Auto Plan, by Kostas Terzidis

Harvard Graduate School of Design
Cambridge, Massachusetts
AutoPLAN:
a stochastic generator
of architectural plans
from a building program
by Kostas Terzidis
In the world of design, computer programs have taken over

many traditionally human intellectual tasks leaving fewer
tasks for traditional designers. From Photoshop filters to
modeling applications and from simulation programs to
virtual reality animation and even more mundane tasks that
used to need a certain talent to take on such as rendering,
paper cutting, or 3D sculpting the list of tasks diminishes
day by day only to be replaced by their computational
counterparts. What used to be a basis to judge somebody
as a talent or a genius is no more applicable. No longer
are dexterity, adeptness, memorization, fast calculation,
and aptitude sought after in a designers skills set, nor
do they elicit admiration and genius-level praise. The
focus has shifted far away from what it used to be toward
new territories. In the process many take advantage of
the ephemeral awe that the new computational tools
bring to design by using them as means to establish a
new concept or form only to be revealed later that their
power was based on the tool they used and not on their
own intellectual ability. After all, the tool was developed
by somebody else, the programmer who discovered the
tools mechanism, and should, perhaps, be considered
the innovator instead.
As a result of the use and abuse of design tools, many
have started to worry about the direction that design
will take in the coming years. As one-by-one all design
tasks are becoming computational, some regard this as
a danger, misfortune, or an appropriation of what design
should be and others as a liberation, freedom, and power
toward what design should be: i.e. conceptualization.
According to the latter, the designer does not need to worry
anymore about the construction documents, schedules,
databases, modeling, rendering, animation, etc. and can
now concentrate on what is most important: the concept.
But what if that is also replaced? What if one day a new
piece of software appears that allows one to input the
84
building program and then produces valid designs, i.e.

plan, elevation, and sections that work. And, worse, what
if they are better than the designer would have ever done
by himself or herself? (Even though most designers would
never admit publicly that something is better than what
Figure 1: (a) A grid, (b) the site, (c) the spaces, (d) the adjacency matrix,
(e) placing a space, and (f) one possible solution.
2007-08 : formZ Joint Study Journal
they would have designed, yet what if deep inside them

they would admit the opposite). What then? Are we still
going to continue demonizing the computer and seeking
to promote geniuses when they really dont exist?
During the peak of enthusiasm for possibilities that opened
up for computational design in the early 1970s, a series of
innovative projects were set as potential targets. One of
them was the automatic generation of plans from building
programs that was proposed by Dietz (1974). It involved
a unit system, a site, a program, and an adjacency matrix
and then the computer system would produce multiple
solutions by trying various combinations of space allocation
based on the neighborhood rules (see image below). This
possibility apart from clever, innovative, productive, and
effective, it also introduced indirectly a radical view on the
role of the designer and the process of design itself. In
its simplest manifestation it calls for the production of an
architectural plan without human guidance. In its so-called
automatic nature, it negates the very premise upon which
architecture, and design by extension, has established
its existence, identity and authority throughout the ages.
It poses a strange paradox where design is redefined
not as an intentional articulation of form in pursuit of an
objective, but as a random reshuffling of information under
constraining rules until a possibility is met that satisfies
a function. Despite its promising potential, automated
design did not take off as one would perhaps expect.
Instead, computers simply became tools that enhanced
the productivity, efficiency, and presentation of design
that led eventually to enhancing the ego of the designer
instead of challenging it.
Figure 2: A series of possible solutions.
In an attempt to shed light on this missing opportunity,

