mum at 24 hours baking and 260o F maximum at 4
tials of both wet wound CFRP and pre-impregnated
hours bake time with an additional 4 hour period nec-
CFRP (Figure 2) resin systems within the context of
essary for kiln ramp-up and down of temperature.
core-less winding.
In addition, the chosen resin had an extremely long
Through a series of early small-scale winding
shelf life of up to 18 months at 0o F and 3 months at
tests, it was determined that a pre-impregnated sys-
room temperature. This additional stability allowed
tem would be utilized for the fabrication of rolyPOLY.
for extended work times, permitting the prototyping
Although wet-wound CFRP held several distinct ad-
and the winding process to take place over extended
vantages over pre-impregnated materials such as
periods of time without material degradation. Mate-
rapid curing time without the need for a kiln, it had
rial winding and the post-cure removal process were
several disadvantages that were difficult to overcome
also simplified by a slight material tackiness that dis-
such as: Difficulty in controlling resin content, work-
couraged slipping without resin pooling.
ing temperatures affected cure time, extreme messi-

FORM FINDING flat surfaces were treated as tensile membranes.
As the project's focus was centered on the explo- Through the utilization of a tensile modeling soft-
ration of CFRP's properties in relationship to craft ware's (MPanel Software Solutions) physics solver,
driven winding, the artifact's formal aspects were de- the form's edges were initially converted to cables
veloped from material deformation created as tow under a predetermined set of constraints (i.e. grav-
was wound over a base primitive. Formally centered ity, material properties, tension) which allowed for
around a sphenoid hendecahedron, rolyPOLY con- the edges to relax into their natural resting positions.
sisted of 10 unique faces of which two were left open Next, each of the eight solid panels was rebuilt uti-
for occupant access and interactive plug-in surfaces lizing the newly curved edges creating a series of ro-
(Figure 3). The primitive's asymmetrical nature, large bust, doubly curved tensile membrane. The new sur-
flat faces, sizable internal volume and stable weight face geometries were then relaxed producing the re-
distribution allowed for the creation of a structure vised overall form. Lastly, to simulate surface defor-
that could not only be oriented in any direction, but mation created by the tensioning of the overlapping
also produce unique internal / external spatial and vi- CFRP layering, the eight tensile membranes were in-
sual implications in each orientation as it was inhab- flated by a factor 10% in the MPanel modeling soft-
ited. ware.
Although the formal aspects of the sphenoid
hendecahedron primitive lent itself well to the de- CORELESS RECONFIGURABLE FORMWORK
sired programming and experiential needs of an in- To minimize material waste throughout the fabrica-
habitable refuge, the form's lack of surface curvature tion process while simultaneously simplifying wind-
created issues throughout early prototyping. Early ing and demounting, a reconfigurable formwork was
on it was found that to create tight bonds between fabricated. The formwork consisted of three simple
layers during the baking process, it was necessary elements, all of which can be simply interchanged
for all composite layers to be tightly overlapping and through bolt on connections:
compressed together. To create the necessary layer
overlap without the use of a solid core or vacuum 1. Steel frame. To create the overall form of roly-
bag system, it was essential to wind in a way that pro- POLY, a modular steel frame was created from
duced double curvature along all faces. ¼" steel round bar for each of the eight panels.
Each panel consisted of three to four uniquely
curved edges that were oriented perpendicu- Figure 3
lar to their corresponding surface then simply rolyPOLY formal
welded together. Every module edge corre- studies
sponded to an adjacent panel edge for easy
assembly.
2. Conduit hangers. Bolting the frames and
grippers together was simply accomplished.
Following the completion of the individual
frames, simple conduit hangers were utilized
to fix frames 1-8 together. Each frame was
bolted in two locations that were easily acces-
sible from the interior volume of the artifact.
As the pre-baked CFRP strands function similar 3. Laser cut CFRP grippers. On the outside of the
to cables when tensioned, all of the primitive's frame, a simple linear gripper was fabricated
out of 1/8" plywood and attached via conduit

hangers (Figure 4). The grippers were 1" deep After gaining a better understanding of the proce-
and had a series of ¼" wide by ¼" deep teeth dure, programming and winding limitations asso-
running along both sides. The topside was ciated with robotic programming a series of hand
utilized as the gripping surface while the bot- wound tests were initiated that would later re-inform
tom surface was cut to allow the teeth to take the robotic winding process. Following is a taxonomy
the curvature of the bent steel frame. Addi- of six successful winding typologies (Figure 5).
tionally, spaced every 12" of the bottom sur-
face was a 1" bolt on connection point for easy 1. Numeric helical offset winding
on-site alignment. 2. Random hybrid winding
3. Helically wound grid shells
4. Alternating helically wound grid shells
Figure 4 5. Core-less alternating polar offset winding
Frame detail. 6. Hybrid alternating shell winding
Although all six methodologies of winding are vastly

different in their appearance, each one began to ex-
emplify specific structural, tactile and craft based ef-
fects that were desired.
Initial testing with numeric helical offset winding
created robust lattice like structures that appeared
random in early layering but lacked the desired vari-
ation in opacity and tactility as additional layers were
added. Random hybrid winding created a robust,
extremely thin solid structure, but found itself with
To avoid having the CFRP cure to the frame, the en- inconsistencies in both structural stability and con-
tire frame was lightly coated in a spray mold release trol during the winding process. The grid shell ty-
before winding. Following the winding, baking and pologies created the most robust structures, but also
cooling of the artifact, the frame is simply unbolted had extremely repetitive patterning while also lack-
from the inside of the refuge and disassembled one ing the projects desired level of variation. Coreless
frame at a time while being removed through one of alternating polar offset winding created both the de-
the two openings in the object. sired structural capabilities and control for the wind-
ing of rolyPOLY. Unfortunately this method relied on
PROTOTYPICAL WINDING STUDIES a fixed series of points that would not work appropri-
Initial small-scale winding was tested through the ately for the asymmetrical and continuous shape of
utilization of a 6-axis, KUKA KR 60 industrial robot the installation.
with custom fabricated end-effectors and ground a To accomplish all of the desired effects, Hybrid
mounted fiber dispenser consisting of a ¼" diame- alternating shell winding was chosen as the projects
ter, lubricated, aluminum tube 24" above the zero method of choice.
plane. As the robot moved through space spiral-
ing around the given effector, fibers were effort- WINDING METHODOLOGY
lessly pulled through the dispenser and placed on Establishing an initial desire for transparency, optical
the effector. Wet winding required an additional step effects, and patterning, while winding across the ten
where the tow was pulled through a pool of resin faces - two of which would be left open - required
prior to application to the effector. design implementation and resulted in the develop-

ment of the hybrid alternating shell winding. Each dition of four widely spaced layers was highly suc-
face was first separated and flattened. Diagrams for cessful in assuring layer consistency throughout the
each face were created to determine the pattern as winding process. Resin tackiness in the initial layers
needed, such as increasing density as a gradient from also helped minimize fiber movement and separa-
edge of face to center. Assembled back together tion ensuring that consistent tow tensions and pat-
in the original sphenoid hendecahedron, a continual terns were achieved.
winding route had to be manufactured, as the carbon Following the laying of the sacrificial CFRP layers,
fiber tow could not be cut or otherwise interrupted a series of alternating winding patterns were layered
without limiting its capabilities as a monocoque shell. that occupied roughly 95% of the winding process.
An additional factor was the degree to which the craft To achieve optimal strength, layer rigidity, consis-
driven method may change or adjust the winding tency, and visual opacity, a variable winding pattern
densities and directions in response to sun, vision, or was developed consisting of three repeating steps:
aesthetic preference. This real-time adaptability to (1) peak winding, (2) valley winding, and (3) spiral
correct or change as needed supersedes the abilities winding. Each step consisted of four distinctive layers
of a preprogrammed robotic arm. of wound carbon CFRP. Following one full winding
After initial pattern design, the overall winding sequence, winding direction was flipped 180 degrees
process was developed in two distinct phases: (1) and restarted. If at any point during the winding pro-
sacrificial formwork and (2) alternating layers. cess deficiencies in craft, visual effect, or structure ap-
peared, a variable could be introduced to produce
1. Sacrificial formwork
the desired resultant. Additionally, winding occurred
2. Alternating layers
over a series of four tangential panels at a time allow-
Through the completion of previous prototypes it ing for a robust bond between all faces and edges of
was determined that upon commencement of the the installation.
winding process, the addition of a tightly wound sac-
rificial layer would aid in the stabilization of the sub- 1. Peak Winding: Placed first, peak winding was
sequent layers. Initial studies found that the bot- utilized as a means of generating the neces-
tom few layers had a tendency to be less consis- sary internal forces to create a stable and con-
tent, as minor variations in CFRP tension and shifts sistent form. Running between the peaks of
within the frame could occur. Therefore, the ad- each panel, high levels of positive surface off-
Figure 5
Winding studies.

set were generated that would later be com- temperatures excess of 2,000o F, extra care was taken
pressed and pulled position by the subse- in the pre-heating process to ensure that the entire
quent two overlapping layers. As the desired volume was heated to a consistent temperature of
surface position was as far off as positive 4" in 260o F. During baking process, a temperature varia-
these layers, winding tension was decreased tion of 40o F could delay the curing process by up to
from 10 pounds of force to 5 pounds allowing 20 hours. In addition, if the temperature within the
for subsequent layers to effortlessly pull them kiln's vertical volume is inconsistent, the overall form
into place. may also be inconsistently cured leading to post bake
2. Valley Winding: In contrast to peak wind- structural failure.
ing, valley winding connected the panel's low
Figure 6 points in a perpendicular direction. As the
rolyPOLY on frame natural resting surface offset for the valley
prior to baking. winding was an average of negative 2" from
the desired surface location, collisions be-
tween peak and valley layers occur through-
out the winding process. The slightly higher
tension within the valley winding aided in
the drawing in of all peak-winding fibers into
their predetermined resting positions. As
surface curvature variation could occur be-
tween panel edges and their respected cen-
ters, slight layer delamination sometimes oc-
curred in panels with the greatest levels of Upon baking for 4 hours, the kiln is disengaged,
curvature. and allowed to slowly return to room temperature.
3. Spiral Winding: To eliminate any remaining When the baked artifact has cooled, it was removed
surface variation and also ensure tight fiber from the kiln and prepared for dismounting from the
overlap, spiral edge winding was applied. frame (Figure 6). The previously discussed dismount-
As the highest levels of layer separation oc- ing process takes only minutes and produces zero
curred in panel edges, a third layer of CFRP waste.
was wound over all intersecting panel edges.
Winding in a tight clockwise direction, spiral
winding not only compressed any loose fibers
CONCLUSIONS
As evidenced from pre-determined design and pre-
but also created strong connections between
set fabrication by robotic procedures, counter to im-
all intersecting panels. As surface centers had
provisational handcraft, the means and methodol-
less deviation, spiral winding was only neces-
ogy of winding contains a forcible outcome. Between
sary on the first 12-18" of each edge.
the two methods is an array of outcomes where a
shared procedure may be profitable. A robotic pro-
BAKING cedure may take on the initial Peak and Valley wind-
The curing process for the chosen CFRP resin system ing after which handcraft may deliver the final cus-
was extremely straightforward. Following the com- tomized wrapping as needed by the designer. With
pletion of the winding process and kiln temperature a greater input and output from feedback sensors, a
ramp up, the full module was loaded into a large gas robotic arm may be able to one-day display real-time
fired kiln. As the available kiln was designed to run at adjustment to control tension or density, for exam-

ple. Having gone through the distinct differences in ACKNOWLEDGEMENTS
robotic winding as well as handcraft production, val- The authors would like to thank the following re-
idation and metrics of value are formed (Figure 7). searchers, institutions, and partners for their con-
tinued assistance throughout this ongoing research:
John Williams & Chad Curtis (CFRP Baking); Tim Figure 7
Rusterholz (Steel Fabrication); Aidan Kim, Joseph Gi- rolyPOLY on display
ampietro, Daniel Lau, Han Kwon, Junghyo Lee, Yue at the Tyler School
Chen, Lyly Huyen (Documentation & Production); of Art.
Temple University Tyler School of Art: Division of Ar-
chitecture and Environmental Design; University of
Pennsylvania School of Design; TCR Composites (Ma-
terials Donation + Design Assistance); MPanel Soft-
ware Solutions (Software Donation).
REFERENCES
Doerstelmann, M, Knippers, J, Koslowski, V and Vasey,
L 2015, 'ICD/ITKE Research Pavilion 2014-15: Fibre
Placement on a Pneumatic Body Based on a Water
Spider Web', Architectural Design, 85(5), pp. 60-65
Gramazio, F and Kohler, M 2014, The Robotic Touch: How
Robots Change Architecture, Park Books
Kudless, A 2013 'Formations of Digital Craft Culture', Pro-
ceedings of ACSA 2013, San Francisco, pp. 368-375
Reichert, S, Schwinn, T, La Magna, R and Menges, A
2014, 'Fibrous structures: An integrative approach
to design computation, simulation and fabrication
for lightweight, glass and carbon fibre composite
structures in architecture based on biomimetic de-
sign principles', Computer-Aided Design, 52, pp. 27-
39
The judgment of one over the other is not necessar- Waimer, F, La Magna, R, Reichert, S, Schwinn, T, Menges,
ily in cost effective time saved or in ad hoc artistic in- A and Knippers, J 2013 'Integrated design methods
spiration, but in the production of meaningful and for the simulation of fibre-based structures', Design
performative artifacts. As the rolyPOLY was a proto- Modelling Symposium 2013, Berlin
type in which to test the above-defined production, Wit, AJ, Eisinger, D and Putt, S 2015, 'Human Interaction-
the next step would be to gauge how well an occu- Oriented Robotic Form Generation', in Reis, LP, Mor-
eira, AP, Lima, PU, Montano, L and Munoz-Martinez,
pant is housed or sheltered within its enclosure. In V (eds) 2015, Robot 2015: Second Iberian Robotics
particular, the winding, as both core-less structural Conference, Springer International Publishing, pp.
shell and separating membrane between occupant 353-364
and the world, would need to respond to lighting, Wit, AJ, Kim, S, Ibanez, M and Eisinger, D 2016, 'Craft
views, privacy and safety. These questions and their Driven Robotic Composites', 3D Printing and Additive
compelling implications would be continued as the Manufacturing, 3(1), pp. 2-9
rolyPOLY expands in size, in color, and in connection
with other modules

Material Active Geometry
Constituting Programmable Materials for Responsive Building Skins
Alireza Borhani1 , Negar Kalantar2

1,2
Department of Architecture, Texas A&M University
1,2
www.thetranslab.com
1,2
{borhani|kalantar}@tamu.edu
This paper is part of a body of research developing an exploratory dialogue

between the built form and the environment, via experimentation with
performative geometry and material. Here, geometry is considered a design
material with the specific capacity to contribute to the performative aspects and
kinetic capabilities of building skins.This work opens with a review of emerging
opportunities for architects to design materials. It then discusses the concept of
Material Active Geometry (MAG) as a means of designing new properties for
existing materials. This is followed by a discussion of MAG principles that inform
the concepts of flexibility and rigidity in a 3D-printed textile called Flexible
Textile Structure (FTS). This research characterizes two FTS types and discusses
their potential to be employed in building skins; it also considers combinatory
approaches to computational models and physical prototyping. The work
concludes with a discussion of the advantages of using FTS, and provides a
trajectory for future research in the field of responsive materials and systems.
Keywords: Programmable Material, Material Active Geometry, Flexible Textile

Structures, Responsive Building Skins, Flexible yet Rigid
RECONCILIATION OF APPEARANCE AND they might simultaneously retain their structural re-
PERFORMANCE LYING AT THE HEART OF silience and resistance. Through a geometric contri-
GEOMETRY bution to the inherent dynamic behavior of material,
In this paper, a primary point of departure is the its properties can be tailored to meet specific struc-
integration of geometry into the conception, com- tural and mechanical needs (Figure 1).
putation, materialization, and ultimate creation of This paper strives to understand some attributes
new material properties that are flexible yet rigid, ro- that cause the ability to change a material's appear-
bust yet light, simple yet complex. Therefore, this ance and performance into whatever specifications
paper lies at the confluence of various fields and the designer desires. For instance, rendering a deter-
seeks to pave a new path for the development of pli- minate material that is geometrically designed can
able materials, enabling them to self-organize so that reinforce the architectural form and spatial compo-

sition of a motion composition. To this end, this pa- gies navigate between vector-active, section-active,
per focuses on the way material unites with geome- form-active, surface-active, and hybrid systems. This
try, which acts as the underlying structure of the mo- study identifies a new typology that refers to a
tion design process. By celebrating the materiality "material-active" system when forces are applied.
of a transformable element in the conceptual phase A material-active system is a system of rigid
of its design process, the properties of the material matter that mainly employs transformable elements
serve as a coherent design feature and tool. In other through the geometry of its material. Regardless
words, this paper thrives on the tension between ma- of the size or scale of the material employed in the
terial articulation and design. transformable element, the motion quality of that el-
ement depends less on the inflexibility or flexibility
of the material and more on its geometry for flexibil- Figure 1
ity (Kalantar and Borhani 2013) (Figure 2). Working Flexible Textile
with rigid materials, this study reflects on how a small Structures that rely
level of flexibility in a material can be incorporated as on the appearance
an active parameter into the movement of the whole and performance of
system (Figure 3). 3D-printed
This paper discusses how the appearance of materials.
material might be intertwined with and dependent
upon its performance, through an exploration of new
shared territories in architecture, engineering, and
technology called "Material Active Geometry" (MAG).
Here, Material Active Geometry is coined by the au-
thors. Figure 2
The authors have developed a process by which Uniting material
MAG can be employed as a means of designing ma- with geometry to
terials. Here, MAG relies primarily on geometry as act as the
a structural and organizational mechanism, rather underlying
than depending solely on a material's properties; the structure of a
goal is to achieve better performance. MAG's pur- motion design.
pose is to improve the adaptability of a structure by
understanding the reciprocity between the geomet-
ric configuration of its materials and their dynamic in-
teraction with external forces.
Within the context of performative geometry as Figure 3
a system of design, MAG should not be understood as Using MAG
seeking to invent new materials per se, but rather to principles to inform
use existing ones, play with their organization, juxta- the motion.
positions, and geometric settings, and thereby reveal
their undiscovered properties.
Although a material-active system is intended
to withstand the maximum stress for which it is de-
MATERIAL ACTIVE SYSTEM signed, its applied geometry enables it to undertake
Structurally speaking, Heino Engel (2007) determines a planned deformation after being stressed and be-
several active structural typologies. These typolo-

fore exceeding the material's elastic limit. In this ometry integrated with the three-dimensional form
manner, a material-active system is not simply one of motion.
that changes the material's steadiness. Rather, this
Figure 4
system carries out elastic deformation when a force
The physical
is applied; the system either returns to its origi-
embodiment of
nal shape or holds a new changeable configuration
geometry in
when the force is removed. To sustain the function of
material.
the system, the geometric design of the material em-
ploys several aspects to avoid any fracturing, buck-
ling, or inelastic (plastic) deformation. In other words,
the applied geometry functions as a framework for
guarding against structural failure. In addition, the
overall applied geometry and resulting configuration
encapsulate a solid architectural idea for the design
Figure 5 of any particular assembly, element, or joint capable
Transmitting of transmitting applicable forces during elastic defor-
applied forces mation (such as deflection or torsion).
through a particular Although a correlation of compressive and ten-
use of geometry to sile forces is necessary to meet a structural load, a
inform the 3D form material-active system puts an emphasis on the gen-
eration of geometrically-informed material behav- In designing a material-active system the following
of motion.
iors that facilitate the load bearing of transformable objectives are outlined:
elements. By offering structural capacity due to a net-
work of relationships among the different parts of the
material, the development of the structural system • Designing material behaviors, rather than
integrates the material behavior into the whole of the simply shaping already existing material
movable structure. • Manipulating material over time from the pas-
A material-active system is able to transmit ap- sive to the active volume due to a set of dy-
plied forces primarily through a particular use of ge- namic properties to provoke a radical change
ometry and not through the bulk and continuity of in its expression and functionalities
its material (Figure 4). Sometimes, the involved ge- • Offering both flexibility and rigidity on de-
ometry generates indirect, interrupted load paths mand: Static-dynamic materials as a future di-
whereby forces should bypass separate components rection of interaction design
that are adequately tied together. In such cases, the • Deriving performance from the local geomet-
movement or displacement of one component can ric organization of materials
be transformed to other components, or to other • Sitting at the intersection of geometry, mate-
parts of the same component (Figure 5). rial, design, and emerging fabrication meth-
Designing a material-active system involves at ods such as additive technologies
least three inseparable counterparts that character- • Exploring the relationships between materials
ize the whole system: (1) the investigation of material and their shaping processes in the generation
properties, (2) the design of a thoughtful geometry to of form through "form finding" exercises in 3D
manipulate the allowable stresses in the movable el- printing
ements, and (3) the physical embodiment of the ge-

THE NEW FUNCTION OF PERFORMATIVE different external and internal stimuli, both in terms
GEOMETRY of its structural configuration and overall behavior.
The design investigation conducted for this paper To this end, the main objective of is to explore the
represents an initial exploration into MAG's larger methodological frameworks of soft-form transfor-
implications as a means of producing physical en- mation and the computational tools for simulating
vironments capable of changing shape to accom- geometrically-driven materials by employing knowl-
modate various spatial performances; moreover, this edge and apparatuses inspired from the field of biol-
work addressed the potential of a reciprocal transfor- ogy.
mation between a building's occupants and its sur- An adaptive system holds a dual promise. On
rounding environment in a responsive manner. the one hand, there is the need to provide local re-
In the interest of exploring negotiations among sponses to stimuli as nastic movement; on the other,
geometry, form, and materials as facilitated through there is the desire to globally alter the whole system's
environmental and computational modeling, as well geometry, sometimes quite radically (when skele-
as the application of these concepts to the fabric tal motion is possible). In nature, geometry recom-
of our buildings, architects have been endeavored mends itself as serious way for altering the structural,
to comprehending responsive systems for building performative, and aesthetic characteristics of mate-
skins. From the responsive brise-soleil of the Los rials, due to its ability to provide a global mode of
Angeles County Hall of Records (designed by Neu- movement.
tra in 1962) (Sample 2012), to the Institut du Monde Since the late 1990s, the development of
Arabe in Paris (designed by Nouvel in 1987) (Ritter computer-aided design (CAD) software and digital
2007), and to the present day, the idea of respon- fabrication tools has led to a significant increase in
sive building skins has been explored by architects complex geometry and curvilinear shapes in archi-
and scholars alike. However, its potential has yet to tectural design. Conversely, in adaptive systems,
be explored within the praxis of architecture. Fortu- there is a direct connection between structural per-
nately, the renewed interest in responsive architec- formance and geometry. When designing process-
ture has been prompted by a collaborative exchange adaptive building skins, more advanced geometry
between engineers, architects, and biologists, as well must be implemented through parametric design
as the maturing suite of computational tools and re- tools in order to achieve optimal geometric perfor-
cent advances in materials science. mance.
Since MAG has not established a lengthy legacy
in the field of architecture, this research adopts sev- ALTERING FLEXIBILITY AND RIGIDITY
eral precedents from the realm of biology and an- THROUGH 3D-PRINTING: SURFLEX AND
alyzes their strengths and weaknesses as elements
SURFMORF
in a system of adaptive materials. For instance, skin
This study is a prototype-based exploration of MAG's
is a complex natural system of adaptation that ful-
formal and aesthetic potential; as such, it examines
fills multiple and often conflicting functions (Vincent
the principles of performative geometry and their ap-
2009). Translated to the architectural scale, skin can
plicability to architectural delineation and transfor-
offer an excellent model for self-reliant building en-
mation [1]. This goal was explored and evaluated
velopes that serve disparate purposes for a variety
through a rigorous process of designing a new and
of reasons. Based on the reviewed precedents, MAG
responsive architecturally-morphing skin called Flex-
can be situated as an application of a biologically-
ible Textile Structure (FTS). To develop FTS, the au-
responsive system in nature within an architectural
thors from [trans]LAB, a design-build practice, collab-
framework attempting to constantly adapt itself to
orated with the DREAMS Lab, an additive manufac-

turing research laboratory in the department of me- and their methods of assembly. These were catego-
chanical engineering at Virginia Tech University (Fig- rized into two genera of cellular aggregation, as fol-
ure 6). lows:
By exemplifying the potential of using
geometry-based design as a structural and organiza- • surFLEX: The first was used to investigate the
tional mechanism, FTS provides MAG's most essential geometrical configuration of 3D-printed tex-
elements. Here, MAG's principles were used to guide tiles through an aggregation of tightly nested
the authors' experiments through geometries used and interlocked torus-like cells. Each torus
to employ FTS in adaptive building skins. was composed of several cord-like elements
that shaped its enclosure. Similar to topo-
Figure 6 logical interlocking systems (Estrin and Dyskin
Flexible Textile 2011), the cellular structure was arranged
Structures, such that an entire assembly was held to-
designed by the gether by kinematic constraints imposed on
authors. the neighboring cells, through their mutual
arrangement and connectivity. In surFLEX,
without requiring any external forces to main-
tain contact pressure between the series of
cells, the structure could make use of the
grasping mechanism of the cells' cords as
part of a greater interlocking system. This
type of FTS could serve as both envelope and
structure. While maintaining a relative kine-
matic constraint, the cells' movements were
not fully locked in any direction - neither par-
allel nor vertical to the structure's plane - and
thus could be used to adjust the performance
of the structure to meet a variety of needs (Fig-
The degree of flexibility or rigidity of a material can
ure 7).
be derived from geometrical principles. As a con-
• surfMORF: The second category of flexible
sequence of making a connection between the for-
yet rigid textiles included a reversible struc-
mation and materialization processes and validating
ture comprised of a loosely-nested assem-
this integration into MAG, FTS can directly inform the
bly of continuous cords that could be used
motion. FTS presents the design of a novel form-
to develop geometrically-differentiated sys-
changing system with the capacity to adapt, in the
tems. One of the main advantages of this
form of an architectural skin that is lightweight, flex-
category is that such materials are unable to
ible, and elastic. FTS research is a nascent field in-
sustain applied external loads, and thus im-
vestigating the use of geometrically and physically-
pact the whole structure. This is attributable
actuated material to prototype shading and air ven-
to the absence of an interlocking mechanism;
tilation systems; these systems can be applied to var-
the cluster-like cells can be freely displaced to
ious architectural components, such as building en-
provide maximum flexibility. Since the tex-
velopes and roofs.
tiles are incapable of resisting bending forces
The development of such FTS skins began with
or even tension, structures can tolerate defor-
an exploration of various individual modules or cells
mations without losing their structural coher-

Figure 7
surFLEX, the first
category of Flexible
Textile Structures.
Figure 8
surfMORF, the
second category of
Flexible Textile
Structures.
Figure 9
Exploration of the
parametric
geometries of FTS.

ence. The overall shape of surfMORF emerges of a cell or module, one considers an interlocking
from the size of its modules, their orienta- mechanism to ensure kinematic constraint between
tions, the ratio of the constituent cord-like el- neighbors. Current animation software's ability to
ements, the method of their connectivity, and simulate the performance of textiles provides only a
their contact faces (Figure 8). simple visual and explorative analysis tool; physical
prototyping remains critical to the study of localized
The morphological behaviors of both FTS categories regions of failure within the system.
seem both necessary and desirable as a means of ac- As explained above, neighboring cells kinemat-
commodating different requirements. The different ically constrain each other. The tangential gap be-
levels of flexibility and rigidity inherent in each cat- tween the cells' cords in the interfacing zone of two
egory offer promising means of elevating the formal modules is very important. This gap is critical to
and performative aspects of a building envelope. defining the tightness of the cell's movements, de-
This study attempted to broaden the formal, ge- pending on certain variables such as the 3D-printing
ometrical, and performative repertoires of FTS mod- method and machine tolerance. Thus, in this re-
ules and their parametric assembly. To this end, in search, different samples were iteratively prototyped
order to populate their complex geometries, a gen- to assess the overall gaps between cells, as well as
eration algorithm must be developed that is capa- their geometric conflicts; this allowed for an appraisal
ble of altering the orientation, dimensions, porosity, of the fidelity of the computational models and print-
and directionality of the modules within the system, ers. Within a proposed FTS geometric configuration,
while providing a larger geometrical repertoire be- a design file cannot simply be scaled up or down; the
yond those seen in mere planar, repetitive, or homo- gaps between the cords are independent of the size
geneous systems (Figure 9). For instance, in the first of the aggregated cells.
FTS category, surFLEX, a parametric geometric algo- Here, the FTS research methodology evolved
rithm was developed based on the size, geometry, through a series of investigations into the develop-
orientation, and location of each cell within the over- ment of physical analog prototypes obtained from
all assembly, the cell-adjacent contact faces, thick- a very narrow material system of 3D printers; this
ness of the constituent cords, gaps between two ad- helped the authors to gain further insight into the
jacent cells and their connected cords, organizational most appropriate forms of transformation and al-
hierarchies of the assembly, and degree of poros- lowed them to test the affordance of FTS being im-
ity of the textile. According to the number of con- plemented in building skins (Figure 10 and Figure 11).
nected points, the algorithm would allow for the gen-
eration of different regular or non-regular spatially- CONSTITUTING A PROGRAMMABLE ENVI-
tessellated patterns, consisting of a variety of possi-
RONMENT
ble parametric and interlocking assemblies and with
One of the main contributions of this research is its
individually differentiated and multi-faceted mod-
exploration into achieving desirable morphing prop-
ules.
erties in newly-developed building skins (Figure 12).
In surFLEX, each cell can be connected to its adja-
Of particular importance to the trajectory of morph-
cent cells at three, four, or six interconnected points.
ing changes are the numerous developments in com-
The general aggregated system easily facilitates cell
puter science, artificial intelligence, and robotics de-
combinations of different sizes and materials. This
signed to produce architecture that can change au-
category is advantageous because it tolerates the re-
tonomously (Sterk 2009).
moval of several cells while maintaining performance
FTS's embedded kinetic properties seem to hold
and structural stability.
great promise for building skins, allowing them to be
surFLEX is achieved when on every desired side

imbued with environmental response, interaction, date new conditions is a key factor of surviving in
and intelligence capabilities. Despite promising to nature (Jeronimidis and Atkins 1995), morphological
provide further responsiveness, at this point, FTS re- changes in response to the immediate environment
mains highly constrained with just geometry. The should be the main trajectory of FTS research, entail-
combination of FTS's kinetic capabilities with inte- ing a seamless and continuous shape-changing de-
grated sensing and actuation has significant poten- formation.
tial and requires further investigation. In their effort to design adaptive architectural el-
ements, a few design scholars pursued the early ex-
ploration of kinetic materiality, including the passive Figure 10
or active response capacities of the properties of pro- Weaving closed
grammable form-changing materials. The works of braids to form a
Menges (Menges and Reichert 2012), Sung (2010), surfMORF.
and Beesley (Gorbet 2010) set a precedent by which
material actuation was applied within the responsive
kinetic architectural context of several installations.
Due to the limited range of movement caused
by their actuation, most form-changing materials are
mainly able to perform locally, at the micro scale.
A hybrid approach combining MAG strategies and
form-changing materials would provide an excellent Figure 11
opportunity to perform actuations by using geome- Using the geometry
try within the materials' properties, in a coordinated of braids to form a
effort to induce a transformation of the overall build- surfMORF.
ing skin at the macro level. Implementing geometry
in the correct manner would help to amplify the con-
traction or expansion of the material actuation that
up to this point has primarily been introduced on a
small scale.
Although FTS's geometric capacity has been ex-
plored through the design investigations conducted
here, there remains the opportunity for FTS to fully
exploit the advantages of deploying both its geomet- Figure 12
rical and material capacities (Figure 13). FTS's use of FTS as
form-changing materials will contribute to changing geometrically and
the shape, geometric arrangement, and mechanical physically -actuated
properties of its neighboring parts as a response to material to explore
environmental inputs. a shading system.
Emerging from the behavioral formation of FTS
through the interaction of its geometry and materi-
als, the goal of the design process is to create self-
organizing and flexible yet rigid structures that mor-
phologically reacts to changes in light, moisture, and
Since the ability to change shape and/or accommo- temperature. By shifting from the explicit design of

form or material, the formal and performative trans- alloy wires will contract when an electrical current
formations of FTS are derived from "the orchestration runs through them. In addition to changing the FTS's
of intensive processes of formation through the de- porosity and permeability, contracting a part of the
sign of the underlying behaviors of matter and geom- structure could cause local convexity and concavity.
etry" (Snooks 2012).
Figure 13 Although programmable materials with the ability
Exploiting the to change form remain less explored, they have
geometrical and great potential, especially for responsive architec-
material capacity of tural design. However, one of the drawbacks of pre-
FTS in programing material properties during the fabrica-
programmable tion process is that the materials are then unchange-
fashion. able during the building's performance (Khoo and
Salim 2013). The pre-programmed operation of such
materials makes them less able to genuinely learn
from and respond to the environment, based on real-
time feedback. Since enabling a building's materi-
als to sense and react to the environment in a de-
Although using electricity to activate shape-memory
sirable manner is one of the main characteristics of
alloy wires would help control the FTS's behavior, the
any adaptive structure, as noted by Sybil P. Parker
idea of a geometrically-actuated material remains
(Parker, 1994), in the next step of FTS development,
a means of moving towards a zero-energy respon-
there should be an attempt to use reactive materi-
sive skin that performs without implementing any
als capable of sensing the environment; such mate-
mechanical or electrical secondary systems; Gary
rials would provide consistent transformations well-
Brown has argued a similar point regarding shifting
suited to support the specialized and autonomous ki-
in adaptation from a complicated system to a bio-
netic tasks of a building skin. Embedding it with sens-
logical paradigm (Brown 2002). Achieved with fewer
ing and actuating capabilities when it is coupled with mechanical components and devices, FTS is a step
collective intelligence on the material and feedback-
toward developing responsive materiality in archi-
loop scale would reframe FTS within architecture as
tecture, as well as cooperating material behaviors
a provider of processes of change rather than an ex-
through transformations in geometry.
pected artifact. In this case, after initiating a set of
controlled constraints, FTS would no longer be per-
ceived as the final form of a building skin, but rather DISCUSSION
the generator of preliminary states that would evolve As part of a larger effort to better understand the op-
in real time to find their forms while responding to ex- eration of performative building skins, this paper in-
ternal stimuli. vestigated how to deal with the geometry necessary
It is possible to embed conductive material into to generate transformations of morphology. Con-
3D-printed parts in order to pass electrical current; sequently, Material Active Geometry sought to re-
this can then activate shape memory alloy wires such conceptualize the formal and material properties of
as Flexinol and allow for precise control of the FTS's building skins within an adaptive system framework.
movement as a performative building skin. When By extending the possibilities of the digital realm
electric current is channeled to individual shape- into multifaceted material behaviors, MAG can be a
memory alloy wires through the appropriate soft- fertile research tool that leads to interdisciplinary col-
ware, the desired parts of a structure can move; the laborations in art, science, and engineering (Kalan-
tar and Borhani 2014). By crossing the boundaries of

various fields such as biology, material science, com- Estrin, E, Dyskin, A.V. and Pasternak, E 2011 'Adaptive
putation, mathematics, and additive manufacturing, mechanical properties of topologically interlocking
the study of MAG aims to highlight the search for a material systems', Mater. Sci. Eng. C, Principles and
Development of Bio-Inspired Materials 31, pp. 1189-
way of thinking about issues of adaptation, change,
1194
and performance in diverse fields of design. Gorbet, R 2010, 'Revealing the Hylozoic Ground Interac-
By embracing computational design, analysis, and tion Layer', in Beesley, P (eds) 2010, Hylozoic Ground:
fabrication tools and procedures, the paper ex- Liminal responsive architecture, Riverside Architec-
plained the experimental probes by which the au- tural Press, Toronto, pp. 112-123
thors investigated Flexible Textile Structures 'various Jeronimidis, G and Atkins, A.G. 1995 'Mechanics of Bi-
ological Materials and Structures: Nature', ARCHIVE
capabilities.
Proceedings of the Institution of Mechanical Engi-
Since a computational design tool that merely neers Part C Journal of Mechanical Engineering Sci-
uses geometrical representations would be insuffi- ence 1989-1996, pp. 221-235
cient for proper simulation of FTS performance, the Kalantar, N and Borhani, A 2013 'Establishing the
discussed parametric geometric algorithm must be Language of Motion Formation for the AURA Ki-
developed; such development should be based upon netic Shading System', International Conference on
Adaptation and Movement in Architecture (ICAMA),
additional data such as areas of stress concentration
Toronto, Canada
between connected cells, means of informing struc- Kalantar, N and Borhani, A 2014, 'Flexible Textile Struc-
tural and mechanical computer simulations, and fi- tures: An Agency for Informing Form and Matter',
nite element models. Accordingly, coalescing FTS's International Journal of Interior Architecture + Spatial
generative and analytical aspects into a single para- Design, ii, pp. 50-55
metric setup capable of constraining and orchestrat- Khoo, C.K. and Salim, F 2013 'Responsive Materiality for
morphing architectural skins', Proceedings of the 33rd
ing the entire design and fabrication process will in-
Annual Conference of the Association for Computer
form the next step in this research. Aided Design in Architecture (ACADIA), Cambridge,
As well as testing the structural or mechanical pp. 243-252
properties of different designs, the close link be- Menges, A and Reichert, S 2012, 'Material capacity:
tween analytical and generative parametric models Embedded responsiveness', Material computation:
calls for an optimization of the material's distribution Higher integration in morphogenetic design, 82, pp.
52-59
and a minimizing of its use, as suggested by finite
Parker, S.P. 1994, McGraw-Hill Dictionary of Scientific and
element models and based on load paths and local Technical terms, McGraw-Hill, USA
stresses. Ritter, A 2007, Smart materials in architecture, interior ar-
Despite using a highly additive manufactur- chitecture and design, Birkhauser, Basel
ing method as a primary departure point for this Sample, H 2012, 'A brise-soleil without a building', in Bor-
research, the time-consuming and expensive 3D- den, GB and Meredith, M (eds) 2012, Matter: Mate-
rial processes in architectural production, Routledge,
printing process requires a search for the best and
Oxon, pp. 329-339
most feasible method of manufacturing FTS within Snooks, R 2012, 'Volatile Formation', LOG, 25, pp. 55-62
various contexts and on a variety of scales, from Sterk, T 2009 'Introduction: Thoughts for Gen X specu-
micro-particles to large-scale building skin modules. lating about the rise of continuous measurement in
architecture', Proceedings of ACADIA Annual Confer-
ence, Chicago, pp. 18-22
REFERENCES Vincent, J 2009, 'Biomimetics of skins', in Favret, E and
Brown, G 2002, 'Freedom and Transience of Space Fuentes, N (eds) 2009, Functional properties of bio-
(Techno-nomads and transformers)', in Kronenburg, inspired surfaces: Characterization and technological
R, Lim, J and YunnChii, W (eds) 2002, Transportable application, World Scientific Publishing, London, pp.
Environments 2, Spon Press, London, pp. 2-20 3-15
Engel, H 2007, Structure Systems, Hatje Cantz, 3rd edition [1] https://www.youtube.com/watch?v=dpL0Y2l_BSI

TOWARDS SMARTER CITIES |
Concepts and Strategies
Smart City: Adding to the Complexity of Cities
A Critical Reflection
Emine Mine Thompson1

1
Northumbria University, Department of Architecture and Built Environment, Suther-
land Building, NE1 8ST, England
1
https://www.northumbria.ac.uk/about-us/our-staff/t/dr-emine-mine-thompson/
1
emine.thompson@northumbria.ac.uk
This paper seeks to further the state-of-the-art knowledge on what a smart city is;
by analysing the smart cities across the world. It also seeks to find out how
different approaches to the smart city creation influence the city. This worked
based on the ongoing review on Smart Cities that was started in 2014. In
particular, it is structured as follows. First, definition of "smart city" are
reviewed, next typologies of smart cities were generated by analysing the different
types of smart cities across the world. Following this, case studies are reported.
Case studies are followed by a reflection. Paper drawn to conclusion following
this section by giving research directions to take this research further.
Keywords: smart cities, future cities, urban planning, city transformations,

number of smart cities
INTRODUCTION nesses, transport and communication networks, ser-

In this critical reflection we are looking to understand vices and utilities.
whether the smart city solutions are considered as From the beginning of the city life in Jericho
technology projects or part of a holistic planning so- (8500-6000 BC) and Catalhoyuk (7000-5500 BC), as
lution which would lead to smart urbanism. The Smith (2012, 33) explains, the fundamentals of human
key question we want to answer is to whether we life in the first cities did not differ greatly from ours to-
are adding another layer of complexity to the aging day. From the love of good food expertly cooked and
siloed structure of the city systems or whether the enjoyed with friends and family, to the need to work
utilization of data and technology will prevail a more and the pleasure of shopping, their daily life mirror ours.
successful urbanism for the future cities. Furthermore, cities have always been places where in-
novations in transport, communications media, print-
Cities ing, publishing, the processing information and the cre-
Cities are complex, organic, self-organizing and non- ation of knowledge have been concentrated (Graham
linear systems and they evolve and change con- 2004). These qualities and the possibilities that city
stantly. Contemporary cities can be considered as life offers are some of the reasons for the growing
immense numbers of interconnected citizens, busi- population in cities since the beginning. On the other
TOWARDS SMARTER CITIES | Concepts and Strategies - Volume 1 - eCAADe 34 | 651

hand, as Albino, Berardi, and Dangelico, (2015) ex- one definition might not be possible. However, there
plains the metabolism of cities generally consists of the are numerous definitions in the scientific literature
input of goods and the output of waste with consistent and in governmental reports. For a thorough ap-
negative externalities, which amplify social and eco- praisal of varied definitions Albino, Berardi, and Dan-
nomic problems. Cities rely on too many external re- gelico's (2015) paper is a great resource. They anal-
sources and, as a matter of fact, they are (and proba- ysed several different definitions and clarified that pre-
bly will always be) consumers of resources. It can be viously technology focused understanding of smart city
speculated here that the joys and opportunities of is changing and the smart city concept is no longer lim-
cities also lead to wicked problems such as poverty, ited to the diffusion of ICT, but it looks at people and
pollution, depletion of resources, traffic, congestion, community needs. With this in mind the definition
waste, and social exclusion. generated by the British Standards Institution (BSI
Growing populations in cities and their unequal 2014) which focuses on integrating diverse systems
consumption of the world resources have become namely, physical, digital and human systems is our
one of the greatest challenges for cities for some time preferred definition: Smart city is an effective integra-
now. The United Nations predicts, by 2050 70% of the tion of physical, digital and human systems in the built
nine billion world population will be living in cities [1]. environment to deliver a sustainable, prosperous and
Urban challenges such as planning, economic inclusive future for its citizens. It is good to point out
development, resilient water supply, integrated data at this stage that as Gil-Garcia, Pardo, and Nam (2016)
and security systems, responsive transport networks, explain being smart is not an end state, but rather can
environment protection, sustainable resource man- be an enabling condition that my lead to other desirable
agement, risk management, sustainable waste man- social, economic, or environmental outcomes.
agement, energy management, education, social It is clear that, data and ICT a big plays part in the
care and support, providing local services and fa- smart future urbanism. But equally there are other
cilities are putting immense pressure on the city major concepts contribute towards creation of smart
resources and city governance. Over the several cities. As Neirottio et al. (2014) highlights, ICT is
decades, the complexity and the speed of change in unable to transform cities without the human capital
cities, amalgamated with the need for integrated so- which brings the liveability of a city to attention. An-
lutions, have been presenting major challenges upon gelidou (2014) explains that, smart cities represent a
local authorities, who traditionally have tackled is- conceptual urban development model based on the uti-
sues in silos. Like many of the previous urban visions, lization of human, collective and technological capital
Smart City concept also tires to tackle the city chal- for the enhancement of development and prosperity in
lenges. This holistic approach is there to enable cre- urban agglomerations. Angelidou (2015) continues
ation of liveable, just, and sustainable and economi- to point that there are two shaping forces of the "smart
cally stable places. city" conception: Urban Futures (the past 100 years of
visions about the city of the future under the influence
Smart Cities of state-of-the-art technology) and the knowledge and
The question of "What is a smart city?" is not an easy innovation economy. Apart from these shaping forces
one to answer. The answer depends on the where it is also important to look into the key players in
the focus is and who is giving the answer. Although the smart city arena. There are three key players to-
still a definitive definition is lacking, in many aspects wards the smart city concept: universities, industries,
this is by no means a terrible void. Smart City con- and local government. The scientific papers and prac-
cept is wide and many players and notions are in- titioner reports reveal the presence of a triple helix con-
volved and therefore capturing all these elements in cept of knowledge-society even if each key player has
652 | eCAADe 34 - TOWARDS SMARTER CITIES | Concepts and Strategies - Volume 1