the author of this paper developed a computer program
called autoPLAN in 2008 that generates architectural
plans out of a building program and a site. The program
was written in the Processing computer language and
can export multiple CAD files, one for each plan that was
then further enhanced using formZ. A series of plans
generated under autoPLAN can be seen in the figures
below. AutoPLAN uses a stochastic search algorithm that
searches for available space to distribute the programs
rooms given the sites boundary and the adjacency
matrix.
The program and its algorithm demonstrate an
alternative approach to the potential of computation as
a design methodology. Is it possible that a design can
be accomplished through the exhaustive search of
possible solutions? Consider the case of all possible
combinations of black or white for nine squares in a 3x3
arrangement. They are 512. If we constrain the choice
to only symmetrical configurations, those are only 32. Or
perhaps all possible combinations of three continuous sets
of three block of three colors in a 3x3 arrangement. Those
are only 60. In that sense, the notion of randomness can
Figure 3: (a) A series of possible plans and (b) a high rise.
85
AutoPLAN: a stochastic generator of architectural plans from a building program
Figure 4: All possible combinations of black or white for nine squares in a 3x3 arrangement (top) and those that are symmetrical (bottom).
86
2007-08 : formZ Joint Study Journal
be seen not as a chaotic disorganizing principle that is

often portrayed, but rather as an ordering mechanism.
Perhaps, the term randomness should be redefined here
to clarify its connection with order. Random presupposes
an exhaustive search of all possible combinations and
therefore can be seen as unexpected sampling. In the
cases shown earlier, such combinations are computable
within a reasonable amount of time. Yet, in other cases,
the combinations are so many that it is not possible to be
computed in a desirable amount of time. In such a case,
randomness functions as a sampling mechanism that
provides possible choices for the designer, occasionally
surprising. Nevertheless, such an ordering device is
based not on a careful premeditated intuitive process but
rather on simple, almost nave, attempts under extreme
repetition. The process, albeit antithetical to that of

traditional design, sets out a new paradigm where design
is laid out, not in the mind of the user, but rather in the
computer program that addresses the issue. The focus of
design is not even in the process itself since that can be
replaced, but rather in the replacement operation itself. In
that realm the new designer constructs the tool that will
enable one to design in an indirect meta-design fashion.
References
1. Dietz, A., Dwelling House Construction, Cambridge:
MIT Press, 1974.
Figure 5: All possible combinations of three continuous sets of three block of three colors in a 3x3 arangement.
Kostas Terzidis is an Associate Professor at the Harvard Graduate School of Design. His current GSD
courses are Kinetic Architecture, Algorithmic Architecture, Digital Media I & II, Cinematic Architecture, and
Design Research Methods. He holds a PhD in Architecture from the University of Michigan (1994), a Masters
of Architecture from The Ohio State University (1989), and a Diploma of Engineering from the Aristotelion
University in Thessaloniki, Greece (1986). He is a registered architect in Europe where he has designed and
built several commercial and residential buildings. His research work focuses on creative experimentation within
the threshold between arts, architecture, and computer science. As a professional computer programmer he
is the author of many computer applications on form-making, morphing, virtual reality, and self-organization.
His most recent work is in the development of theories and techniques for algorithmic architecture. His
book Expressive Form: A Conceptual Approach to Computational Design published by London-based Spon
Press (2003) offers a unique perspective on the use of computation as it relates to aesthetics, specifically in
architecture and design. His book Algorithmic Architecture, (Architectural Press/Elsevier, 2006), provides an
ontological investigation into the terms, concepts, and processes of algorithmic architecture and provides a
theoretical framework for design implementations. His latest book Algorithms for Visual Design by Wiley (2009)
provides students, programmers, and researchers the technical, theoretical, and design means to develop
computer code that will allow them to experiment with design problems.
87
AutoPLAN: a stochastic generator of architectural plans from a building program

Auto Plan, by Kostas Terzidis

Uploaded by

Copyright:

Available Formats

Auto Plan, by Kostas Terzidis

Uploaded by

Document Information

Original Description:

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

Auto Plan, by Kostas Terzidis

Uploaded by

Copyright:

Available Formats

Harvard Graduate School of Design

In the world of design, computer programs have taken over

building program and then produces valid designs, i.e.

2007-08 : formZ Joint Study Journal

they would have designed, yet what if deep inside them

Figure 2: A series of possible solutions.

In an attempt to shed light on this missing opportunity,

Figure 3: (a) A series of possible plans and (b) a high rise.

2007-08 : formZ Joint Study Journal

be seen not as a chaotic disorganizing principle that is

repetition. The process, albeit antithetical to that of

You might also like