Figure 1
What makes a
Smart City?
different aims: research and knowledge spreading for the technology driven concepts and tools such as
universities, business and profit for private companies, Open Data, Big Data, IoT, urban sensing, volunteered
local well-being and political consensus for public bod- geographic information, and electronic democracy.
ies (Dameri et al. 2016). And these concepts and tools are redefining the city
The scientific papers and practitioner reports re- and how to manage and govern the city. These re-
veal the presence of a triple helix concept of knowledge- cent explosions of ICT technologies within every as-
society even if each key player has different aims: re- pects of city life also altering the urban planner's
search and knowledge spreading for universities, busi- job description to where a planner needs to practice
ness and profit for private companies, local well-being smart urbanism with more awareness of the data-rich
and political consensus for public bodies (Dameri et al. city governance. Traditionally, the challenge for urban
2016). planners was the generation of meaningful and timely
information. Today, the urban planners' challenge is no
Smart City and Urban Planning longer the timely generation of urban data, rather, it is
In first account, it is obvious that smart cities, although in relation to how much information can be exploited
specifically developed as an idea during the past 20 and integrated successfully into the contemporary spa-
years, have a long history dating back to previous cen- tial planning and governance (Thompson et al. 2106).
tury's visions about urban futures. Although elusive, Luque-Ayala et al. (2016) points out smart urbanism
the visions were and continue to be an essential part is the consolidation and expansion of relatively novel
of urban planning and development discipline and they ways of knowing and thinking the city. They continue
have always been urging to move forward (Angelidou to tell us that in the interface between digital technolo-
2015). Urban planning as discipline is changing with gies and urbanism, the city comes to known through

data, algorithm, modelling and a combination of visual our study (Table 1). Another classification also influ-
and media channels. We think there is no doubt that enced our work comes from Neirottio et al. (2014)
technology plays a fundamental role in making cities where they classify Smart Cities based on domains
smarter however how cities utilizing the technology and sub-domains (Table 2). Although these various
and the data for the specific needs and requirements classifications differ slightly from research to research
for the citizens, is as important as the technology im- (more at, Lee J-H 2012; Manville 2014; Thompson
plementation the smart city process. 2015), the typologies in general cover similar areas.
Table 1
METHODOLOGY Six application
Since September 2014, data on smart cities, defini-
areas of Smart
tions, applications and locations have been collected
Cities.
for this ongoing project. This desktop survey in-
cluded:
• City resources- city websites, articles with re-

gards to strategies, visions, plans, projects,
After this initial analysis, compression between the
etc.
identified typologies from literature and from our
• General sources including websites
survey is completed. Any differences from these
• Smart City project websites
guides were noted to report back.
• Academic books and journals
Focus then given to the smart cities itself and
• Reports from government and non-
more in depth information on some of the countries
governmental organization
and the EU were in order to understand the smart
• Seminar and conference notes, presentations
city process in these geographic locations since have
After gathering data from these resources, sorting top most smart city applications and initiatives in the
and analyses this raw data took place. Initially the world. Following this section a reflection given fol-
total number of smart cities was found out. In this lowed by the concluding remarks and future research
paper, we classed cities as "Smart" even though they directions.
might have one smart city project or an initiative and
not a "smart city" per se with a view that the smart
Limitations
The results of this conference paper have to be inter-
city is a process rather than an end product. One of
preted within the limitations of the research method-
the main reason behind this, that smart city term en-
ology and choices that were made by the researcher.
compasses 3 layers according to Manville (2014) and
Limited number of case studies has several implica-
these layers are Smart City projects, Smart City Initia-
tions. Initially, it is almost impossible to generalize
tives and Smart Cities (Smart City projects are a sub-
from this sample size. Following this, it should be
category of Smart City Initiatives which in turn are a
pointed here that this quantitative research is the ini-
sub-category of Smart Cities).
tial stages of this ongoing research and consulting
In order to find out the typologies of smart cities,
all the actors in Smart City developments will make
similar work from previous research were identified in
the basis for the qualitative research which will make
the literature and six application areas defined by the
the second stage of this study. Within the time-frame
Centre of Regional Science at the Vienna University of
and scope of the research qualitative research was
Technology (smart economy, smart mobility, smart en-
not possible.
vironment, smart people, smart living, and smart gov-
ernance) (Giffinger et al. 2007) set the foundation for

Table 2 smart cities were available to analyse. Out of these
Classification of applications 31% based on Living Labs, 22% on Smart
Smart City domains Government, 16% on Smart Environment, 13% Open
and sub-domains, Data, 11% on Smart Mobility, 7% Smart Economy, 5%
(adapted from Smart Living, 4% Smart People, 1% on Digital Initia-
Neirottio et al. tives, and rest of the 2% of the applications were a
2014). combination of Smart City from scratch, education,
health, innovation, planning, public safety, smart in-
frastructure ( Figure 2).
It can be said that the "smart city solutions"
mainly brought by big IT companies initially and in
general they are created for a specific city or a com-
munity. However as Chourabi et al. (2012) explain,
it is also clear that the smart city projects have an
impact on the quality of life of citizens and aim to fos-
ter more informed, educated and participatory citizens
and smart city initiatives allow members of city to par-
ticipate on the governance and management of the city
and become active users.
80% of all smart city applications in our survey
came from Chinese (20%), European (54%) and In-
dian (6%) cities. For this reason examples from some
of these locations are further examined.
Approaches to Smart City Developments

RESULTS AND DISCUSSION Our Smart City survey indicates that there are two dif-
As explained previously, by surveying the smart cities ferent approaches to develop a smart city. A typical
across the world we aimed the find out what smart top-down approach can be observed in the India's
city applications or domains are popular choices to Smart Cities Mission prepared by the Government of
implement. applications India, Ministry of Urban Development in June 2015
In our survey of smart cities, we have observed [2] and the Chinese Government's latest China's Five-
1101 smart cities (proposed, just embarking on the Year Plan (13th Five year plan: 2016-2020) encom-
project or classed as smart city already) and exam- passes the National New-type Urbanisation plan for
ined the available data from these initiatives. This the same period. The previous Five Year plan had also
process indicates that apart from the six application planned investment in Chinese Smart Cities agenda.
areas mentioned above, there are different smart city However in Europe although there are some top-
application types also in operation, for example; Liv- down approaches, in city level Europe has more of a
ing Labs (Figure 2). It should be pointed here that bottom-up approach for development of smart cities.
many cities analysed have more than one smart city
China. China is one of the countries that adopted
application. Also it is important to note that some
top-down approach. In January 2013, the Ministry of
of cities although they are classed as "smart" in the
Housing and Urban-Rural Development (MOHURD) for-
literature we could not access the types of applica-
mally announced the first list of national pilot Smart
tions that is in operation for those cities yet. How-
Cities (Yang et al. 2016) and as September 2013 total
ever, overall 1707 smart city applications from 1101

Figure 2
Smart City
Application types.
of 311 cities in China have proposed or are embarking Cities, National New Urbanisation Plan (2014-2020),
on smart city developments, including all cities above and Guidance on Promoting the Sustainable Develop-
the sub-provincial level, 89% cities at the prefectural ment of Smart Cities (Yang et al. 2016).
level and above, and 47% cities at the country level India. Smart City Mission is the one of the most com-
and above. The 12th Fiver plan budgeted to spend ap- prehensive selection process of developing Smart
proximately 220 million Euros on its Smart Cities project Cities. The Mission's objective is to promote cities that
(Yanrong and Whyte 2014). It is speculated that the provide core infrastructure and give a decent quality of
Chinese government is more focused on the tech- life to its citizens, a clean and sustainable environment
nological issues in developing Smart Cities. With and application of 'Smart' Solutions. The focus is on sus-
this understanding in mind, as Li, Lin and Geert- tainable and inclusive development and the idea is to
man (2015) explains the technology is seen as the cen- look at compact areas, create a replicable model which
tral importance, and 'smart city' can be divided into will act like a light house to other aspiring cities (MoUD,
four layers, that is, sensor layer, network layer, plat- 2015). Each aspiring city needed to compete in two
form layer and application layer. Two types of fund- stages of the selection process. These are: Stage 1 of
ing approaches are set for Chinese Smart Cities: Fully the competition: Shortlisting of cities by States and
supported by government or PPPs (More informa- Stage 2 of the competition: The City Challenge round
tion can be found at Li et al. 2015). There are cur- for selection. The proposals needed to be modelled
rently no laws and regulations directly governing Smart on one of the strategic components, namely: city
Cities in China. The Chinese government has, how- improvements (retrofitting), city renewal (redevelop-
ever, introduced a number of guidance notices: No- ment), and city extension (greenfield development)
tice to Speed up the Project Implementation of Smart and a Pan-City initiative in which Smart Solutions are

applied covering larger part of the city. The imple- and under the UK Goverment's Future of Cities Fore-
mentation of selected smart city projects will run by sight project; City visions: Newcastle 2065, City vi-
body (Special Purpose Vehicle - SPV, a limited com- sions: Birmingham and West Midlands 2060 are some
pany in which State ad Urban Local Bodies 50:50 eq- examples from UK.
uity shareholding). This organisation will plan, ap-
praise, release funds, implement, manage, operate City Level Cases
and evaluate Smart City development projects [2]. Although many to chose from, for this paper two
Europe. Our survey showed that more than 78% of cities selected from China and Europe since these lo-
the smart cities in Europe came from 8 countries (Ger- cations where the main contributors towards the to-
many 16.46%, UK 13.01%, Spain 12.80%, Italy 9.96%, tal number of smart cities we have surveyed origi-
France 8.33%, Poland 8.13%, Romania 5.08%, Nether- nally.
lands 4.67%). Rest of the %21.5 comes from 21 dif-
Bristol, UK.
ferent countries in Europe (It should be pointed here
that percentages comes from the amount of Smart • UK's eight largest populous city.
City applications in that specific country and does • Bristol, in May 2016 were selected as the UK's
not take into account the size and population of the leading smart city, sharing this spot with Lon-
country in this initial analysis). don.
The European Innovation Partnership for Smart • Smart City Bristol is a collaborative pro-
Cities and Communities was launched in 2013, and it gramme between the public sector, business
seeks to establish strategic partnership within the Smart and community and led by the City Council.
City arena and aims to boost the development of smart • Aim: Use smart technologies to meet City's
technologies. EU Policy Framework for Smart City De- targets to reduce CO2 emissions by 40% by
velopments based on the EU 2020 goals (20% reduc- 2020 from a 2005 baseline, as well as the city's
tion of greenhouse gas emissions from 1990 levels, 3% wider social and economic objectives
EU's GDP to be invested in R&D and innovation, 75% of • Drivers are based on economic, environmen-
20-64 year olds in work, and 20 million fewer people in tal, efficiency and transparency.
or at risk of poverty and social exclusion). When look- • Funded through sources such as the City
ing at the member state level and at the level of indi- Council budget and funding from the Euro-
vidual cities and regions, it is not surprising to find out pean Commission, UK Government, Technol-
an abundance of approaches to development and op- ogy Strategy Board, UK Research Councils and
eration of smart cities. This strong degree of indepen- private company funding.
dence of EU Communities results in very different so- • Investments: Metro bus, cycle city, enterprise
lutions being implemented with respect to any aspect zone, smart parking, broadband connections
city modernisation. It is also makes it somewhat diffi- solutions, smart energy, smart data, open
cult to create national or EU level approaches to a com- data, city data platform, Apps, smart mobility.
mon and coherent smart city development. However
this lack of top-down decision making authority is offset Beijing, Haidian District, China.
by the possibility to incentivise and encourage the cities
follow targets developed on EU or national level by way • Northwest of Beijing, the Haidian District is
of providing additional support (Yang et al. 2016). an important area and it is famous for its sci-
There are many city level documentations such ence and technology, culture, education and
as VIENNA 2050: Ensuring Quality of Life Through In- tourism.
novation - Adopting the Smart City Wien Framework, • Smart Haidian Top-level Design, Smart Haid-
ian Development Program, and Smart Haid-

ian Construction Program brings together the a strategic vision, infusion of intelligence to
overall programme for smart city. these various systems in soiled fashion is not
• Aim: establish smart administration, smart the correct approach. However these one sys-
parks, smart urban areas, smart homes and IT tem solutions if they are applicable can act as
industry HUB demonstrators and should be utilized as part
• Smart Haidian is built under the full respon- of the whole smart city / future city vision.
sibility of Haidian District Government, with • Biggest numbers of smart cities comes from
Haidian's smart city industry alliance, project locations that have either a direct top-down
management units, consulting and design approach (China and India) or an in-direct top-
units. down approach (Europe).
• Haidian invested about 300 million Yuan an- • Achieving the first implementation stage is
nually in 2011 and 2012 and 500 million Yuan only the part of the problem. Barriers such
in 2013, which mainly comes from district as moving beyond pilots, securing further fi-
government financial allocation. nance, digital infrastructure, human capital
• Wireless and fiber optic and other basic net- (education and establishing understanding)
works and data centers in Smart Haidian are and most importantly an overall vision to take
shared by different projects. ICT infrastructure these projects further needs to be considered.
is shared via a cloud platform. Haidian spa- • Cities also need to consider the vulnerabil-
tial data sharing platform is built for the Haid- ity of complex technological processes to the
ian GIS technology and other business appli- many unintentional and / or malicious inter-
cations. ventions.
Information on the sample cities are gathered from, Whether we are adding another layer of complex-
Yang et al. (2016), [3], [4], [5]. ity to the aging siloed structure of the city systems
or whether the utilization of data and technology
REFLECTIONS will prevail a more successful urbanism for the future
Increasingly, neither state intervention, nor neoliberal cities is still remain to be seen since many of smart
market solutions are seen as satisfactory approaches city initiatives are new and therefore it is hard to as-
to urban planning challenges. The former is criticised sess the outcomes yet. However, it should be pointed
for its managerial inefficiency while the latter is criti- here that smart city concept is not only for creating
cised for its neglect of external and community interests new holistic systems but also it is about improving
(Thompson et al. 2016). We believe that smart ur- current systems as well.
banism shapes and manages complex and multidi- The majority of strategies for becoming "smart" are
mensional places by engaging multiple stakeholders. not something that can be achieved here and now but
This integration enables better design, delivery, gov- they imply a strategic approach to fulfilling a long-term
ernance and maintenance of cities. aspiration The vision about the city of the future is an es-
sential driver of the smart city discourse, be it within, or
• As (Albino et al. 2015) points out the concept out of immediate grasp (Angelidou 2015). Recent lit-
of the smart city is far from being limited to the erature that explains the pillars of the smart city con-
application of technologies to cities. Cities and sist of connections (as networks and technological in-
"smart" cities for that matter, is for citizens and frastructures), Open data and Sensors (including citi-
therefore their involvement is a top priority. zens able to actively participate in a bottom-up way in
• Cities consist of many systems that inter- city activities). However we believe citizens will needs
act with each other continuously. Without to be considered as one of the fundamental support

mechanism for a successful smart city development not a standalone platform needs to be one of the first
(Figure 3). things to be established.
Smart City should not be only considered as cre-
Figure 3 ating new solutions/systems but also improving the
Smart City existing processes with the knowledge and under-
Framework. standing established from these new process, appli-
cations etc.
Balance between the data requirements and
data privacy needs to be well-adjusted and consid-
erations needs to be taken on a case by case basses
until standards established.
Solutions, although cities are different and can
have varied problems, should be replaceable.
Further Research
Future research will focus on maintaining the collect-
ing of data on smart cities across the world however
CONCLUSION apart from the quantitative analysis more in depth
qualitative analysis will be conducted to further this
Points to consider
research.
Smart city framework should have vision, people and
processes as the main drivers.
Political willingness and long-term commitment REFERENCES
is one of the key factors needs to be taken into ac- Albino, V, Berardi, U and Dangelico, RM 2015, 'Smart
count. Strategic vision is profoundly related to the cities: Definitions, dimensions, performance, and
initiatives.', Journal of Urban Technology, 22(1), pp. 3-
long-term commitment and will provide continuity 21
beyond routine election related changes in cities. Angelidou, M 2014, 'Smart city policies: A spatial ap-
Tackling common and local challenges should be proach', Cities, 41, pp. 3-11
the priority and these challenges needs to have a di- Angelidou, M 2015, 'Smart cities: A conjuncture of four
rect relation to the demands from citizens, cities, re- forces', Cities, 41, pp. S3-S11
gions and the country. Local strengths should also BSI, British Standards Institution 2014, Smart Cities – Vo-
cabulary (PAS 180:2014, 3.1.62), British Standards In-
influence the project selection process. stitution, London
Leadership with professional mechanisms to co- CAICT, China Academy of Information and Communica-
ordinate and consolidate needs to be in place and tions Technology 2014, Comparative Study of Smart
this needs to be backed up with concrete strategic Cities in Europe and China- Current Chinese Economic
objectives for the city. Report Series, Springer-Verlag GmbH, Berlin Heidel-
The impact of the smart city work on the city berg
Chourabi, H, Nam, T, Walker, S, Gil-Garcia, JR, Mellouli,
should not be underestimated. Necessary precau- S, Nahon, K, Pardo, TA and Scholl, HJ 2012 'Under-
tions in the form of public consultations, openness standing smart cities: An integrative framework',
and processes etc. need to be in place. And financial System Science (HICSS), 2012 45th Hawaii Interna-
and managerial challenges need to be considered at tional Conference, IEEE, USA, pp. 2289-2297
all stages. CRS, VUT Centre of Regional Science, Vienna UT 2007
Instead of creating another new silo for the city 'Smart cities – Ranking of European medium-sized
cities', report
processes, the understanding of that the smart city is

Dameri, RP, Negre, E and Rosenthal-Sabroux, C 2016 (ST/ESA/SER.A/352)., UN
'Triple Helix in Smart cities: a literature review about Yang, K, Clery, A and Di Liello, D 2016, Sector Report Smart
the vision of public bodies, universities, and private Cities in China,, The EU SME Centre and China-British
companies', 49th Hawaii International Conference on Business Council
System Sciences (HICSS), IEEE, pp. 2974-2982 Yanrong, K and Whyte, J 2014, Comparative Study of
Giffinger, R, Fertner, C, Kramar, H, Kalasek, R, Pichler- Smart Cities in Europe and China. EU-China Smart
Milanovic, N and Meijers, E 2007, Smart cities. Rank- & Green City Cooperation, White Paper. DG CNECT,
ing of European medium-sized cities, Final Report, EU Commission, China Academy of Telecommunica-
Centre of Regional Science, Vienna UT tions Research.
Gil-Garcia, JR, Pardo, TA and Nam, T 2016, 'A Compre- [1] http://esa.un.org/unpd/wup/Publications/Files/W
hensive View of the 21st Century City: Smartness UP2014-Highlights.pdf
as Technology and Innovation in Urban Contexts', [2] http://smartcities.gov.in/writereaddata/SmartCi
in Gil-Garcia, JR, Pardo, TA and Nam, T (eds) 2016, tyGuidelines.pdf
Smarter as the New Urban Agenda, Springer Interna- [3] http://www.huawei.com/en/news/2016/5/UKs-
tional Publishing, pp. 1-22 leadin
Graham, S 2004, The cybercities reader, Psychology Press g-smart-cities
Li, Y, Lin, Y and Geertman, S 2015 'The Development of [4] http://www.connectingbristol.org/
Smart Cities in China', CUPUM 2015, USA [5] http://www.europarl.europa.eu/RegData/etudes/et
Lonbardi, P and Di Torino, P 2011 'The Smart Cities, The udes/join/2014/507480/IPOL-ITRE_ET(2014)50748
Triple Helix of Smart City Visions', Presentation 0_EN.pdf
Luque-Ayala, A, McFarlane, C and Marvin, S 2016, 'In-
troduction', in Marvin, S, McFarlane, C and Luque-
Ayala, A (eds) 2016, “Smart Urbanism: Utopian Vision
Or False Dawn?, Routledge, pp. 1-15
Manville, C 2014, Mapping Smart Cities in the EU, Eu-
ropean Parliament, Directorate General for Internal
Policies, Policy Department–Economic and Scien-
tific Policy-IP/A/ITRE/ST/2013-02
MoUD, Ministry of Urban Development Government of
India 2015, Smart City Mission Transfrom-Nation, Mis-
sion Statement and Guidelines, Government of India
Murgante, B and Borruso, G 2015, 'Smart Cities in a Smart
World', in Rassia, ST and Pardalos, PM (eds) 2015, Fu-
ture City Architecture for Optimal Living, Springer In-
ternational Publishing, pp. 13-35
Neirotti, P, De Marco, A, Cagliano, AC, Mangano, G and
Scorrano, F 2014, 'Current trends in Smart City ini-
tiatives: Some stylised facts', Cities, 38, pp. 25-36
Smith, PD 2012, City: a guidebook for the urban age,
Bloomsbury Publishing , USA
Thompson, EM 2015 'City “is” Real-time', Proceedings of
the 33rd eCAADe Conference - Volume 1, Vienna Uni-
versity of Technology, Vienna, Austria
Thompson, EM, Greenhalgh, P, Muldoon-Smith, K, Charl-
ton, J and Dolník, M 2016, 'Planners in the Future
City: Using City Information Modelling to Support
Planners as Market Actors', Urban Planning, 1(1), pp.
79-94
UN-DoESA, United Nations, Department of Economic
and Social Affairs, Population Division 2014, World
Urbanization Prospects: The 2014 Revision, Highlights

From Moving Cube to Urban Interactive Structures
A case study
Vasilija Abramovic1 , Henri Achten2

1,2
Faculty of Architecture, Czech Technical University
1
vasilijaabramovic@gmail.com 2 achten@fa.cvut.cz
When thinking about the future vision of a city, having in mind recent
development in digital technologies and digital design tools we are inclined to
expect new building structures which incorporate this technology to better help us
manage the complexity of life, and to simplify our daily lives and tasks. The idea
behind this research paper lies in design of such structures, which could be put
inside an urban context and engage in creating a built environment that can add
more to the quality of life. For us Interactive architecture is architecture that is
responsive, flexible, changing, always moving and adapting to the needs of today.
The world is becoming more dynamic, society is constantly changing and the new
needs it develops need to be accommodated. As a result architecture has to
follow. Spaces have to become more adaptive, responsive and nature concerned,
while having the ability for metamorphosis, flexibility and interactivity. Taken as
a starting point of this idea is a specific module from graduation project in 2014
"The Unexpected city", where it was possible to test out first ideas about
interactive and flexible objects in an urban environment.
Keywords: Flexible architecture, Interactive architecture, Responsive systems
INTRODUCTION to achieve much better building functioning, being

In this paper we report on a PhD research which user friendly and communicative on both ends: user
has recently started. The field of the PhD research - building - user. We believe the logic of design has to
project is Interactive Architecture. We explore what be changed due to the rapid change of the technol-
is the phenomenon of Interactive architecture, and ogy and human demands. (Figure 1)
how objects (building structures) within an urban or-
ganism may be both flexible and interactive with its PROBLEM EXPLORATION
users. The word "organism" is deliberately used to We can observe around us in society that it seems
accentuate the author's belief that interactive means to develop more dynamically than before and that
responsive, flexible, changing and always moving. needs also have become more dynamic. The com-
We aim to develop a design by which architecture mercial lifespan of goods and products is decreasing
can become more interactive in the future. We hope because they quickly do not fulfill our needs. Unless

Figure 1
Interaction
relations.
they can follow and change with us, they become House" from 2008 [2], many examples of media/in-
expendable. The challenge we are facing in this re- teractive facades, or the Spielbudenplatz in Hamburg
search is how to establish an environment that is part from 2006 (Schumacher et al. 2010, 204-205). The fu-
of developing human culture? ture is happening. And it looks pretty much similar as
"The world around us is evolving. We are living in- the one drawn back in 60s on the images of the Living
side an evolution. As a practicing architect I find noth- city.
ing more natural than to look around me and imple-
ment relevant changes in to my own profession." (Oost- VISION OF RESPONSIVE AND CHANGING
erhuis 2012)
ARCHITECTURE
In our view, architecture is more than the phys-
We as people change constantly. We are given free
ical realization of a space. Like for example Tschumi
will and free movement. We adapt and adopt the
(1994) we believe that the events which take place
spaces, thus why the buildings would not do the
in architectural space are as important as that space
same? Leon C. Megginson said: "It is not the strongest
itself. We can observe a proliferation of events in
of the species that survives, nor the most intelligent. It is
contemporary buildings that are contrary to mono-
the one that is most adaptable to change" (Megginson
functional types that have been favored in Modernist
1963, as cited in Seydel 2012).
approaches. Peter Cook said that cities should gen-
Adaptability in architecture may be an essential
erate, reflect and activate life, their structure orga-
component in creating sustainable architecture. Pre-
nized to precipitate life and movement. In short, the
serving and adaptive reuse of a building instead of
cities should breathe together with the people, turn-
demolishing it and erecting a new one in its place can
ing them into an alive organism . "The pulsation of the
contribute significantly to the environmental sus-
city life is fast, so why not that of its environment? It re-
tainability. However, adaptability comes as a cost
flects rise and fall, coming and going...change. So why
(the whole sum of the system that is capable to
not build for this?" (Cook 1972)
change may be more expensive than a passive op-
50 years after Archigram's visionary Living City
timized system). Therefore the return of the higher
project we see proposals and prototypes that refer-
cost of an adaptive building should lead ultimately to
ence back to the ideas from the Living city and its
a better solution. Highly dynamic systems such as lo-
way of above ground and diagonal transportation,
gistical nodes, places of events with large numbers of
such as in Elon Musk's ambitious 2013 whitepaper for
people, and so on may benefit in this way from adapt-
Hyperloop [1]. Other examples are N55's "Walking

ability (Achten 2011). In particular we are looking at public spaces which
"Intelligent building systems are used to create in- are abandoned and today represent holes inside the
teractive architecture that responds to users' require- city structure.
ments in automatic or intuitive ways. It is architecture Architecture of industrial heritage is as well an in-
that is receptive to people's needs to alter their envi- teresting point of focus which could be a new field of
ronment and has mechanisms in place to do so eas- action for this "new" type of architecture. There are
ily.(..)...sensors operate actuators that can trigger a wide many examples throughout the world where indus-
range of actions - kinetic systems that physically alter trial buildings are being converted to cultural cen-
space, services that alter the environment or materials ters. Their flexibility gives an option for hosting mul-
that alter their state." (Kronenburg 2007) tiple different functions, at various scales. Often in-
With our idea of responsive structures architec- dustrial buildings are located in areas that have eco-
ture should become more self-conscious, and ori- nomically stagnated in cities, and which thus require
ented towards all individual changes in urban matrix revitalization. For this specific research industrial her-
as well as in more private sector such as residential itage is of importance in a way that it could be used as
buildings. Technological advances are making such ground for technological and design improvements
ideas increasingly feasible (see for example Schu- by the means of interactivity and flexibility in archi-
macher et al. 2010 for a comprehensive overview of tecture. These types of buildings will be included in
technologies applied in adaptable buildings). the research process from theoretical view, but also
"Public, the people almost always stand behind the as a case study material.
traditional architecture, behind the traditionalists. In
the public eye, architecture is about comfort, about Residential sector
shelter, about bricks and mortar. However, there is In this research project we start in an urban context,
more and more of those for whom architecture in not and later in a residential one. The residential sector
necessarily about comfort, but is also about advanc- is different from urban sector in many obvious ways -
ing society and its development" (Architecture and Dis- scale, number of people involved, environment, con-
junction, Bernard Tschumi, 1994). text, and so on. However, we can apply a number of
principles derived from experiences with urban con-
PLACE OF INTERVENTION text to residential sector. First, in both cases we start
In our research, we look at interventions by means of with an anonymous base of users. In urban environ-
interactive structures on two levels: the urban envi- ments this base may stay anonymous, whereas in res-
ronment and in the residential sector. idential context the system is more likely to get to
learn the principal inhabitants. Second, in urban en-
Urban environment vironments the number of actors involved with an in-
Author's belief is that the public space is the first teractive system can vary greatly from just a few peo-
point of action where these designs must take place, ple to large numbers of people. Deciding on the right
and where they must be tested. By observing peo- interaction approach means reading and balancing
ple's actions we would be able to realize what is out many inputs. Such decision making is also neces-
needed within an urban space, and to answer the sary in a residential context, although the number of
questions around our design. In the first stage of the actors is much more limited. Finally, by looking at the
research this will be done in a virtual environment, in urban context we may get clues what people need
particular the Desk-Cave system at the Faculty of Ar- from the urban context that they cannot get at home.
chitecture at CTU Prague. Following this step, more These clues may be important pieces of information
focused experiments will take place in real situations. to improve the residential context.

UNIVERSAL OR LOCAL? like, being nomads out of need or wonder.
The question after determining the place of inter-
vention would be whether it should start as some- ETHICS CONCERNS
thing more universal - as a structure that has no con- Our perception of what is comfortable, of how the
nection to where it is being built, or rather as some- building should behave at the moment, is not going
thing more local, with respect to a specific location. to be the same in the future. With ever-changing per-
Both approaches have different implications. Some- ception of us - architects and creators of living space,
thing local would have to inherit cultural and local el- the main question we need to ask ourselves is "How
ements, thus respond to them, while something uni- to design a building that easily changes with us in the
versal should be able to be utilized and put anywhere future?". A building that can be "upgraded". Without
in the world. To establish something interactive that such changes the building becomes rapidly old and
makes sense globally has the risk that the interac- unsuitable, in need for replace
tions are reduced to lowest common denominator. Another important point is how compelling
This could lead to rather banal and simplistic inter- should the interaction become - we should avoid that
actions, of which additionally it would be difficult to the building dictates our lives, and habits, and pushes
prove the universality as we cannot test a system ev- us - users, to comply with its performance. In fact, we
erywhere. As we stated before, an interactive archi- could say that in such an overbearing situation there
tecture would have to be responsive and interactive is no case of interaction, but rather one-way commu-
not only to the users but also to the surrounding - nication or even control.
this surrounding then also plays a significant role. We Interactive architecture absorbs inspiration from
might even argue that different locations inside the other industries, such as for example car manufac-
same city, or a country differ enough for the structure turing, user interface design, and aero-spatial de-
not to be designed specifically for one - "no more lo- sign. Developing technologies contribute to the pos-
cating in a fixed place, but a new heterotopia" (Tschumi sibility of new and better constructional and opera-
1994). Therefore we need to strike a balance between tional strategies. Principle is creation of spaces which
generic and specific interaction strategies. are able to maintain a dialogue with their users, not
This is a matter of concern for spatial urbanism only responding to their demands, but pro-actively
as well. The idea of the nomad, of mobility, of the engaging themselves in all kinds of featured spatial
transformable and permutational became one of the activities. There are however many problems how
main obsessions of the spatial urbanists in 20th cen- to reach true "interactive architecture". The least of
tury. This was a function both of prospective thought them are of the technological nature, but the most
- the desire to predict the change, or at least to be difficult once to overcome relate to theoretical, cul-
prepared for an unpredictable change. What is un- tural and social questions. Important questions are
clear in this happening is whether this is a condition what to design, and how to design those interactions
or a remedy. If man were being liberated to wonder, - a lot of this is being tried out without established
or if they were only displaced (Busbea 2007). Prob- theoretical or methodological frameworks (Achten
ably both is the answer, since we have witnesses of and Kopřiva 2010).
the first case on an everyday cycle, and the second These problems should not be only concerning
one can be connected to the current world crisis and the "machine", but the identity as well. Similarly as in
with refugees being nomads out of need. Here we re- Mass-customization flexibility and client-orientation
fer back to the original proposed solution - building is desired, we should "mass-customize" interactions
flexibly and in an interactive manner, so people can so that they best fit user-needs and desires. The phys-
find their place in whatever environment they feel ical part of interactive systems (sensors, actuators,

and material components) are less flexible than the project in 2014 "The Unexpected city", which was also
software part, which can employ learning strategies part of an international competition given by Daniel
to adapt its behavior. In all cases, some kind of user- Libeskind. The proposed interactive structure in this
adaptable behavior and identity needs to be estab- project is a cube of 7m x 7m, which would serve as
lished. an experiment model. The module can be moved up
Overall mass production has its influence on the and down thanks to four telescopic columns which
city's image as well. Global corporations and busi- are part of its construction. These columns expand
ness trends tend to make cities look more like one and descend when needed. The ceiling is attached to
another so the success and individuality gets lost. the columns and therefore rises and descends along
The fear which Archigram shared was that architec- with them. Walls are made of special textile water-
ture alone cannot be enough to give this feeling of a proof texture. As such they are stored in the com-
place, to give identity. partment attached to the ceiling, and they can roll
"I have a desire for out, or roll in based on the need. This structure
The built environment would already give a starting point towards the re-
To allow me to do search being flexible and able to change : 1 able to
My own thing ." go up to 4m of height and therefore become an ob-
ject, or 2 able to go down to -0,4m and therefore be-
come either a platform to walk on/sit on, or used as
What happens with nature?
a stage for an urban setting. It can provide seating
There are social groups which are troubled with tech-
elements, light sources, atmosphere creation, am-
nological advances discussed above and who have
bient and direct communication, active monitoring,
the fear to lose touch with nature, while making this
acoustic control, access control and many other. Spa-
world a fast and a dynamic machine. The poem
tial volumes change and continuously redefine their
"All Watched over by Machines Of Loving Grace"
relationship with the ground. This volumetrically
by Richard Brautigan (according to [3] published in
chameleonic structure can accommodate multitude
1967; according to [4] published in The Realist, 1968,
of functions. This is where all simplicity ends and the
Issue 81, August, p. 11) explains it very well in just a
whole set of complications and technical tricks enter
few lines.
the scene. These cubic modules, represent the idea
"I like to think of architecture of transformation or so called meta-
(right now , please !) morphosis.
of a cybernetic forest Mentioned structure is just taken as a proposal -
filled with pines and electronics
starting point, whilst it will probably be changed dur-
where deer stroll peacefully
past computers
ing the design process and moved towards a differ-
as if they were flowers ent solution/strategy. The main research questions
with spinning blossoms ." which are proposed are originating around the no-
tion "interactive" itself. Research question number
Similar as in the poem, the author feels that the face one would be: "What types of interaction technology
of the city of tomorrow should be the integration of do people in public places prefer? ". That is some-
the natural with the artificial. thing that can be tested and analyzed in the con-
text of the public sphere. More specific formulations
CASE STUDY of this question could be for example "If provided a
Many of the ideas presented above have their origin choice set of various interaction technologies, which
in a case study for the design of an interactive and do people choose in a particular public situations?",
flexible module. This study is the author's graduation

"What is the acceleration speed at which the walls of 4) ARCHITECTURE (main host)
the module should move so it can be pleasant and in- 5) FLEXIBILITY (main option)
teresting for people?", "How is the shift of control of 6) INTERACTIVITY (main way)
such modules perceived by the public?"
Another important factor is the building skin
since this is the most common interface between Figure 2
public-private space between building and urban Elements of
space, and also one of the most dominant elements interactive
that makes up the visual boundary of urban space. architecture.
Thus the building envelope is not anymore just
something which is closing the inside space. What if
the envelope, the façade could be changing in such
a way that when needed it could be invisible? Or if
we could make the building skin react to people, for
example based on the occupancy within? Also since
the starting idea is to work within the public spaces,
it would be as equally important if the object/build-
ing envelope could react based on the public activity
in the outdoor space.
Figure 3 shows the relations introduced earlier in Fig-
Flexibility - approach ure 1, applied to the modular cube from the case
If we trace processes that form living organisms, it's study (user, event, surroundings, architecture, flex-
obvious that no organism had initially been shaped ibility and interactivity). The user causes a certain
in all its intricacy. They always start with a single number of events inside of a building, which then
cell which multiplies itself numerous times. When communicates back to the user, whereas at the same
a critical mass is reached, cells start to differentiate; time both factors are being affected by the surround-
they begin to form tissues and organs. This process ing as well.
is called cell differentiation in biology, and leads to
autopoiesis of stable natural systems. Analogically, Figure 3
buildings can be designed and created in a similar Relation between
manner. In the case study we can see the same prin- six elements
ciple: a simple modular cube which was then multi- applied to module
plied several times until the needed number and area cube.
for that particular location was reached. Later what
gave the required "shock" to this complex was the
juxtaposition of the events behind the façade.
From here we can propose that basic elements of
how we should start developing new way of thinking
and designing the architecture should be (see Figure
2):
1) USER (main player),

2) EVENTS (main action),
3) SURROUNDINGS (main place),

GOAL make it each time a step closer to our goals. Several
Architecture should not only sustain, but also needs ways would be possible in achieving such results and
to entertain, instruct, explore, and optimize perfor- analyzing them, such as: creating a prototype and
mance in various degrees (Achten 2013). While some observing the reactions by ways of videotaping, ob-
of this may be achievable through passive means, it serving (professional observers), surveys, and simula-
is evident that the future of architecture is a design tions.
which would be simple for use, but changeable due
to the rapid change of the lifestyle today. Change- CONCLUSION AND METHODOLOGY
able design is perceived as a flexible architecture in- The world is changing, and so is architecture. We
corporated with interactive design technologies, and have to figure out a way for architecture to follow
new design principles. In this research we aim to de- these changes, by designing complexity with sim-
velop and test one or more simple structures which ple methods. This research aims to develop architec-
offer multitude of functions, while at the same time tural solutions capable of sustaining themselves in its
being responsive and user active. dynamic spatial, social and natural environment. At
"Since contemporary ways of life are so rapidly the same time these solutions should actively engage
evolving in all their aspects, there is an urging neces- with its users and the surroundings. Cities should
sity for architectural spaces to be enhanced in ways be designed in such a way that they can breathe
that would allow them to perform an active dialogue along with the city life. By this we would be creat-
with their fluctuating content; to dynamically deal with ing solutions which can become interactive and flex-
changing needs of social groups, as well as to directly ible while adding quality to our lifestyles, and to cities
serve particular individuals. This trend forces architects in general. The ideas which were analyzed and men-
to design flexibly, to take into account potential emer- tioned in this paper will be used to help developing
gence of new spatial requirements that cannot be an- this philosophy further more in this research, which
ticipated before the actual building use comes to place can contribute to better designing of an interactive
and which can dynamically change over time." (Jask- and flexible structure in the future.
iewicz 2008) Metamorphosis and transformation of the space
In the first stage of the research, we expand the are the qualities which we must seek. Flexibility is the
theoretical literature review. Based on this we iden- main component an object needs to demonstrate. If
tify a number of interactive urban case studies that we want to build in a more flexible and varied man-
are first tested in a virtual environment (CTU Desk- ner, the architect must design at a higher level of ab-
Cave) and after fine-tuning implemented in a real sit- straction - to allow the end user to fully engage with
uation. As stated earlier, we first aim at urban envi- the final design product.
ronments and then make the step towards the res-
idential sphere, where we aim to apply the lessons
learned from the urban environment. This research
ACKNOWLEDGEMENTS
Presented work is a part of the PhD research project
project plans to focus on this specific problem - How
"Interactive Architecture" carried out by the author -
an interactive urban structure should look like
Vasilija Abramovic, at Faculty of Architecture Czech
and how can it be implemented inside an urban
Technical University in Prague, Cabinet of Architec-
matrix? What would be its abilities? How could it
tural Modelling (MOLAB) and supervised by Henri
serve to people? Our main methodology is research
Achten. The PhD researcher is supported by a grant
by design - in which we develop our understanding
received from Ministry of Science of Montenegro
of interactivity through a sequence of designs. By
through "National scholarship program for excel-
analysis, we would be able to improve the design and
lence."

In the text is referenced Master thesis project Birkhäuser Verlag, Basel
"The Unexpected City", done by the author - Vasilija Seydel, ER 2012, Veerkracht. Rede uitgesproken bij het
Abramovic, at Faculty of Architecture Czech Techni- afscheid als hoogleraar Communicatiewetenschap en
Psychologie aan de Universiteit Twente op 12 septem-
cal University in Prague in 2014, supervised by prof.
ber 2012 door Prof.dr. Erwin R. Seydel, Universiteit
Vladimir Sitta. Twente, Twente
Travel to the eCAADe 2016 conference was spon- Tschumi, BT 1994, Event-Cities, The MIT Press, Cam-
sored by a scholarship given by Foundation "Nadání bridge, Massachusetts
Josefa, Marie a Zdenky Hlávkových", from Prague [1] http://www.spacex.com/hyperloopalpha
Czech Republic. [2] http://www.wysingartscentre.org/archive/exhibit
ions/walking_house/2008
[3] http://www.ep.tc/realist/81/11.html
REFERENCES [4] http://www.brautigan.net/machines.html
Acharya, LA 2013, Flexible Architecture for the Dynamic
Societies, Master's Thesis, Faculty of Humanities, So-
cial Sciences and Education University of Tromsø
Achten, HA 2011 'Degrees of Interaction: Towards a Clas-
sification', Proceedings of the 29th International Con-
ference on Education and Research in Computer Aided
Architectural Design in Europe, Ljubljana: University
of Ljubljana, pp. 565-572
Achten, HA 2013 'Buildings with an Attitude. In Stouffs,
R. and Sariyildiz, S. (eds.), Computation and Perfor-
mance', Proceedings of the 31st eCAADe Conference –
Volume 1, Delft University of Technology, Delft, The
Netherlands, pp. 477-485
Achten, HA and Kopřiva, MK 2010 'A Design Method-
ological Framework for Interactive Architecture. In
Schmitt, G., Hovestadt, L., van Gool, L. et al. (eds.) Fu-
ture Cities', Proceedings of the 28th International Con-
ference on Education and Research in Computer Aided
Architectural Design, Zurich: ETH Zurich, pp. 169-
177
Cook, PC 1972, Archigram, Princeton Architectual Press,
New York
Jaskiewicz, TJ 2008 'Dynamic design matter(s)', First In-
ternational Conference on Critical Digital: What Mat-
ter(s)?
Jenks, CJ (eds) 1997, Theories and Manifestoes of contem-
porary architecture, Academy Press
Fox, Kemp, MF, MK 2009, Interactive architecture, Prince-
ton Architectual Press
Oosterhuis, K 2001, Architecture Goes Wild, 010 Publish-
ers Netherlands, Rotterdam
Oosterhuis, K 2012, 'Simply complex, toward a new kind
of building', Frontiers of Architectural Research, 1(4),
pp. 411-420
Ota, KO (eds) 2011, Project Japan. Metabolism Talks…,
Taschen
Schumacher, SM, Schaeffer, SO and Vogt, VM 2010, Move,

SenCity - Piloting Intelligent Lighting and User-Oriented
Services in Complex Smart City Environments
Henrika Pihlajaniemi1 , Eveliina Juntunen2 , Anna Luusua3 ,
Mirva Tarkka-Salin4 , Johan Juntunen5
1,3,4
Oulu School of Architecture, University of Oulu 2,5 VTT Technical Research
Center of Finland
1,3,4
{henrika.pihlajaniemi|anna.luusua|mirva.tarkka-salin}@oulu.fi
2,5
{eveliina.juntunen|johan.juntunen}@vtt.fi
New operation frameworks and user-oriented design methods are needed to make
better use of new innovative technologies within smart city contexts. This paper
addresses the design and research of intelligent lighting and user-oriented
services for smart city environments. It presents the problem setting and research
and development methods of the SenCity project. The project will pilot smart
lighting solutions in six Finnish cities in different kinds of urban environments. In
the pilots, the target is to employ lighting infrastructure as a service platform - an
Internet of Things backbone - in the intelligent city. Together, separate pilots in
different cities around Finland will create a living lab ecosystem for developing
and testing innovative solutions. The specific objective of this paper is to present
the concept of a platform as defined and applied in SenCity project. The
presented framework forms an operational model for creating intelligent lighting
and digital services in smart cities by integrating relevant technologies, users'
needs, and business into an interactive system. In the paper, the framework is
applied to a selection of pilot cases with specific themes to introduce its usability
in real world implementations.
Keywords: smart city, sensing, intelligent lighting, smart lighting, user-oriented

design
INTRODUCTION paper addresses the design and research of intel-

The world-wide interest to smart city development ligent lighting and user-oriented services for smart
as well as impetus to implement intelligent lighting city environments. It presents the problem state-
solutions for energy savings is opening up new col- ment and research and development methods of
laboration patterns between urban development, re- the project "SenCity - Intelligent lighting as a Ser-
search and business. New operational frameworks vice Platform for Innovative Cities". The project pilots
and user-oriented design methods are needed to smart lighting solutions in six Finnish cities in differ-
make better use of new innovative technologies. This ent kinds of urban environments. The research has

three focus areas: 1) to study users' needs and expe- complex. Thus, it is important to reflect upon what
riences of intelligent lighting, 2) to develop and test the users' needs are in different kinds of environ-
sensoring, data analysis and communication meth- ments, what kind of information is relevant for the
ods and technologies needed for the user-centric de- users' needs, how the information can be analysed
signs, and 3) to generate business opportunities for and how lighting should behave to best serve users'
smart, data-based services. In the pilots, the aim is needs. In order for lighting to be truly intelligent, not
to employ lighting infrastructure as a service plat- only the system should be smart but the overall de-
form - an Internet of Things backbone - in an intelli- sign should be smart as well: Intelligent lighting can,
gent city. Together, separate pilots in different cities besides energy savings, offer added value for urban
around Finland will also create a living lab ecosys- environments on various levels of experience. (Pih-
tem for developing and testing innovative solutions lajaniemi 2016.) However, even though almost ev-
in the future. The SenCity project group consisting erything is possible with new technology, it is chal-
of research institutions, cities, and companies is pre- lenging to find technologically reliable and econom-
sented in Figure 1. ically feasible means to operate in complex environ-
ments which have many changing factors. How can
we strike a balance between the desired simplicity of Figure 1
a system and the complexity of the design task? SenCity project
Technologically, intelligent lighting can be seen pilots intelligent
as part of a larger Internet of Things (IoT) develop- lighting and smart
ment. IoT can be defined as an intelligent and inter- services in different
operable node interconnected in a dynamic global kinds of urban
infrastructure network. IoT means connecting every- contexts.
thing around us, such as household devices, mobile
phones, cars, even cities and roads, to the Internet
and/or to each other. As such, IoT seeks to imple-
ment the connectivity concept of anything from any-
where at any time (Ali et al. 2015). This leads us to our
broader research context: smart cities.
Smart City development - From technology

towards better urban living
According to the large number of varying definitions
found in literature, a smart city is a combination of
modern ICT technology, economic ambitions, and
Intelligent lighting social and communal objectives in urban environ-
Intelligent lighting, or smart lighting, can be defined ments. Due to increasing environmental awareness
as a lighting system which gains real-time informa- and economic and social importance of cities, more
tion of its environment and users through sensors, and more focus has shifted from technology to peo-
and adapts its behavior accordingly. The information ple and community. Besides conserving natural re-
is analysed and processed in the system into lighting sources, the opportunities for increasing the quality
control commands based on pre-defined parametres of urban living, regarding city administration, educa-
and rules (Guo 2008). The principle in itself is sim- tion, healthcare, public safety, real estate, or trans-
ple but the complex real world setting, i.e. the city portation for example, have been recognized. Also,
with its inhabitants, renders the design task highly

sustainable economic growth more related to mod- amining this literature, the crux of the issue seems
ern technology and services is considered important to be the lack of participation and democracy in
in the current smart city environment. (Albino et al. the field of smart city development, both on the
2015.) level of viewpoints and practices. (Kitchin 2014,
The interest on smart city development is global. Hollands 2008) To surmount these issues, the op-
For example, India has a grand project of develop- erational framework that is utilized in the SenCity
ing 100 smart cities, while Masdar City in United Arab project, is introduced.
Emirates and Songdo in South Korea have launched
projects of their own. Also, South Africa has a smart OBJECTIVES AND METHODS
city project in Modderfontein. In Europe, the de- The general objective of the SenCity project is to pilot
velopment towards smart cities is active, and sev- intelligent lighting solutions and to employ lighting
eral European cities, like Amsterdam in Netherlands, infrastructure as a platform for user-oriented services
Santander in Spain, Lyon in France, and Morgen- in different kinds of smart city environments. The pi-
stadt in Germany are investing to develop into smart lots vary in complexity level concerning both design
cities. (Frost & Sullivan 2015) Amsterdam with dozens and research targets. The specific objective of this pa-
of smart city related projects is an interesting exper is to present the concept of a platform as defined
ample of European activities. Somewhat similar to and applied in the SenCity project. The presented
our project, in Amsterdam, smart lighting is imple- framework forms an operational model for creating
mented with sensors to control street lighting and intelligent lighting and digital services in smart cities
the lighting infrastructure is also used for services be- by integrating relevant technologies, users' needs,
yond the lighting functionality. For example, traffic stakeholders, and business into an interactive sys-
flows are controlled with help of coloured lighting. In tem. In the article, the general framework of the plat-
addition, a smart lamp post has been sketched with form is presented. In addition, the framework is ap-
functions like sensing various data such as pedestrian plied to a selection of pilot cases with specific themes
count or environmental data, communication, assist- to introduce its usability in real world implementa-
ing in navigation, and delivering notifications, alerts tions.
and advertising. [1] An energy provider, Alliander, The pilot cases are presented discussing design
has an "Open Smart Grid Platform" which allows us- and research aims and methods to be applied. The
ing and managing of different objects independent relevant users' needs and users' experience aspects,
of the different suppliers. The Alliander's system has technological needs, stakeholders and service re-
been employed on lighting control in Amsterdam. lated business potential are presented. The pilot
[2], [3] Also in Santander, there are over 25,000 wire- projects will be realized and evaluated between sum-
less devices over the town monitoring the environ- mer 2016 and autumn 2017. Those results will be
ment. The sensors help in lighting control, planning published in the future.
of garbage collection, tracking traffic, and monitor- Methodologically, the basis of our research can
ing air quality [4]. The project has reduced energy be found in traditions of research-by-design, quali-
costs by 25% and garbage pickup costs 20% [5]. tative research and technological research. Our re-
The smart city development already carried out search can be defined as transdisciplinary research,
around the world sets good reference to our ob- referring to Gibbons et al. (1994). The research sub-
jectives in the SenCity project. In addition to the ject concerns various research disciplines such as re-
smart city definitions and actual implementations search of lighting design and experience, architec-
discussed above, there is also a literature which ex- tural research, engineering and HCI. Typical to all
amines smart cities in a more critical light. When ex- of these disciplines is that their research problems

are not usually set within a one singular disciplinary and experiences as a User node parallel to Technol-
framework. In addition, the research problem and ogy node. The pilots are designed applying user-
knowledge production operates within a context of oriented and participatory design methods, such as
practice. In the research process, the framework of scenario working, workshops with users, and cultural
methods is being developed and modified through- probes. Networking with relevant stakeholders as
out the process to respond to the needs of research well as communication are part of understanding the
and to the clarified research questions. The research users' needs and anticipated users' experiences of pi-
follows a mixed-method or combined strategies ap- loted lighting and services. Evaluation of users' expe-
proach, where multiple methods from diverse tra- riences of each pilot is a central part of our process, as
ditions are applied in a single research endeavour. it provides feedback applicable in further design iter-
This approach is suitable for researching complex ations and has value as such from research perspec-
phenomena, as in this case intelligent lighting so- tives.
lutions and digital services in smart city contexts. A deep understanding of users' needs helps
Each method brings with its particular strengths and companies develop smart technology to serve cities'
weaknesses and "combining methods provides ap- aims about better urban living and sustainable ser-
propriate checks against the weak points in each, vice and maintenance processes. Boosting economy
while simultaneously enabling the benefits to com- in the pilot regions by providing new types of busi-
plement each other." (Groat and Wang 2013) ness opportunities based on data is one of the fo-
cus areas in piloting, thus forming the third node,
SENCITY PLATFORM: GENERAL FAME- i.e. Business, in the framework. The methods for
this are connecting developers with their customers,
WORK FOR USER EXPERIENCES, INNOVA-
supporting development projects and providing test
TIONS AND BUSINESS and piloting opportunities for new products and ser-
In the cityscape, lighting is always utilized along
vices, arranging innovation competitions, and gener-
routes, where people move, and within areas, where
ating data for data-based business opportunities.
people spend time and act. With intelligent lighting
In the framework, the SenCity group with its
this will mean that, in the cities, a relatively evenly
partner cities, companies, and research units, con-
distributed network of electricity supply as well as
nect three focus nodes with interactive collaboration,
communication exists near people. These networks
creating new links between technology, users, and
can be employed for integrating different kinds of
business. The project collaborates with other relating
sensors which gain and analyze data in real-time for
research projects and negotiates with relevant stake-
smart city services. This forms the starting point for
holders, for example, city and business organizations
the SenCity project and the technological core of the
and communities. Contextual factors of each pilot
platform concept (see Figure 2). The Technology
areas, such as environmental conditions, regulations
node of the framework consists of both the hard-
and social and cultural patterns, set requirements for
ware and software, which form the smart infrastruc-
pilots as well as offer possibilities.
ture of the cities, enabling lighting control, gather-
ing sensor and other type of data, communication as
well as data analysis and fusion. This infrastructure PILOTING IN SMART CITY CONTEXTS
is partly lighting control related, but can serve other In the SenCity project, each pilot project has a fo-
purposes, too. cus on a different theme or application context. The
Smart technology offers endless potential. How- themes and contexts include: presence-based bi-
ever, it is useless if it does not meet users' needs. cycle route lighting, traffic safety of pedestrians in
Thus, in the framework, we have situated users' needs crossings and housing areas, data harvesting and ap-

Figure 2
General framework
of the SenCity
platform concept.
plication, educational outdoor environment, as well Case 1: Smart applications for safe traffic
as guiding and informative route lighting. The pilots environment
vary in complexity level concerning both design and This case presentation (see Figure 3) combines fea-
research targets. In this section, three pilot contexts tures of three pilot projects, which will be conducted
are scrutinized through the aforementioned frame- in different kinds of traffic environments in three
work. The themes are Traffic environment and safety, cities. From a users' experience perspective, the fo-
Big data in road environments, and Smart lighting cus in all the pilots is enhancing traffic safety and en-
and services for educational outdoor environment. abling safe and comfortable moving by foot or by

Figure 3
Framework of
SenCity platform
application with the
Case 1: Safety in
traffic
environments.
bike. At the moment, energy savings are sought in on bicycle routes? The aim is to design and test
outdoor lighting by intelligent control and presence- optimal lighting behavior which saves substantial
sensing and this should not affect safety or sense of amounts of energy without lessening cycling safety
security negatively. or the sense of security of the route users during the
A pilot in Helsinki will approach intelligent light- dark. Two well-designed lighting behaviors will be
ing on the simplest level, setting a singular research tested: the one will be perceptible by route users
or development target: What kind of a detailed light- and the other imperceptible. Feedback of the real
ing behavior is suitable for presence-based lighting world pilot will be collected from cyclists and walk-

ers as well as from people living near-by with inter- and emergency vehicles. Buses communicate their
views and questionnaires. Technically, the pilot will location through GPS and their routes can be fol-
develop and test the ability to detect direction of the lowed in screens and by mobile devices. [6] Be-
movement. Otherwise, already existing lighting con- sides that, buses can be attached with different kind
trol products and technology will be applied. of sensors, thus, creating a mobile sensoring net-
Another pilot, in the city of Salo, will be carried work system of real-time environmental data in dif-
out on a collector road of a housing area. In this ferent districts of the city [7]. Beyond the transport
case, presence data will not be used to adjust illumi- examples, also environmental information, such as
nation for every road user individually. However, the open weather data, is already available [8]. In our pi-
presence data will be averaged over a certain time- lots, combining this information to intelligent light-
slot and used to dynamically control illumination in- ing control is studied. In practice this requires collect-
stead of a fixed dimming schedule currently applied ing and combining available data, and transforming
in Finnish cities. Similar to previous example, feed- it as a lighting control message. In Finland, roads can
back will be collected from the area inhabitants and be very slippery at winter time causing significant risk
the technical implementation will be based on com- of accidents for both the vehicles and pedestrians.
mercially available components. So, an interesting case example is to combine slipper-
Due to several recent accidents, the issue of traf- iness data from the bus and/or weather service and
fic safety in pedestrian crossings is unfortunately very to utilize lighting infrastructure to deliver warning to
topical in Finland at the moment. Alerting lighting road users. The research combines two aspects: tech-
behavior could be used to enhance safety in cross- nical aspect of collecting and delivering the data to
roads or in housing areas with lots of kids playing in the correct location in the city, and the user aspect
the neighbourhood. This requires that the lighting how to communicate the information with accept-
system recognizes the presence of pedestrians and able and understandable way.
even profiles the risk group - young children. In the Today, the typical approach is to monitor in
pilot implementation in Raahe, we will study differ- road environments, for example, traffic density, with
ent scenarios of intelligent lighting behavior in such rather expensive and extensive computing demand-
context. Even though in our research, the focus is ing camera or radar based solutions, which are in-
in real world testing, in this case, virtual modelling stalled only in critical locations of the road network.
could be applied for safety reasons in researching In our project, the research objective is to study the
users' experiences of different kind of alerting light- ubiquitous sensing as an alternative to these kinds
ing behavior. Traffic environment has many bright of solutions. Sensors will be integrated with the lu-
objects, illuminated elements, and signal lights al- minaires and use the lighting control system as their
ready presently. Thus, it is important to study the communication network. The approach is to use sim-
visual load and legibility of this kind of a new infor- ple, low cost sensors and wide geographical cover-
mative lighting layer. In real world implementation, age over the pilot installation. In practice, this means
the research will focus on technical aspect of how to that a sensing device will locate at every lamp post,
identify different road users, vehicles, and pedestri- and the data that a single sensor generates is sim-
ans for example. ple enough to be delivered through the lighting con-
trol system's communication method. As a result, the
Case 2: Big data in road environment installations will generate local data in the road en-
There is already lots of information available about vironment to be transferred to the cloud for further
Finnish cities. For example, traffic lights in city cen- analysis and fusion with other information available.
ters are known to adapt based on traffic conditions The research approach is not to pre-define specific

events or circumstances, such as accident, snow, or ments and school yards can become active learning
a traffic density, to be detected. Instead, the pilot in- environments during school days and leisure time.
stallations are designed to sense road environment In the piloting (see Figure 5), understanding the
with versatility, and the generated data is analyzed to users' needs has had a major role, and during the
study which kinds of events can be identified. The pi- winter 2015-2016, a participatory design process has
lot implementations will follow this approach in sev- been carried out. The phases of the process have
eral locations, for example, in Tampere and Salo, and been cultural probe tasks (Gaver et al. 1999) for both
the research serves to communicate the benefits and personnel and pupils, interviews, scenario working
opportunities of constantly increasing IoT solutions (Rosson and Carrol 2002) and an on-line question-
to different stakeholders in road lighting community naire. The cultural probes and interviews provided
with concrete examples. us with teachers' and pupils' analysis of the current
Besides generating data for the research, the situation and activities in the yard and its lighting
ubiquitous sensing will be used to deliver energy conditions and development ideas. The scenarios
savings to the cities and new services to the end were based on that material, as well as on state-of-
users. The presence data will be used to implement the-art research. They describe with short stories
presence-based lighting as described above. Also, it and comic strips different ways to implement smart
is anticipated that collected data will be useful for city outdoor lighting and digital services as part of the
operations, for example, for road maintenance. As a every-day life of the school. In the scenarios, intel-
part of our research, we will carry out a survey in city ligent lighting applications had several motivations
organizations in order to discover needs and identify in school yards. Besides energy savings, lighting can
most potential utilizers of the smart city data that is orchestrate children's actions in the school yard, for
available with the ubiquitous IoT approach. Finding example, inform pupils of right routes, game areas,
potential use cases and needs will help to generate passing of time, and attract them to physical exer-
new data-based business. The framework of this pi- cise. Lighting connected with digital services can
lot case concerning big data in road environments is extend educational environment from classroom to
presented in Figure 4. the yard and digitally augment the familiar outdoor
space adding new educational contents there. Chil-
Case 3: Smart lighting and services for edu- dren and teachers can interact with light as well as
cational outdoor environment play and create with coloured light. In addition, the
In the city of Oulu, our pilot will concern the outdoor scenarios covered also ideas of digital services related
area for Educational, leisure and culture centre in the not to lighting but to, for example, maintenance and
Metsokangas area. The community centre consists of space management, enhancing security and virtual
several buildings containing mainly educational fa- education. Finally, with the help of on-line question-
cilities for primary and secondary schools and will naire, we got feedback about the scenarios from par-
eventually host 1500 pupils with all its extensions. ents, teachers, and pupils, which will helps us to de-
Other services, which are located in the buildings, are velop further the designs for piloting. The participa-
kindergarten and communal and sports facilities. tion process will still continue. After realization of the
The Finnish National Board of Education has con- pilot in the summer 2017, we will evaluate the results
firmed the new core curriculum for basic education with qualitative methods to gain feedback from ex-
and it will be introduced in schools in August 2016. periences.
The new core curriculum emphasizes the joy of learn- Technologically, the pilot implementation re-
ing and the pupils´ active role [9]. Smart city solutions quires producing intelligently controlled general
can bring sensor data to digital learning environ- lighting and effect lighting using RGBW spotlights.

Figure 4
Framework of
SenCity platform
Case 2: Big data in
road environments.
Also, there is an interest on testing the pilot imple- the pilot design.
mentation infrastructure as an installation site for fast To generate business and interest on smart ser-
data communication and computing for future digi- vices development beyond our own project activi-
tal services, for example, a virtual reality learning aid, ties, we will arrange an innovation competition for
which will have many users at the same time. Other Oulu area ICT start-ups. This is a method to boost lo-
technological research aspects, such as user recog- cal business activity with long tradition in ICT field.
nition, presence monitoring, and needed IoT cloud The competition will be arranged in collaboration
solutions, will be clarified during the continuation of with another project 6Aika which aims at develop-

Figure 5
Framework of
SenCity platform
Case 3: Educational
outdoor
environment.
ing an open innovation platform for smart city oper- DISCUSSION

ations and business. The same approach was taken In general, the SenCity framework of a three node
also in Tampere and Lahti, where our project has al- platform has so far proved to be useful and effective
ready organized similar events with local actors with in creating collaboration between partners in order
good results. to develop intelligent lighting solutions and smart
city services and we will expect interesting results
from the pilots. The project is challenging, as it has
many participating cities and companies with vary-

ing interests. Our method in addressing this com- those service concepts and business actors which re-
plexity - the plethora of diverse starting points and in- ally would find the users, develop scalable and suc-
terests - has been choosing different focus themes for cessful business, and pay cities back for building and
the pilot areas, which will complement each other in maintaining the infrastructure - sensor networks and
the end. This living lab environment, the testing plat- fast communication - necessary for the services.
form which combines the results of each pilot case, Reflecting the three case studies presented it can
will eventually serve all participating cities and com- be noted that even though there is always interaction
panies and build a coherent picture of a smart city. between all the nodes of the framework, there can be
Answering to the former critique against smart seen different main drivers for the pilot development
city concept (Kitchin 2014, Hollands 2008), we have (see the dark grey arrows in Figures 3, 4, and 5). In
emphasized users' role in the piloting processes and the case 1, the clear initiator is found in users' needs,
in the platform framework. Understanding the users' which in the selected pilot contexts, obviously are re-
needs and experiences forms one of the three nodes lated to safety issues. The framing of the research
in the piloting processes, interacting with technology question as well as methods has been kept relatively
and business, which are the other two focus areas. simple. The piloting in those sites aims at testing and
With participatory design methods, we have brought finding suitable and reliable solutions to support ex-
the users into the design processes as our partners. periences of safety in traffic environments, and the
Metsokangas school pilot in Oulu is a good example technological research and development clearly fol-
of that kind of process. The community has actively low that driver. Then, the case 2, which concerns
participated in all the phases and is waiting the re- big data in road environments, can be considered as
alization with great expectations. The process has technology-driven. Technologically, the complexity
provided us with many good ideas as well as criti- level is in that case higher, and that will eventually
cal viewpoints to be taken care of in further develop- lead to a bunch of user and business related research
ment, concerning, for example, privacy issues. Both questions. However, in that case the main method
industry and city partners are keen on understanding is to start with implementing a wide range of sen-
users' experiences and finding ways to serve users' sor and data fusion technology to see what new ap-
needs. plications and use cases will emerge from analyses
From our processes so far, we can indicate that of the complex data. The process will be supported
the business aspect in relation to genuine user needs with stakeholder interviews to find potential applica-
is quite crucial in developing smart city solutions. It tion areas. Thinking about user aspects in the case 3,
has partly been challenging to find motivations for the context of the case - an educational and commu-
investing in smart infrastructure, which goes beyond nity centre - could almost be compared in complexity
already used lighting control technologies, due to level to a small scale "smart village" with its function-
typically higher purchase costs and lack of practical alities and user groups. Thus, here the main driver for
use cases of smart data-based services. Changing to development was chosen to be the understanding of
LED lighting with time-related dimming or presence- the variety of users' needs whereas technology and
sensing, is already saving substantial amount of en- business development goals are subordinate to that
ergy and costs. These functionalities, which are eas- analysis.
ily justified with energy savings, do not in itself need Although the business aspect may seem to lack
great capacities of local data communication and attention in defining the drivers in our research, it is
computing, which many more developed IoT solu- well understood that in the end the success in busi-
tions and smart city applications demand. In our pi- ness, or in economy in general, is a relevant point of
lot contexts, the critical point seems to be finding view. After the finalized pilot implementations and

evaluation processes, we will be able to analyse the [2] http://amsterdamsmartcity.com/projects/detail/i
results and reflect the success of chosen drivers for pi- d/62/slug/flexible-street-lighting
loting. This will presumably complement the general [3] https://www.alliander.com/en/our-activities/sta
rt-ups/open-smart-grid-platform
framework with new clarifying, context-related inter-
[4] http://www.digitalmanifesto.telefonica.com/mani
action factors between the three nodes: technology, festo/big_picture/santander-project/
user and business. [5] http://www.governing.com/topics/urban/gov-santa
nder-spain-smart-city.html
[6] http://developer.reittiopas.fi/pages/en/home.ph
ACKNOWLEDGEMENTS p
SenCity project is funded by TEKES The Finnish Fund- [7] http://www.vtt.fi/sites/livinglabbus/
ing Agency for Innovation and by participating cities, [8] http://en.ilmatieteenlaitos.fi/open-data-manual
companies and research institutions (see Figure 1). [9] http://www.minedu.fi/OPM/Verkkouutiset/2015/03/
curricula.html?lang=en
REFERENCES
Albino, V, Berardi, U and Dancelico, RM 2015, 'Smart
cities: definitions, dimensions, and performance',
Journal of Urban Technology, 22(1), pp. 3-21
Ali, ZH, Ali, HA and Badawy, MM 2015, 'Internet of Things
(IoT): Definitions, Challenges and Recent Research
Directions', International Journal of Computer Appli-
cations, 128(1), pp. 37-47
Gaver, W, Dunne, T and Pacenti, E 1999, 'Cultural probes',
Interactions, 6(1), pp. 21-29
Guo, L 2008, Intelligent road lighting control sys-
tems—Experiences, measurements, and lighting con-
trol strategies, Ph.D. Thesis, Report 51. Helsinki Uni-
versity of Technology, Department of Electronics,
Lighting Unit, Espoo, Finland.
Hollands, RG 2008, 'Will the real smart city please stand
up? Intelligent, progressive or entrepreneurial?',
City, 12(3), pp. 303-320
Kitchin, R 2015, 'Making sense of smart cities: address-
ing present shortcomings', Cambridge Journal of Re-
gions, Economy and Society, 8(1), pp. 131-136
Pihlajaniemi, H 2016, Designing and experiencing adap-
tive lighting. Case studies with adaptation, interaction
and participation, Ph.D. Thesis, University of Oulu,
Acta Universitatis Ouluensis, H3 Architectonica
Rosson, MB and Carrol, JM 2002, 'Scenario-Based De-
sign', in Jacko, J and Sears, A (eds) 2002, The
Human-Computer Interaction Handbook: Fundamen-
tals, Evolving Technologies and Emerging Applica-
tions, Lawrence Erlbaum Associates, pp. 1032-1050
Frost & Sullivan, Inc. 2015, Analysis of the Global Sensors
in Infrastructure Monitoring and Smart Buildings Mar-
ket, Frost & Sullivan
[1] http://amsterdamsmartcity.com/projects/detail/i
d/93/slug/smart-light

Adaptive Designs with Distributed Intelligent Systems
Building Design Applications
Andrzej Zarzycki1
1
New Jersey Institute of Technology, USA
1
andrzej.zarzycki@njit.edu
This paper discusses and demonstrates an integration of embedded electronic

systems utilizing distributed sensors and localized actuators to increase the
adaptability and environmental performance of a building envelope. It reviews
state-of-the-art technologies utilized in other fields that could be adopted into
smart building designs. The case studies discussed here, sensors are embedded in
construction assemblies provide a greater resolution of gathered data with a finer
degree of actuation. These case studies adopt the Internet of Things (IoT)
framework based on machine-to-machine (M2M) communication protocols as a
potential solution for embedded building systems. stract here by clicking this
paragraph.
Keywords: Adaptable Designs, Arduino Microcontrollers, ESP8266, Smart

Buildings, Internet of Things
INTRODUCTION work. Elements like shutters and sun louvers continu-

The concept of an adaptive building envelope as a ously and in real time adjust to external environmen-
performance enhancer has been in use for a long tal conditions in the attempt to optimize indoor cli-
time. Façade elements such as sunscreens, win- mate based on preprogrammed algorithms.
dow shutters, or removable window sashes helped With the introduction of building environmen-
to adapt to climatic and seasonal changes by aug- tal control systems, the automation started to con-
menting building use and its performance. While trol temperature, air quality, and lighting levels.
these elements were dynamic in nature-opening and These automated controls not only provided in-
closing shutters or louvers--architects usually refer to creased comfort for living but also translated into
them as passive environmental techniques for man- more energy-efficient and environmentally friendly
aging building performance because of the manual buildings (Guillemin 2001). However, automated
nature of their operations. The inclusion of mechani- controls were often structured around a central con-
cal and electrical systems within a building, and more trol dashboard system with a limited number of sens-
recently embedded electronic intelligence that con- ing points delivering the same solution to a broad
nects automated control systems with networked number of space conditions. A more deliberate and
data sets and environmental sensing, changed this fine-tuned approach to building controls that goes
passive approach to an active and dynamic frame- beyond the size of a single room into the scale of in-

dividual building components is needed. tal system equipment. While these systems are ef-
While buildings and cities are increasingly filled fective and deliver significant cost savings (˜20%) as
with smart objects and devices that monitor traffic, compared to non-BAS/BMS buildings [1] , they in-
track inventories, or respond to occupants, building clude various levels of autonomy and intelligence. In
assemblies themselves do not demonstrate a similar many cases, they respond to matrix-oriented algo-
level of interactivity and autonomy (Achten, 2015). rithms without understanding the real-time consid-
There is an opportunity to extend these smart and in- erations of building occupants or building assembly
terconnected device networks into the very matter of conditions. Furthermore, BASs/BMSs are usually im-
buildings and their assemblies. plemented in non-residential buildings where a sin-
This paper discusses and demonstrates an inte- gle owner or interested party is in control of central-
gration of embedded electronic systems utilizing dis- ized building systems.
tributed sensors and localized actuators to increase While building automation is an example of the
the adaptability and environmental performance of smart environment approach and is often referred
a building envelope. It reviews state-of-the-art tech- to as the framework behind intelligent buildings, it
nologies utilized in other fields that could be adopted currently limits itself to controlling already mecha-
into smart building designs. The case studies dis- nized and electric/electronic devices such as heating
cussed here, sensors embedded in construction as- and cooling systems, without necessary broader im-
semblies, provide a greater resolution of gathered plementation of embedding sensors and actuators
data with a finer degree of actuation. These case into building components and assemblies. It is par-
studies adopt the Internet of Things (IoT) framework tially because BASs/BMSs are developed by compa-
based on machine-to-machine (M2M) communica- nies that manufacture building system components
tion protocols as a potential solution for embedded and their controls (HVAC or air handling units), not by
building systems. construction companies or building component fab-
ricators. They facilitate an improved performance of
FROM AUTOMATED TO INTELLIGENT NET- installed equipment, not necessarily of the building
itself. What is needed in the next wave of transfor-
WORKS
mation of the building industry and buildings them-
Building automation is a centralized and automatic
selves (Osman 2005), and is advocated in this paper,
control of lighting, heating, and cooling, as well as
is to develop technologies that integrate and take ad-
other systems including security, fire, and occupant
vantage of the embedded systems within building
safety, through a building automation system (BAS)
assemblies. Windows, doors, floors, and wall panels
or building management system (BMS). The goals
all could and should function as part of the building
of a BAS are improved efficient operation of build-
digital interface, sensing user and environmental in-
ing systems, including reduction in energy use and
puts as well as actuating desired spatial outcomes.
operating costs, as well as increased occupant com-
fort and the life cycle of the building. Recently con-
structed buildings include some sort of BAS/BMS, ADAPTATION OF INTERNET OF THINGS
and many older buildings have been retrofitted with FOR BUILDING ASSEMBLIES
these systems. BASs/BMSs include software and Smart and connected devices are objects embedded
hardware architecture that integrates controls for all with microcontrollers, sensors, and actuators, with
or most building systems within one dashboard (in- connectivity that allows data exchanges between the
terface). They are offered by many established com- product and its environment, user, manufacturer,
panies, such as Siemens, Honeywell, and Cisco, that and other products and systems. They allow for en-
already manufacture various building environmen- hanced interactions with people and other objects,

utilizing radio-frequency identification (RFID) tags or form basic building automation, such as traditional
wireless networks. Connectivity enables certain ca- sensor-based (or even mechanical) thermostats or
pabilities of the product to exist outside its physical automated louver systems, to sophisticated and pre-
form as part of a larger data set (cloud). Collected dictive devices such as the Nest thermostat utilizing
data can be analyzed to inform decision-making and machine learning algorithms.
enable operational efficiencies of the product and Independently of their level of autonomy and
the entire system. As envisioned by Mark Weiser in "smartness," smart objects commonly exhibit the fol-
his Scientific American article titled "The Computer lowing three typologies or design dimensions:
for the Twenty-First Century" (1991), "When almost
• Awareness is a smart object's ability to under-
every object either contains a computer or can have
stand (that is, sense, interpret, and react to)
a tab attached to it, obtaining information will be
events and human activities occurring in the
trivial." BAS/BMS platforms discussed earlier are of-
physical world.
ten beneficiaries of these collected data points-tabs
• Representation refers to a smart object's ap-
attached to physical objects. The IoT is the network
plication and programming model-in particu-
of smart objects with embedded technologies able
lar, programming abstractions.
to communicate, sense, and interact with the out-
• Interaction denotes the object's ability to con-
side environment (Kortuem et al. 2010). However,
verse with the user in terms of input, output,
this network of smart and interconnected devices
control, and feedback (Kortuem et al. 2010,
provides opportunities for greater interoperability
31).
and resiliency of the entire system, with data com-
ing from and access to individual subcomponents- These three design dimensions are further organized
devices and objects-as compared to BASs/BMSs. into three types:
In addition to sensing and actuation, smart sys-
tems incorporate decision-making abilities, utilizing • Activity-Aware Smart Objects: An activity-
previously gathered data in a predictive or adaptive aware object can record information about
manner that often employs machine learning algo- work activities and its own use.
rithms. In these cases, the "smartness" of the sys- • Policy-Aware Smart Objects: A policy-aware
tem is attributed to autonomous operation meeting object is an activity-aware object that can in-
its performance and user satisfaction expectations. terpret events and activities with respect to
While current BAS/BMS platforms follow an estab- predefined organizational policies.
lished (pre-programmed) set of rules, the expecta- • Process-Aware Smart Objects: Processes play
tion is that the underlying reasoning (algorithm) for a fundamental role in industrial work manage-
smart systems would adapt over time based on envi- ment and operation. A process is a collection
ronmental and user feedback. of related activities or tasks that are ordered
according to their position in time and space
(Kortuem et al. 2010, 32-34).
PROPERTIES OF SMART OBJECTS AND EN-
VIRONMENTS While these typologies are rather general, Das and
While smart objects can function autonomously and Cook (2005) define smart environment with more nu-
perform complex performance optimizations or user ance as having the following features:
tracking, they do not necessarily need to exhibit in-
1. Remote Control of Devices: the ability to con-
telligence in the sense of artificial intelligence (AI).
trol devices remotely or automatically.
As such, the term "smart objects" is a rather inclu-
2. Device Communication: the ability of devices
sive category of object types, from those that per-
to communicate with each other, share data,

and retrieve information from outside sources proach minimizes network bandwidth and device re-
over the Internet or wireless communication source requirements to ensure communication relia-
infrastructure. bility and some degree of assurance of delivery. This
3. Sensory Information Acquisition/Dissemina- approach is particularly effective in the context of
tion: the ability of sensors to share informa- interconnected IoT devices with limited bandwidth
tion and make low-level decisions. and battery.
4. Enhanced Services by Intelligent Devices: in- Currently, neither IoT nor M2M platforms have
cludes context and location awareness. standardized connected device protocols, with many
5. Predictive and Decision-Making Capabilities: systems custom-built to facilitate particular tasks or
full automation and adaptation that rely on utilize certain devices. Once the adoption of IoT and
the machine learning, or acquiring informa- M2M devices becomes more prevalent, manufactur-
tion that allows the software to improve its ers will need to agree on standards for device-to-
performance. device communications. Security and safety are sig-
nificant concerns associated with IoT and M2M com-
munication, since many original systems were not
designed as open Internet-connected networks. [5]
CURRENT IMPLEMENTATIONS OF INTER-
NET OF THINGS Wireless Communication Technologies
M2M Communication Wireless mesh network topology allows for each
M2M communication forms the basis for IoT. It is node to relay data for the entire network. All mesh
routinely used for remote monitoring and controls, nodes participate in the distribution of data, which
including traffic, inventory, and transportation fleet provides a number of benefits over traditional net-
management, and for robotic applications, includ- works, where a small number of wired access points
ing autonomous vehicles. Big Belly trash bins [2], or wireless hotspots connect users and control all
a waste and recycling management system, utilize communication. A distributed nature of mesh net-
M2M communication to message for trash pickup work communication makes them (1) highly adapt-
when the bin is full. Similarly, vending machines can able and expandable, as individual mesh nodes can
track purchases, record customer preferences, and be added to or removed from the network; (2) able
inform a distributor when a particular inventory is to support high data demands, as each node is con-
running low. [3] nected to multiple other nodes and each device par-
MQTT [4] stands for MQ Telemetry Transport and ticipating in the network would function as a node;
is an M2M and IoT connectivity protocol designed (3) reliable and resilient sources of wireless connec-
as an extremely simple and lightweight publish/sub- tivity for a broad range of public safety applications,
scribe messaging transport utilizing a client-broker as an off-line or damaged device would be super-
relationship structure. Individual clients can be both seded by other nodes and communication routes;
publishers and subscribers, providing or acting upon and (4) less expensive than traditional networks, as
provided sensor data. The broker is counterpart to a they require less wiring and can service significantly
MQTT client and is responsible for receiving all mes- larger areas.
sages, processing them, deciding which client may The disadvantages of mesh networks include the
be interested in them, and then sending messages to complexity associated with their management and
subscribed clients. potentially high initial capital investment if not uti-
MQTT is suitable for wireless networks that expe- lizing already existing device infrastructure.
rience varying levels of latency due to occasional low- The implications for embedded environments
bandwidth or unreliable connections. A design ap-

are that a broad distribution of wireless communica- high range and coverage area is offset by extremely
tion devices would not only benefit from mesh net- low data rates as compared to Wi-Fi networks. How-
works but also benefit the network itself by provid- ever, this may be acceptable for IoT applications that
ing a dense grouping of nodes, mitigating one of are not data transfer heavy, such as sporadic (ev-
the main disadvantages of mesh networks. However, ery couple of seconds or minutes) reporting from a
adaption of mesh networks would mean that the cur- weather station or of the temperature of a road sur-
rent implementation of many IoT technologies would face. LoRa radio modules use 868- and 900-MHz ISM
have to be rethought once a critical mass of embed- bands, which make them suitable for global applica-
ded components was to be implemented in build- tions, as these are commonly shared radio frequen-
ings. The relevance of mesh networks to embedded cies. These frequencies have less interference and at-
environments becomes particularly important when tenuation than the highly populated Wi-Fi spectrum
considering larger-urban and regional-scales of de- (2.4 and 5 GHz).
sign. While buildings could effectively function with The main benefit of long-range radios as com-
an access point approach, cities and open spaces pared to WiFi networks is the low number of radio
could be better served with mesh networks. nodes necessary to provide the coverage for large ar-
eas such as cities and rural regions. However, the low
Specific Technology Considerations data transfer rates make LoRa radios more suitable for
The core of the IoT concept is inter-device communi- communicating low volumes of data from individual
cation achieved through wireless connectivity (usu- sensors or actuators.
ally Wi-Fi) and/or high-frequency radio communica-
Figure 1 tions such as LoRa radios. Both technologies could
ESP8266 module be used in the client-server or mesh network con-
with a motor shield. figurations. As such, IoT networks can be struc-
tured hierarchically or with peer-to-peer connectiv-
ity. The benefits of Wi-Fi-based communication in-
cludes the ability to connect to already established
networks, and through these networks to the In-
ternet, as well as communicating with other Wi-Fi-
enabled devices such as smartphones and tablets.
This approach is popular, since it uses personal mo-
bile devices as control dashboards and data inputs.
Wi-Fi-based IoT could function as part of the larger
Internet-connected network, providing remote ac-
cess and data sharing, or could also function as an
independent and isolated network in either client-
server or mesh network configurations. It provides
high data transfer rates as compared to other radio
communication options, but it also comes with a lim- ESP8266: System of Chip Solution
ited coverage area. One of the recent approaches to IoT, particularly in
LoRa [6] radio-based wide area networks pro- the do-it-yourself (DIY) community, involves ESP8266
vide the opportunity for long-distance communica- chip-based architecture. ESP8266 is implemented in
tions suitable for monitoring urban-scale infrastruc- a number of designs and can be integrated with sen-
ture with a range of up to 5 km, and even 10 km. A sor and actuator shields (Figure 1). ESP8266 com-
bines functionalities of the microcontroller and Wi-

Fi communication module (2.4 GHz, 802.11 b/g/n, Node-RED: Interfacing with Internet of
supporting WPA/WPA2) in a simple system-on-a-chip Things
(SoC) design. ESP8266 is a highly integrated SoC so- Node-RED is an event-processing engine and an
lution with 16 general-purpose input/output (GPIO) open-source visual editor (Figure 3) for wiring the
pins, analog-to-digital conversion (10-bit ADC), Inter- IoT developed by IBM. It integrates hardware de-
Integrated Circuit (I²C), Serial Peripheral Interface vices, APIs, and online services in ways easy to inter-
(SPI), I²S interfaces with DMA (sharing pins with face by lowering technical experience requirements
GPIO), UART (on dedicated pins, plus a transmit-only and allowing developers to focus on creating ap-
UART that can be enabled on GPIO2), and pulse- plications rather than on coding. It allows for vari-
width modulation (PWM). It is running at 80 MHz (or ous forms of IoT system management without having
overclocked to 160 MHz). Most recent ESP-12F mod- an in-depth understanding of underlying technolo-
ules are FCC (USA) and CE (EU) approved. For those gies. The Node-RED eco-system uses drag-and-drop
using the Arduino platform, the ESP8266 chip allows "nodes" that represent components of a larger sys-
for a convenient migration path to IoT architecture. tem such as wireless devices, software platforms, and
It is also compatible with NodeMCU [7] open-source Web services that are to be connected. It also offers
IoT platform and the Lua scripting language. various functionalities ranging from simple passing
of data payloads or a simple debug to more involved
MQTT client handling or posting HTTP GET requests. Figure 2
Like many other visual editors, it does not take away ESP8266 module
the need for scripting, but it significantly reduces the with an analog
need for it. multiplexer
The Node-RED implementation uses a Web (74HC4051 Texas
browser, which is usually hosted in a server such as Instruments)
a Raspberry Pi module (MQTT broker). It allows for provides additional
administration of the entire network (MQTT clients) analog inputs to
without the need to reprogram individual controllers connect an array of
with sensors and actuators. The management and sensors.
controls can be accessed remotely from any Web-
capable device.
As a very capable processor and communication

Figure 3
module ESP8266 is an intentionally highly compact
Node-RED: visual
design with only a single analog input. While this
editor with
may be a limitation in some sensor implementa-
interconnected
tions it can be resolved with an additional analog-
nodes allows for
to-digital converter (ADC) or an analog multiplexer
easy management
(Figure 2). However, in most applications there is
of a large number
no need to up-size the original ESP8266 module,
of IoT devices,
since most IoT implementations use compact, single-
services, and
functionality modules that address a singular sensing
programs.
or actuation functionality, such as the monitoring of
temperature and moisture, or an actuation of individ-
ual relays, motors, or lights.

Figure 4
Transparent Touch
Interface project
using a rigid and
flexible base
focused on
embedding smart
systems into the
conventional
building materials.
PLUG-AND-PLAY BUILDING ASSEMBLIES will be an integration of wireless communication

A critical aspect of future embedded assemblies is with a localization protocol, possibly using RFID/NFC
a plug-and-play (PnP) architecture of building com- tags and energy sources or storage. While this may
ponents. What we have come to expect from com- seem like a lot packed into a window or an individual
puters and electronic devices is their ability to in- building component, current material research sup-
terface with other objects without a need for sig- ports future implementations of these designs. For
nificant user involvement or understanding of tech- example, translucent lithium-ion batteries charged
nology. The same should apply to building com- with sunlight, developed by researchers at Stan-
ponents, their ability to get connected and inte- ford University [8,9], could be integrated into glaz-
grated into an overall framework of a smart building or frit panels and provide the necessary energy
ing. A window, or any other building component, to power embedded components. However, glazing
should be able to localize itself and recognize its and other visually present components need to sat-
role within the overall building, as well as its perfor- isfy not only technological but also aesthetic and user
mance and users' expectations toward it. The history experience considerations. Skeletouch, developed
and building knowledge should be passed to, or at by Hiroyuki Kajimoto (2012), is an example of hep-
least accessible to, newly installed components for tic/tactile display that uses transparent glass plate
self-configuration and performance optimization. In with electrodes to provide a tactile feedback to users.
this instance, mass-produced building components While this particular project implements the tactile
would need to be localized and reconciled within display as an overlay to a smartphone's video dis-
their assembly, respond to their physical and spa- play, the same technology could be applied to build-
tial configurations-windows facing south may need ing glazing as part of building user interface. A sim-
to perform differently than those facing north-and ilar investigation was part of the Transparent Touch
understand regional and microclimatic conditions. Interface project (Figure 4) that looked into devel-
Thanks to embedded technologies, generic manu- oping transparent and flexible printed circuit board
factured building components would adapt to local (PCB)-like components that would respond to touch
circumstances and acquire highly specialized proper- using an MPR121 capacitive sensor controller mod-
ties, possibly hard-reconfiguring their original com- ule driven by an I2C interface (Figure 5).
position.
An important part of these embedded systems

provides on-demand functionalities and contextual- Figure 5
izes knowledge. This is evident not only in the Trans- A capacitive sensor
parent Touch Interface and Skeletouch projects men- controller
tioned earlier, utilizing touch-based inputs or tactile integration with a
feedback, but also in many current consumer prod- mirror as a possible
ucts, such as LED faucets and showerheads relating user interface (UI)
water temperature to light color (Figure 6), with em- for IoT applications.
bedded devices providing users with the relevant
feedback.
While smart environments-buildings and cities-
are initially intended to increase performance and Figure 6
efficiencies (Nakama et al. 2015) as well as to en- Temperature
hance user experience, there are many other side sensing LED faucet;
benefits to technologies deployed in smart environ- an example of
ments. Sensing and actuating technologies em- building user
bedded into buildings are compatible and can in- interface.
terface with autonomous mobility agents, not only
providing navigational clues but also facilitating au-
tonomous wheelchair driving. Similarly, the local-
ization techniques that are used to track and inter-
act with building occupants can also be deployed
to assist nonhuman robotic agents during building
MACHINE LEARNING
construction and post-occupancy phases (Schwartz
To optimize user comfort, minimize operational cost,
2015).
and adapt to occupant needs, smart buildings must
rely upon sophisticated tools for intelligence behav-
ior and performance such as learning, anticipation, ESP8266: CASE STUDIES
and decision making. These predictions can only be A number of the technologies discussed above were
based on past behavioral and occupancy patterns implemented and tested in adaptive façade and
and on occupants' interaction with sensing devices building assembly prototypes developed as part of
such as motion detectors, light controls, or video academic and research work at the New Jersey In-
monitors. stitute of Technology (NJIT) (Figure 7). While ini-
Then recorded data can be used to build sophis- tially a number of "smart" prototypes used Arduino
ticated statistical models that can be used in predic- microcontrollers to interface sensors and actuators,
tion algorithms. and for data communication, it quickly became ev-
ident that for some of the projects a more robust
microcontroller eco-system with wireless communi-
BUILDING AS DYNAMIC INTERFACE
cation would be required. A number of projects
While distributed sensor systems provide a dense
adopted the ESP8266 chip for its integrated micro-
network of data input points, the same channels of
controller and WiFi capabilities. Most of the projects
communication can be used to actuate and inter-
used a small number of controllers (3-4) but one
face with building occupants. Embedded systems
of the projects tested twelve concurrently reporting
not only provide localized intelligence in materials
ESP8266 units continuously connected and access-
and objects but, more importantly, can serve as an
ing outside Internet services with a minimal band-
informational and control interface: an interface that

Figure 7
Adaptive façade
and building
assembly
prototypes
developed as part
of academic and
research work to
test embedded
systems. From the
left: Adaptive width impact on the WiFi network. During heavy a highly reliable configuration with only occasional
Facade 1 and 2, data transfers on the same WiFi network (large file communication interactions, which resulted in a cou-
Algae Bio-Facade, downloads and uploads) module connection time- ple of seconds' delay of the actuation time of one of
and Adaptive Media outs would be a common occurrence with no effec- the modules.
Facade projects. tive data transfer between modules and the cloud One of the embedded prototype projects re-
database. While this could be associated with a data quired several analog inputs. To extend a basic
loss, these situations could have been resolved with ESP8266 module a shield was designed with a 12-bit
additional code establishing protocols that would ADC (Figure 8) that allowed for greater precision in
verify received communication. In the context of the
Figure 8 research projects, this was not necessary since any
ESP8266/WeMos D1 unreceived data from environmental sensors (light,
Mini shields temperature) would have been updated within a
developed as part short period of time. An important finding was that
of the Adaptive once the WiFi bandwidth became available again,
Building ESP8266 modules were able to resume the commu-
Componentes at nication with no impact on the functionality or sta-
NJIT to extend and bility of the system. Similar tests with Arduino Uno
augment base and Ethernet Shield showed that connection time-
module outs often resulted in an unresponsive module and
functionalities: (A) the need to reboot a microcontroller.
sensor and servo While client reporting of sensor data may be in-
extension shield, (B) frequent (every couple of minutes) the connection to reading analog inputs 4096 (2ˆ12) as compared to
12-bit ADC shield clients handling various actuators may require much Arduino's 1024 (2ˆ10) values. This is particularly
extends a number faster response time particularly when a human in- useful with very precise measurements or in cases
of analog inputs put is involved. This need is further compounded when one would like to cover a wide range of val-
and increases the considering the pull nature of client communication ues as would be the case in measuring noise levels
range of values to protocol implemented in discussed prototypes-data and mapping them to decibels (dB)-another research
12 bits, and (C) a could not have been pushed into a client, only a project currently underway in the lab.
motion (PIR), client could have pulled it by establishing a connec- Only some of the projects that implemented the
temperature tion. To reduce the time between data pullings, the ESP8266 chipset with WiFi communication used a
(DHT22), and client would have to be continuously or frequently Node-RED implementation with Raspberry Pi. The re-
illumination level connected to the network, which could impact net- mainder deployed a conventional WiFi network con-
[Lux] (TSL2561) work efficiency. Again, this configuration was tested nected to the Internet and HTML/PHP connections
sensor shield. with twelve ESP8266 modules. The results showed to MySQL databases. The majority of the projects
Gerber file and
additional info at
emergentmat-
ter.org. TOWARDS SMARTER CITIES | Concepts and Strategies - Volume 1 - eCAADe 34 | 689
stayed within the HTML/PHP and MySQL platform cessed at www.emergentmatter.org
partially for a legacy reasons and partially to provide
a parallel testing environment to compare ESP8266 REFERENCES
and Arduino Uno with Ethernet Shield implementa- Achten, H 2015 'Closing the Loop for Interactive Archi-
tions. tecture - Internet of Things, Cloud Computing, and
Wearables', Proceedings of eCAADe 2015
Das, S and Cook, D 2005 'Designing Smart Environments:
CONCLUSION A Paradigm Based on Learning and Prediction', in
This paper and the projects discussed therein inves- Pattern Recognition and Machine Intelligence: PReMI
tigate strategies to shift from individual smart object Guillemin, A and Morel, N 2001, 'An innovative lighting
or smart assembly implementations to orchestrate controller integrated in a self-adaptive building con-
larger intelligent systems for architectural applica- trol system', Energy and Buildings, 33(5), p. 477–487
tions. This means, in most cases, leaving behind the Kajimoto, H 2012 'Skeletouch: Transparent Electro-
Tactile Display for Mobile Surfaces', Proceedings SA
familiar territory of Arduino and getting involved in
'12 SIGGRAPH Asia 2012 Emerging Technologies, Arti-
interfacing various building systems and assemblies, cle No. 21, ACM New York, NY, USA
integrating them with databases and mining data Kortuem, G, Kawsar, F, Fitton, D and Sundramoor, V 2010,
with machine learning algorithms. While this may 'Smart Objects as Building Blocks for the Internet of
feel like stepping outside of established notions of Things', IEEE Internet Computing, 14 (1), p. 44–51
architecture, embedded systems provide new tools Nakama, Y, Onishi, Y and Iki, K 2015 'Development
of Building Information Management System with
to redefine human-made environments and design
Data Collecting Functions Based on IoT Technology',
more efficient and resource-considerate buildings. Proceedings of eCAADe 2015
Osman, A 2005, 'Integrating Digital and Building Tech-
ACKNOWLEDGEMENTS nologies: Towards a New Architectural Compos-
ite', International Journal of Architectural Computing,
The following projects are the research contributions
3(2), pp. 181-190
from NJIT students: Schwartz, M, Frias, R, Dolgener Cantu, E, Stoffel Saul,
G and Park, J 2014 'Humanoids ‘Performing’ Man-
1. Transparent Touch Interface project devel- ufacturing', Proceedings of Workshop on Humanoid
oped by Anthony Samaha (Figures 4 and 5) Robots and Creativity at IEEE-RAS, Humanoids 2014
2. Adaptive Façade 1 developed by Jorge Cruz, Weiser, M 1991, 'The Computer for the Twenty-First Cen-
Alan Mera, and Carlos Rodrigues (Figure 7 left) tury', Scientific American, 265 (3), p. 94–104
3. Adaptive Façade 2 developed by Daniel Bel- [1] www.siemens.com/energyefficiency
[2] www.bigbelly.com/solutions/stations/
tran and Patryk Kleba (Figure 7 second from
[3] www.bizjournals.com/bizjournals/how-to/technolo
the left) gy/2015/04/the-internet-of-things-is-transfor
4. Algae Bio-Façade developed by Samantha ming-vending.html
Bard, Mary Lopreiato, and Libertad McLellan [4] www.mqtt.org
(Figure 7 center panel) [5] www.zdnet.com/article/m2m-and-the-internet-of-t
5. Adaptive Media Façade developed by An- hings-how-secure-is-it/
[6] www.lora-alliance.org/
thony Morrello and Anthony Samaha (Figure
[7] www.nodemcu.com/index_en.html
7 right panel) [8] www.nature.com/news/2011/110725/full/news.2011.
6. ESP8266 shields designed by George Hahn 439.html
(Figure 8) supported by a Facutly Seed Grant [9] www.technologyreview.com/s/424802/a-battery-
from NJIT. you-can-see-through/
Additional information about discussed projects in-

cluding a PCB designs and Gerber file with can be ac-

Animating the Static. Case Study of The Project
"Urbanimals"
Enhancing play in the cities through an augmented and interactive
environment
Sebastian Dobiesz1 , Anna Grajper2

1,2
Wroclaw University of Science and Technology, Department of Architecture
1,2
{sebastian.dobiesz|anna.grajper}@gmail.com
This article delineates the process of developing the project "Urbanimals" - an

interactive installation designed and realised in Bristol, UK, in 2015. As the case
study research, it draws attention to the difficulties in designing interactive
structures in urban spaces - from an architects' idea to a construction stage.
There are four areas that are being investigated: (1) Modelling interactions, (2)
Negotiating locations and logistics, (3) Developing hardware and (4) Performing
the on-site observations. The project draws from the idea of Smart City (SC) as
the concept of the urban environment with a certain level of responsiveness
through implementing a technology-driven matter that expands city offer
perceivable, but gentle and not hindering way. It highlights the possible
applications of projection technology and the utilisation of the 3D modelling
software which provides complex tools for creating animations, movements and
interactions with future users. The article gives clues how to design more
engaging interactions and how to deal with implementing them in public realm.
Keywords: Smart Cities, Interactive Architecture, public realm, art installations
INTRODUCTION life through play [2]. In this respect, the project re-
The investigation presented in this paper is a case lates to the notion of "urban acupuncture" (Lerner
study of a development - and a further performance 2014) which involves the actions that, through small-
- of an Urban Interactive Art Installation called "Ur- scale contributions, transform and revive forgotten
banimals" which was realised in Bristol, UK, in 2015 as and neglected places.
the winning proposition for the "Playable City Award As authors of "Urbanimals", we focused on the
(PCA)". The concept of the project and the "PCA" is spaces of transition in cities - spaces that only offer
built upon an idea of implementing small urban in- an opportunity to pass through. We wanted to try
terventions within cities to improve the appearance to transform them into more pleasurable and vibrant
and perception of their deprived areas, as well as places of public interest - into places that stimulate
to engage citizens in active participation in urban creative behaviour. By "spaces of transition" - follow-

ing Gabriel Fuentes - we understand passages, gate- and entertainment. He advocates for intelligent solu-
ways or aisles and a like, which connect the attrac- tions that not only improve the efficient performance
tive and well maintained public spots such as main of the city, but more importantly, improve citizens
roads, squares or parks (Fuentes 2011). These areas themselves. From the variety of research branches,
were chosen as a great opportunity for an urban re- such researchers as Drew Hemment (Hemment 2013)
search which intends to observe the changes in be- indicate that probably the most essential fact about
haviour of street users. Smart City is that it is as intelligent as its planners and
Urban Acupuncture performed in the project architects are, thus implying that without smart com-
and presented in this research paper is rooted in the munities represented by smart inhabitants, the con-
studies of the Smart City (SC) as the specific urban cept of Smart City lacks its essence and purpose.
testing ground for technological experiments (Saun- In this regard, this research paper correlates with
ders and Baeck 2015). In this regard, we used the In- the technology-driven approach in designing urban
teractive Architecture as the SC-oriented tool to re- environment which can increase the positive devel-
design existing urban spaces into responsive and yet opment of the society in the smart - and any other -
creative places. To understand the main objectives city. To achieve this, it uses the findings in the field
of Smart Cities and Interactive Architecture that are of Interactive Architecture as a tool to influence com-
being addressed in this study, both movements are munities through the design with a real-time, cre-
being expounded in the next paragraphs. ative approach. For the authors of this article, Inter-
active Architecture can be used for augmenting the
Smart city Smart City with crucial layer of social well-being and
Smart City (SC) is one of the concepts of the city healthy relations that are expected to construct such
development which was widely deliberated in the emerging form of settlement as SC.
turn of XXI century, both in public and academic dis-
course. It could be defined - following ie. Tim Camp- Interactive Architecture
bell, Anthony Townsend or Antoine Picon (Campbell Interactive Architecture (iA) enriches the idea of the
2012, Townsend 2014, Picon 2015) - as the idea of Smart City not only by making an urban environment
organising the city in a sustainable and intelligent more technologically and functionally developed,
manner and making it ever more efficient through but also by creating surroundings that may evoke
the process of learning by sensing and analysing its peoples' creativity, hence, their smartness. Ideolog-
current states of performance. That process is based ical roots of Interactive Architecture date back to the
on the use of the latest advances in technologies for mid-twentieth century, to the project "Fun Palace"
data collection and for sending data to citizens in redesigned by Cedric Price and to the technology-
sponse. As Serwant Singh [1] points out, this fast and driven interactive art movement represented by such
easy exchange of knowledge improves such aspects artists as Myron Krueger or Roy Ascott. The ma-
of the Smart City functioning, as smart governance, jor characteristics of interactivity they advocated for,
smart energy, smart building, smart mobility, smart was a novel attribute of creating space that puts a
infrastructure, smart technology and smart health- man in a role of not just a passive user but rather
care. Yet that is just one - and very top-down - side a creator and generator of events and actions. In
of understanding the benefits of the smart city. On that sense - as Kaas Oosterhuis (Oosterhuis and Xia
the other hand, Yuan He (He et al. 2014) draws at- 2007) stresses out - "iA is based on the concept of bi-
tention to more citizen-oriented aspects of urban liv- directional communication, which requires two ac-
ing in SC, such as: public safety, telemedicine, home- tive parties". This point of view is implemented in the
care, interpersonal communications, social activities, "Urbanimals" project also through understanding in-

teractivity in the following way "...something is inter- crease their cognitive development. As Xu Lin (Lin et
active when people can participate as agents within al. 2015) suggests, there is a lot of research done on
a representational context. (An agent is 'one who ini- Interaction Design in urban environment, however,
tiates actions)" (Laurel 1993, p.112). Following that this research paper refers specifically to the investiga-
understanding of interactivity, we anticipated that it tion conducted by Matthew Anijo Punnen, presented
gives extraordinary features for architecture to stim- in "Urban Research Machines: Engaging the Mod-
ulate behaviour of the citizens and in a result, to in- ern Urban Citizen through Public Creativity" which fo-
Figure 1
The set of four
Urbanimals.
Pictures depict: a
dolphin (top-left), a
kangaroo
(top-right), a rabbit
(bottom-left) and a
beetle
(bottom-right).

cuses on using public creativity as a tool for urban The project was intended to run for 2 months
engagement. Author presents the case study of the over the autumn in all selected locations in the city
range of interactive installations on top of which he of Bristol. The development of the project took 3
builds a framework that helps in creating more valu- months and involved 8 persons on a regular, daily
able interactions. Nevertheless, he draws attention basis. It used commonly known technology of short
to the fact that there is still "the lack of research pro- throw projectors (actuators) that are able to display
viding guidance and advice on designing for interac- high quality images from the very acute angle. With
tivity in urban contexts" (Anijo Punnen 2015, p. 61). such technology - allowing to reduce the occur-
He stresses out that the framework for designing Ur- rence of interrupting shadows - users were able to
ban Interactive Installation needs more complex ex- closely interact with the surfaces on which the pro-
pertise. Thus, herewith this research, authors are try- jections appeared. This effort implied strengthen-
ing to contribute to this field of Urban Studies, by ing the awareness of the built environment. The "Ur-
deepening the aspects of implementing art installa- banimals" project also exploited such tools as depth-
tions in a strict urban settings, and by depicting prac- cameras (sensors) for mapping human motion. Us-
tical solutions this process requires. age of these sensors and actuators in the project, was
Therefore, the paper tries to answer such re- the main method for producing the final Interactive
search questions as: (1) what are the legal obstacles Installation.
in developing interactive installations in urban con-
text? (2) How the long term presence of the interac- Construction
tive structure on-site affects the number of interac- The basic component of each installation was a wa-
tions? (3) How the complexity of proposed interac- terproof box with all electronic hardware situated in-
tions influences the behaviour of the users? (4) Could side of it (Figure 2). The hart of each unit was an
the Installation be multiplied and implemented in Intel NUC - micro-computer able to run a standard
other locations to achieve similar results? software on Ubuntu operating system. To monitor
the condition of the equipment, such as its idle/stress
MATERIALS AND METHODS temperature, and to control Urbanimals' program,
The project called "Urbanimals" was proposed as a the NUC was connected to an USB Broadband don-
competition entry for the Playable City award 2015 gle which provided a wireless connection with the
in Bristol. It consisted of the eight installations that internet. The main actuator in a set, was the Short
were built and spread across the city. Each of the Throw Projector with the spherical lens sticking out
structure displayed one of the four origami-shaped, of the box which was protected from being damaged
colourful reminiscences of commonly liked animals by a transparent, polymeric dome. The projector had
such as a dolphin, a kangaroo, a rabbit, and a bee- nearly 500 cd of luminosity which allowed the pro-
tle onto a floor or a wall surfaces in the city (Figure jection to be visible only during the late afternoon
1). The project was designed as an experiment in a and evening hours, and mostly in the spots located
public space that was conducted to arise an invisi- far enough from street lamps. The major sensing de-
ble link between citizens and the city. Thus, the pro- vice was the Microsoft Kinect - a depth-camera which
jections were designed and programmed to interact allowed to track and digitise the motion of users. Fur-
with passers-by and with occurring elements of the thermore, the box was also equipped with a cool-
build environment such as stairs, gutters, windows ing systems constructed from the two efficient fans,
and such. It was therefore intended that the long and protected from insects and birds by the fine grain
term provision of the project should increase peo- metal nets. The water resistance of the unit was en-
ples' conscious perception of the urban fabric. sured by building it from the Spider Flight-Case that

Figure 2
The construction
and a sample
position of the main
unit.
was originally sealed with rubber gaskets. There- elements such as lamps, roof gutters or balustrades
for it was protected during the wet autumn months. in each location. In order to achieve the best area
Each of the boxes was attached to the rigid handle for sensing and displaying animations (roughly 3m x
made from perforated, steel C-profiles, mounted at 4m), the vast majority of the boxes were situated at
the sides of the box and fixed to the existing street least 3,5 meters above the ground.

Interactions brought to "life" with a 3D Open Source Software.
The interactions - understood as the language of Initially, the basic "bodies" of a dolphin, a rabbit, a
communication between each animal and architec- kangaroo, and a beetle, was designed with the fo-
ture and between animal and human (Figure 3) - cus on a proper structure of their armature and set
were designed to examine the ability of digital tech- of movements only. The preliminary construction of
nology to encourage people to play and by this, in- the body consisted of bones with proper constraints,
troducing a new quality of the surrounding without such as a degree of rotation in 3D axis, a ratio of re-
complex and time consuming spatial (physical) rear- scale while being moved, or a relation to other bones;
rangements of the existing urban environment. To which allowed the natural movement of each part of
achieve that, the main challenge of the project was their trunk. Such core of each Urbanimal was a base
to set up an intimate connection between partici- for creating movement scenarios and it gave enough
pants and the city structure through the behaviour of flexibility for changing the external appearance of
each Urbanimal. For this purpose the range of inter- the bodies in later stages of the design, without the
actions and animations was produced and tested to necessity of re-working animations.
attain the highest smoothness in their performance. Subsequently each of the creatures - with a spec-
The amount of predefined situations in which Ur- ified armature and basic postures - was exported to
banimal could enter into an interplay with the user, the Unity to setup the model for interactions be-
reached tens and even hundreds per one creature. tween each animal, a person, and architecture. In-
They were all created in "Blender" - an open source stantaneously, all Blender models were being ad-
software which allowed to build a three dimensional vanced to build a database of all required movement
body of our creatures, and to draft and record the scenarios, as well as to acquire the final shapes of
scenarios of movements. This software gave the abil- Urbanimals. The basic types of movements were:
ity to manage all knotty models with simplified bone running, jumping, sniffing, turning or crouching,
armatures, which ultimately made the process more but each Urbanimal also consisted of characteristic
efficient. The high complexity of rough animations movements like: a dolphin - clapping flippers, splash-
was tackled, merged and multiplied with the "Unity ing water etc., a rabbit - scratching the ear or sniff-
3D" gaming software which provided fluent transi- ing around etc., a kangaroo - making a boxing mill,
tions between produced assets, and thus created a jumping on the tail etc., a beetle - flipping on the
smoothly animated environment in which the inter- back etc. These motions were the essential attributes
actions with participant could occur. of their behaviour. All these behaviours were taken
from nature and prepared as singular actions which
RESULTS AND DISCUSSION were able to be merged and blended into sequences
The whole process of the development of the instal- in Unity, in order to achieve the naturalness of the
lation was simultaneously conducted in three stages designed 3D models. Additionally, every database
within which the cooperation between various spe- of movements included a range of pre-assumed sce-
cialists was crucial. The production was composed of narios in certain situation that might occur while in-
the following actions: teracting with Urbanimal. Vast majority of such sit-
uations consisted of the "eye-contact" animations.
• Modelling interactions For instance: the rabbit starring at the approach-
• Negotiating locations and logistics ing person and jumping back while the person gets
• Developing hardware too close; the kangaroo jumping on the tail to catch
somebody's attention and making angry pose if no-
Modelling interactions. The whole concept of Ur- one comes closer; or the dolphin performing dou-
banimals was based on four animated animals

Figure 3
Passer-by entering
into interaction
with one of
Urbanimals.
ble backflip above the one who get close enough. for an extended site-recognition in order to select
In summary there were almost 800 such animations the place with characteristics formulated by the au-
developed for Urbanimals of which rabbit and Kan- thors of the installations. These characteristics were
garoo consisted of more than 250, dolphin of nearly among all: proper lighting conditions, access to the
150, and beetle of less than a 100. electricity or rich composition of the architectural el-
Simultaneously, a range of varying "bodies" for ements that the Urbanimal could play with. More-
Urbanimals was developed, aiming for the best qual- over, every location needed to be consulted with
ity of movement, but also maintaining the simplic- local authorities in order to obtain required agree-
ity of origami shaped creatures. The final outcome ments. This stage of the development was very de-
was a low-poly model which was using mainly tri- manding because: (1) preselected places had to meet
angular surfaces and which was corresponding with a strict constraints regarding their mono-functional
the paper style models most accurately. Both, move- and transitional character, in which no other activ-
ments of Urbanimals and their appearance (including ity could be normally performed and (2) they needed
a colour) was widely consulted with graphic design- to be a property of the public body to attain proper
ers and interaction designers to achieve the best re- permissions for installing the units. In this respect,
sults and most appealing and fluent outcome. Thus, it also involved a long term official correspondence.
the hole process of designing interactions required While granted with approval for specific location, the
dozens of on-site testing. In result, every Urbanimal insurance company needed to complete the secu-
gained its own personality that was easy to perceive rity checks on-site and recommended an external rig-
by interacting person. ging company to proceed with the assembling of the
Negotiating locations and logistics. The founda- units. It was due to securing the locations accord-
tional idea of the "Playable City Award" is to engage ingly to the local law.
citizens from all over the city into a play. Therefore, Designing hardware. This stage of development re-
the locations for Urbanimals were spread across ev- quired a to conduct a set of tests of devices antici-
ery district in the city of Bristol. Each area was visited pated to be used in the final unit. At the beginning

there were many try-outs of cardboard, plywood and order to obtain permissions for installing Ur-
metal boxes - as the main body of the unit- before banimals. Unfortunately that process took
the regular flight-case was chosen as the most solid more time than expected which caused the
and cost-efficient option. Also few alternative pro- intensification of work short before the launch
jectors were borrowed to check their performance date. There were also many additional is-
in a daylight and the quality of images they pro- sues with the ownership of the street ele-
duce on the rough surfaces of the streets. Likewise, ments - such as lamp hanging on the wall or
many experiments with various cooling systems for balustrades and other - which often resulted
the boxes and all equipment inside were conducted in cancellation of chosen location due to the
to assure no overheating after a longer period of op- rights obstacles. Local authorities had some
erating. Within this stage, every devices was properly reservations to some locations also regarding
adjusted in terms of: (1) Kinects - the range of sens- the smooth flow of pedestrians that might be
ing, (2) projectors - colour, brightness, angle of im- interrupted while installation is turned on and
age, and focus setup, (3) fans - speed of rotation, air- entertain passers-by.
flow direction, (4) power strip - length of the cable. • Designing hardware - The flight-boxes used as
a main protection for all the electronic devices
Observations within the unit worked perfectly well and pro-
Every stage of the design process was carefully ob- tected the equipment from the weather con-
served and recorded in order to collect all relevant in- ditions. They were also very easy in perfo-
formation for the evaluation process. The following rating them to allow the sensors and actua-
aspects of the development were analysed accord- tors to reach the area of interactions. Multi-
ingly: ple perforation did not weaken the stiffness
of the boxes, thus they securely held heavy
• Modelling Interactions - The process of devel- projectors and other equipment, and solidly
oping the "Urbanimals" project was initially stood in one position once attached to the
started with main designers working on 3D rigging structure on site. The projectors cho-
assets in Poland, and with the rest of the sen due to their high luminosity were lack-
team operating in UK. After the first month, ing of remotely controlled focus of the image.
the authors moved to Bristol to strengthen They were fitted only with manual knob which
the cooperation with other specialists. Forth- could not be adjusted while inside the box. To
coming two months were performed as the ensure a sharp image, each projector needed
team collaboration within one office space to be adjusted to pre calculated distance from
which highly improved the communication the projection surface, before it was installed
within the team and effectiveness of per- inside the unit.
formed tasks. • On-site performance - It was observed that
• Negotiating locations and logistics - Finding complexity of the designed interactions and
proper location was a case of finding minor a huge amount of animations implemented
semiotic differences in the appearance of the within the Urbanimals' program, did not
site. Therefore it required presence of main cause an increase in the amount of interac-
designers to fully participate in that process. tions with people participating in the ven-
After the selection of several spots across the ture. In fact, those Urbanimals who acted
city, which were suitable for installing Urban- more calmly engaged more users. The ac-
imals, the producer and production assistants tions performed by passers-by were in such
started negotiations with local authorities in

situations, often very long and with positive erty ownership and lack of required media on site
attitude. On the other hand, the more com- might disrupt the project even at the final stage of
plicated were the movements of Urbanimals its development.
the less active were people in interacting with The research proved that complexity of inter-
them. All cases of people interacting with Ur- actions proposed in the "Urbanimals" project, had
banimals were recorded as gif files and stored an impact on the interactions initiated be passers-
on the server for further analysing the par- by. Less elaborated movements of animated animals
ticipants' behaviour. The camera recorded turned out to more appealing for participants. Long
GIF images both (1) when "Urbanimals" were periods of stagnancy in fact triggered the curiosity of
displayed and (2) when they were switched pedestrians. While on the other hand, those diversi-
off (Figure 4). The dataset obtained after fied movements seem as a closed spectacle with not
the 2 months of the installations operating in too much room for other participants.
the city, showed a significant change in the Within the two months of displaying "Ur-
behaviour of the passers while "Urbanimals" banimals" no significant reduction of interactions
were operating. Moreover, throughout the recorded by the system, was observed. Thus, the
research time-span, the number of observed project proved that a long term exposition does not
interactions did not change significantly, and affect the amount of interactions performed in the
remain on the similar level till the very end of selected areas. Yet we must admit, that those mea-
the project. sures might be different depending on the time of
the year, the city and location within the city, and the
type of exhibited project too.
Figure 4
The process of producing each installation
Recordings from
showed that with just a portion of effort to prepare
the Kinect camera,
the first installation the other seven were easily man-
depicting
ufactured and, with slight changes, implemented
behaviour of
into varying sites. Thus, the methods of develop-
pedestrian when
ing "Urbanimals" might represent a good practice in
the installation is
dealing with activating large number of neglected,
turned off (left) and
mono-functional and unattractive areas. Neverthe-
when it is turned on
less, the results obtained from various locations are
(right).
not comparable due to their different spatial affor-
dances. Thus, it was impossible to identify whether
the same installation has the same impact on the
CONCLUSIONS behaviour of passers-by in each spot.
Through the case study on the development of the Project "Urbanimals" was a joint collaboration of
Interactive Installation in urban context, we wanted designers, architects, urban planners, programmers,
to draw attention of prospective designers to the creative technologists and theatre producers; and as
range of obstacles that they might face in their prac- such elaborated initiative, it resulted in highly com-
tice. Through an in-depth elaboration of this pro- plex digital environment for composing interactions
cess we pointed out that most crucial for effective in urban environment. It was not only a great at-
workflow is the constant communication with local traction for participants but we hope it become a
authorities to secure the best locations for potential valuable case study for prospective designers dealing
projects. We stressed out that planning the endeav- with interactive projects in public realm.
our in advance is a fundamental case because prop-

Acknowledgment / Credits REFERENCES
Designer: LAX laboratory for architectural experi- Campbell, T 2012, Beyond Smart Cities, Routledge
ments (arch. Anna Grajper, arch. Sebastian Dobiesz) Fuentes, G 2011, 'Between History and Modernity :
For Playable City: Searching for Lo Cubano in Modern Cuban Architec-
ture', in Riobó, C (eds) 2011, Cuban Intersections of
Producer: Hilary O'Shaughnessy
Literary and Urban Spaces, SUNY Press, New York, pp.
Executive Producer: Clare Reddington 65-81
International Director: Mark Leaver He, Y, Stojmenovic, I, Liu, Y and Gu, Y 2014, 'Smart City',
Playable City Judging Panel 2015: Claire Doherty, International Journal of Distributed Sensor Networks,
Richard Roberts,Miguel Sicart, Tim Brooke 2014, pp. 1-2
Production team: Hemment, D 2013, Smart Citizen, FutureEverything,
Manchester
Creative Technologist: David Haylock
Laurel, B 1993, Computers as Theater, Addison-Wesley
Developer: Phill Tew Publishing Company, New York
Unity Developer: Lukas Roper, Opposable Lerner, J 2014, Urban Acupuncture, Island Press, Washing-
Games ton
Interaction Assistant: Andrew Cook Lin, X, Hu, J and Rauterberg, M 2015 'Review on Inter-
Production Support: Richard Harbert action Design for Social Context in Public Spaces',
Cross-Cultural Design Methods, Practice and Impact,
Product Design: Molly Price
Los Angeles, pp. 328-338
Watershed Beards of IT Wisdom: Oliver Oosterhuis, K 2007, iA # Interactive Architecture, Episode
Humpage, Alex Heal and Stewart Morgan Publishers, Rotterdam
Rigging Technician: Utopium (CPB Trans LX) Picon, A 2015, Smart Cities : A Spatialised Intelligence,
Production Coordination: Rachael Burton, Akiko John Wiley & Sons, Chichester
Takeshita, Hazel Grian Anijo Punnen, M 2015, 'Urban Research Machines: En-
gaging the Modern Urban Citizen through Public
Intern (Production Assistant): Steven Hartill
Creativity', in Araya, D (eds) 2015, Smart Cities as
Intern (Communications and Archive): Tanya Democratic Ecologies, Palgrave Macmillan, London,
Charteris Black pp. 43-63
Communications: Catrin John, Louise Gardner, Saunders, T and Baeck, P 2015, Rethinking Smart Cities
Claire Stewart From The Ground Up, Nesta, London
Public Relations: Ladbury PR Townsend, AM 2014, Smart cities : big data, civic hackers,
and the quest for a new utopia, W.W. Norton & Com-
Evaluation: Pippa Jones
pany, New York
Trailer: Benjamin Croce and Dominic Gaskell [1] www.forbes.com/smart-cities-a-1-5-trillion-mark
Graphic Designer: Smith and Milton et-opportunity/
Photographer: Paul Blakemore [2] http://www.watershed.co.uk/playablecity/overvie
Special thanks: Verity Mcintosh, Laura Walder, w
Ted Buffery, Helen Bream, Bristol City Council, Wa-
tershed, SteveNightingale at Bloq Management Ser-
vices, Richard Powell, Billy Lambert and Kate Barnes
at the Royal Marriott Hotel, Rikki Temi and staff at
Clifton Down Shopping Centre, Tom Metcalfe, Mor-
risons in Hartcliffe, Georgette Van Hoof
Website: www.urbanimals.eu

Index of Authors
A Cannaerts, Corneel 1-397
Abdelmohsen, Sherif 1-529 Carl, Timo 1-591
Abramovic, Vasilija 1-661 Cerovsek, Tomo 1-485
Achten, Henri 1-63, 1-613, 1-661 Cerrone, Damiano 2-473
Agkathidis, Asterios 1-291 Chelkoff, Grégoire 2-275
Ahmadi, Shahab 1-213 Chen, Kian Wee 2-583
Akleman, Ergun 1-69, 1-335 Chen, Nai Chun 2-447
Al-Qattan, Emad 2-621 Cheng, Chi-Li 1-133
Alacam, Sema 1-253 Chiuini, Michele 2-511
Allani, Najla 2-653 Cojocaru, Vasile Danut 2-409
Almeida, Ana de 2-349 Coraglia, Ugo Maria 2-503
Altin, Burcak Ozludil 2-485 Cursi, Stefano 2-503
Androutsopoulou, Eirini 2-71
Asanowicz, Aleksander 1-195 D
Aschwanden, Gideon 2-457 Das, Subhajit 2-117, 2-217
Aydemir, Zeynep 1-387 Day, Colin 2-217
Ayres, Phil 1-413 Decker, Martina 1-603
Aşut, Serdar 1-321 Dengg, Ernst 2-573
Dewberry, Michael 2-217
B Dobiesz, Sebastian 1-691
Baharlou, Ehsan 1-549 Dokonal, Wolfgang 2-573
Bajsanski, Ivana 1-185 Doumpioti, Christina 1-521
Balaban, Ozgun 2-89 Dounas, Theodoros 1-229, 1-241
Baquero, Pablo 1-73 Duarte, José Pinto 1-501, 2-189, 2-365
Baranovskaya, Yuliya 1-571 Dörstelmann, Moritz 1-571
Barczik, Günter 2-109
Bardyn, Jean-Luc 2-275 E
Beirão, José 2-305 Eisenbach, Philipp 1-93
Beirão, José Nuno 2-31 ElGhazi, Yomna Saad 2-399
Białkowski, Sebastian 1-283, 2-255 Eloy, Sara 2-349
Bingöl, Cemal Koray 2-41 Elsen, Catherine 2-159
Blonder, Arielle 1-581 Elshafei, Ahmed 1-529
Borhani, Alireza 1-69, 1-335, 1-639 Erdine, Elif 1-163
Bouattour, Mohamed 2-653 Erioli, Alessio 1-219, 1-435
Buchanan, Alexandrina 2-493 Evers, Henrik Leander 2-643, 2-673
Budig, Michael 1-83
F
C Ferreira,
Caetano, Inês 1-405, 1-465 Maria da Piedade 1-501
Authors - eCAADe 34 | 701

Ferreira, Victor 2-365 Holland, Alexander 2-555
Fillwalk, John 2-511 Holzheu, Stefanie 2-209
Fioravanti, Antonio 2-315, 2-503 Hou, June-Hao 1-133
Flores, Hector 2-19 Humppi, Harri 2-601
Hysa, Desantila 1-357
G
Galanter, Philip 2-621 I
Gamal, Noha 2-275 Ilčík, Martin 2-417
Garcia, Manuel Jimenez 2-265
Gemme, Frederic 2-591 J
Gerber, David 1-455 Jabi, Wassim 2-147
Giannopoulou, Effimia 1-73 Jancart, Sylvie 2-159
Gomaa, Mohamed 2-147 Janssen, Patrick 2-59, 2-583
Goodship, Paul 2-79 Jovanovic, Marko 1-185
Grajper, Anna 1-691 Juntunen, Eveliina 1-669
Grobman, Yasha Jacob 1-581 Juntunen, Johan 1-669
Grohmann, Manfred 1-93 Jupp, Julie 2-95
Gruber, Anna 1-179
Guerritore, Camilla 2-189 K
Gül, Leman Figen 1-493 Kalantar, Negar 1-69, 1-335, 1-639
Kallegias, Alexandros 1-163, 1-475
H Katodrytis, George 1-329
Hahn, George 1-603 Kepczynska-Walczak,
Halici, Suheyla Muge 1-253 Anetta 1-59, 1-299
Halin, Gilles 2-653 Khabazi, Zubin 1-83
Halıcı, Süheyla Müge 1-493 Khalili-Araghi, Salman 2-633
Ham, Jeremy 1-377 Khallaf, Mohamed 2-95
Hansen, Ellen Kathrine 1-275 Kieferle, Joachim 1-229
Hanzl, Malgorzata 2-49 Kieferle, Joachim B. 1-329, 2-565
Harding, John 1-423 Kim, Jaehwan 1-53
Haring, Mario 1-179 Kim, Simon 1-631
Harris, Libertad M. 1-603 Klein, Bernhard 2-437
Harvey, Lawrence 1-377 Klerk, Rui de 2-305
Hauck, Anthony 2-217 Knight, Michael 2-573
Haymaker, John 2-117 Koenig, Reinhard 2-249
Heinrich, Mary Katherine 1-413 Kolarevic, Branko 1-25, 2-633
Henriques, Konatzii, Panagiota 1-169
Goncalo Castro 1-511 Kontovourkis, Odysseas 1-169
Henriques, Kos, Jose 2-31
Gonçalo Castro 1-537 Kotnik, Toni 1-43, 1-561, 2-391
Herneoja, Aulikki 1-63, 2-179 Kouchaki, Mohammad 1-213
702 | eCAADe 34 - Authors

Kovács, Ádám Tamás 1-269 Menges, Achim 1-549, 1-571
Kretzer, Andreas 1-501 Michalatos, Panagiotis 1-445
Kreutzberg, Anette 2-547 Min, Aram 2-529
Kruse, Rolf 2-109 Mitov, Dejan 1-185
Kulcke, Matthias 2-359 Mohamed, Basem Eid 2-591
Kwiecinski, Krystian 2-349 Mohanty, Akshata 2-59, 2-583
Kępczyńska-Walczak, Mohite, Ashish 1-561
Anetta 1-283 Moleta, Tane 2-137
Monesi, Roberto 1-435
L Montas, Nelson 1-73
Langenhan, Christoph 2-199 Montenegro, Nuno 2-31
Langley, Phillip 2-663 Mullins, Michael Finbarr 1-275
Larson, Kent 2-447
Lee, Ji-Hyun 2-379 N
Lee, Jihyun 2-529 Nagakura, Takehiko 2-447, 2-521
Lee, Jongwook 2-529 Nan, Cristina 1-345
Lee, Sang 2-209, 2-297 Narangerel, Amartuvshin 2-379
Leitão, António 1-405, 1-465 Noel, Vernelle A. A. 1-311
Li, Ruize 2-59 Nourbakhsh, Mehdi 2-117
Lima, Fernando 2-31 Nováková, Kateřina 1-613
Lo, Tian Tian 2-137
Lorenz, Wolfgang E. 1-205, 2-229, 2-359 O
Luusua, Anna 1-669 Ochmann, Sebastian 2-643
Luís, Romão 2-339 Okuda, Shinya 1-113
Olascoaga,
M Carlos Sandoval 2-19
Mahdavinejad, Oliveira, Rui 1-123
Mohammadjavad 1-213 Ozer, Derya Gulec 2-521
Mahmoud,
Ayman Hassaan Ahmed 2-399 P
Marchal, Théo 2-275 Pak, Burak 1-387
Markkanen, Piia 2-179 Panagoulia, Eleanna 1-621
Martens, Bob 1-485 Pantazis, Evangelos 1-455
Martins, Pedro Filipe 1-153 Papadopoulou, Athina 1-367
Mascio, Danilo Di 1-59, 2-511 Paraizo, Rodrigo Cury 1-537
Massoud, Passaint 1-529 Passaro, Andrés Martin 1-537
Matcha, Heike 1-237 Pasternak, Agata 2-371
Mateus, Luis 2-365 Patterson, Charlie 1-345
Meagher, Mark 2-663 Pattichi, Eleni 1-475
Meijer, Winfried 1-321 Pauwels, Pieter 2-511
Melioranski, Martin 2-473 Payne, Andrew 1-445

Pedreschi, Remo 1-345 Stouffs, Rudi 2-59, 2-329, 2-379
Peterson, John Robert 2-493 Stricker, Didier 1-501
Petzold, Frank 2-127, 2-199 Svidt, Kjeld 2-541
Pihlajaniemi, Henrika 1-669 Szoboszlai, Mihály 1-269
Pirhosseinloo, Hengameh 2-275 Sørensen, Jesper Bendix 2-541
Plotnikov, Boris 2-127
Poinet, Paul 1-549 T
Prado, Marshall 1-571 Taborda, Bruno 2-349
Prohasky, Daniel 1-377 Takagi, Naoya 2-463
Puusepp, Renee 2-473 Takizawa, Atsushi 2-169, 2-463
Tamke, Martin 2-643, 2-673
R Tan, Elvira 2-59
Raducanu, Doina 2-409 Tarkka-Salin, Mirva 1-669
Raducanu, Vlad Andrei 2-409 Tauscher, Helga 2-611
Rejepava, Alina 1-179 Tepavcevic, Bojan 1-185
Remy, Nicolas 2-275 Tessmann, Oliver 1-93
Retsin, Gilles 1-143 Thompson, Emine Mine 1-651
Roman, Miro 2-287 Thompson, Ulysee 2-485
Roudavski, Stanislav 2-555 Thomsen, Mette Ramsgaard 1-33
Rumpf, Moritz 1-93 Thurow, Torsten 2-199
Tian, Lo Tian 1-77
S Tomarchio, Ludovica 2-437
Santos, Filipe 2-349 Toulkeridou, Varvara 2-217
Sara, Garcia 2-339 Toyama, Takumi 1-501
Schenkenberger, Benjamin 1-501 Trento, Armando 2-315
Scherer, Raimar J. 2-611 Treyer, Lukas 1-261
Schleicher, Simon 1-621 Triantafyllidis,
Schnabel, Marc Aurel 1-77, 1-377, 2-137 Georgios 1-275
Schubert, Gerhard 2-127 Tuncer, Bige 2-89, 2-437
Schwartz, Mathew 1-53 Turunen, Heidi 1-103
Schwinn, Tobias 1-549
Siala, Aida 2-653 V
Simeone, Davide 2-503 Varela,
Sousa, Jose Pedro 1-123 Pedro de Azambuja 2-427
Sousa, José Pedro 1-153, 2-427 Varinlioglu, Guzden 1-253
Spaeth, A. Benjamin 1-229 Varoudis, Tasos 2-249
Spaeth, Benjamin 1-241 Vella, Irina Miodragovic 2-391
Sprecher, Aaron 2-591 Verbeke, Johan 1-63
Stals, Adeline 2-159 Verniz, Debora 2-365
Stepper, Frank 1-591 Vock, Richard 2-643
Stojakovic, Vesna 1-185

W
Webb, Nicholas 1-59, 2-493
Weber, Markus 1-501
Wendell, Augustus 2-485
Wessel, Raoul 2-643
Wimmer, Michael 2-417
Wit, Andrew 1-631
Woessner, Uwe 2-565
Wortmann, Thomas 2-239
Wurzer, Gabriel 1-205, 2-229
X
Xu, Wenfei 2-19
Y
Yan, Wei 2-621
You, Linlin 2-437
Z
Zali, Parastoo 1-213
Zarzycki, Andrzej 1-53, 1-681
Zhenyi, Lin 1-113
Zolfagharian, Samaneh 2-117
Zupancic, Tadeja 1-63
Zwierzycki, Mateusz 2-643, 2-673
Zünd, Daniel 1-261
Ä
Äikäs, Topi 1-93
Ç
Çolakoğlu, Birgül 2-41
Ö
Österlund, Toni 2-601
Özkar, Mine 1-357


eCAADe2016 Volume1 Lowres

Uploaded by

Copyright:

Available Formats

eCAADe2016 Volume1 Lowres

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

eCAADe2016 Volume1 Lowres

Uploaded by

Copyright:

Available Formats

eCAADe 2016

Complexity & Simplicity

1st Edition, August 2016

Cover Design: Piia Markkanen and Toni Österlund

Complexity & Simplicity

24.-26. August 2016

Alfred Kordelin Foundation

Complexity & Simplicity

Theme - Volume 1 - eCAADe 34 | 5

eCAADe 2016 Acting Conference Chairs

6 | eCAADe 34 - Theme - Volume 1

Acknowledgements - Volume 1 - eCAADe 34 | 7

8 | eCAADe 34 - Acknowledgements - Volume 1

eCAADe 2016 Acting Conference Chairs

Acknowledgements - Volume 1 - eCAADe 34 | 9

List of Reviewers - Volume 1 - eCAADe 34 | 11

12 | eCAADe 34 - List of Reviewers - Volume 1

List of Reviewers - Volume 1 - eCAADe 34 | 13

81 FABRICATION | Design & Application

Contents - Volume 1 - eCAADe 34 | 15

121 FABRICATION | Robotics: Design & Assembling

161 FABRICATION | Robotics: Generative Design

193 CAAD EDUCATION | Design Concepts & Strategies

16 | eCAADe 34 - Contents - Volume 1

251 CAAD EDUCATION | Design Tools

309 CAAD EDUCATION | Fabrication

Contents - Volume 1 - eCAADe 34 | 17

395 GENERATIVE DESIGN | Design Concepts & Strategies

433 GENERATIVE DESIGN | Applications

18 | eCAADe 34 - Contents - Volume 1

509 SMART AND RESPONSIVE DESIGN

559 MATERIAL STUDIES | Concepts

601 MATERIAL STUDIES | Applications

Contents - Volume 1 - eCAADe 34 | 19

649 TOWARDS SMARTER CITIES | Concepts and Strategies

20 | eCAADe 34 - Contents - Volume 1

Mette Ramsgaard Thomsen

Keynote Speakers - Volume 1 - eCAADe 34 | 23

24 | eCAADe 34 - Keynote Speakers - Volume 1

Keywords: Simplexity, Complicity

INTRODUCTION ily complicated designs and tectonic details; how-

Keynotes - Volume 1 - eCAADe 34 | 25

26 | eCAADe 34 - Keynotes - Volume 1

Keynotes - Volume 1 - eCAADe 34 | 27

28 | eCAADe 34 - Keynotes - Volume 1

Keynotes - Volume 1 - eCAADe 34 | 29

to introduce resource economy (time-, material-, and Figure 4

30 | eCAADe 34 - Keynotes - Volume 1

Keynotes - Volume 1 - eCAADe 34 | 31

Mette Ramsgaard Thomsen1

This paper outlines the keynote given at eCAADe 2016, Oulu.

Keywords: Simulation, Adaptive Parametrisation, Open Topologies, Multi Scale

INTRODUCTION engineered materials from the artiﬁcial silk of nylon

Keynotes - Volume 1 - eCAADe 34 | 33

CONSTRAINTS IN CURRENT MODELLING

34 | eCAADe 34 - Keynotes - Volume 1

Keynotes - Volume 1 - eCAADe 34 | 35

36 | eCAADe 34 - Keynotes - Volume 1

Keynotes - Volume 1 - eCAADe 34 | 37