Robot Automation Prototypes Service Applications
Robot Automation Prototypes Service Applications
Robot Automation Prototypes Service Applications
1. Introduction
It has been suggested that the robotics and automation research will move more and more
towards the service sector, rather than the more structured factory jobs as in the past
(Engelberger 1989), (Schraft 1994). A number of service robots have been developed
worldwide in recent years, including an increasing number for the healthcare sector.
Examples include the HANDY-1 robotic arm for feeding the severely disabled (Whittaker
1992), (Jackson 1993), Meal Assistance Robot System (Ishii et al. 1991), MANUS wheelchair
mounted robotic arm (Kwee 1998), KARES rehabilitation robotic system (Song et al. 1998),
Care-O-bot mobile home care system (Schraft et al. 1998), THR hip replacement surgical
robotic system (Paul et al. 1992), CERO robot for assisting partly motion-impaired people
to transport light objects in an office environment (Huttenrauch et al. 2002), and
WorkPartner for interactive work with humans outdoors (Ylonen et al. 2002).
While the above examples provide direct services to the end users, the authors in City
University of Hong Kong (CityU) have focused their attention on developing facilities to
support maintenance or auxiliary-aid services such as cleaning, delivery, and inspection in
vast but difficult to reach/manage places. Although these are less glamorous or
prominent chores, they are nevertheless essential routine services often associated with a
substantial tedium and lack of stimulating challenges. Additionally, these services may
also include safety risks, as in the case of washing windows on high-rise buildings,
moving through a maze of dilapidated ventilation ducts, or delivering to, or cleaning, a
hospital ward accommodating infectious or infection-sensitive patients.
Such safety risks or convenience of service provided further emphasise the benefits of
automating these monotonous tasks. CityU robotic systems reviewed in this text have
been arranged approximately in the order of diminishing risk to operators from the
targeted activities. Some originally developed supporting technology has also been
included.
cleaning
head
PLC
Two main rodless cylinders are controlled by two sets of solenoid valves in parallel, one of
type noted for large air-flow capacity and the other for high-speed response. The motion
speed is thus not compromised for precision. The main cylinders are also equipped with
pneumatic locking units for more agile and precise stopping. Two additional short
cylinders are used for the robot rotation (in 1.6 ° increments) to align it with the window
frame. The position sensors on the
main cylinders are rotary encoders
with rack and pinion mounting.
One ultrasonic sensor is located at
each end of both main cylinders.
Their feedback is used to devise the
window-pane cleaning path and
cross to the following pane when
appropriate.
Cleanbot II shown in Figure 2.2
uses one large vacuum gripper and
electrically actuated wheels. It
moves on large flat surfaces
including tile walls with small air-
gaps or unevenness, and can cross Figure 2.2 Climbing robot Figure 2.3 Climbing robot
10mm high obstacles. It is “Cleanbot II” Cleanbot III
534
relatively fast, low cost, and easy to manoeuvre. Cleanbot II provides smooth motion. It is
shown in Figure 2.2 with a wet scrubbing cleaning attachment.
Cleanbot III (Figure 2.3) uses a chain-track to move its 52 legs, each with a suction cup and
passive compliance (CIDAM 2004). The robot provides smooth continuous motion and can
step over window frames and obstacles 35mm high.
Ampli-
fier A/D
PC
Signal
analysis
535
3.1 Theoretical Background
It can be readily shown that the fundamental frequency of flexural resonance of the tile
increases with diminishing size of the void underneath it − for the same tile thickness (Tso
et al., 2000). The impact-generating side of the problem is modeled here by a two-degree-
of-freedom spring-mass system, Figure 3.2. One spring with stiffness Kf represents the tile
deflection, and the other spring with stiffness Kc represents the nonlinear contact stiffness.
The two masses, M2 and M1, represent the tile and the impacting sphere, respectively.
Considering the energy distribution in the system, the original kinetic energy of the sphere
deforms the structure during the impact. Assuming that the structure is elastic, as it
reaches its maximum deformation the velocity of the sphere is zero and all of the initial
kinetic energy has been converted to the energy stored by the deformation of the structure.
Therefore, ignoring the shear and membrane components of structure deformation, the
energy balance equation can be given as:
1
E sum = M 1 v0 2 ≈ E f + E c = E f + E c1 + E c2
2
where v0 is the initial sphere speed, the subscripts f, c refer to the energy stored in the
elastic deformation of the structure and sphere indentation in the contact region (c1
pertains to the sphere and c2 to plate).
It can be shown that the ratio of energy converted into flexural vibration depends on the
thickness and radius of the plate. In the tile-wall structure, the thin tile layer caused by
serious bonding degradation has small thickness and effective stiffness, leading to much
stronger flexural vibration under impact compared to a solid tile-wall. Based on acoustics
theory, the intensity of sound radiation is proportional to the vibration energy. Thus, the
intensity of sound excited by flexural vibration after the impact can be used as an indicator
for the structure-integrity identification for the tile-wall.
According to theoretical analysis for a degraded tile-wall, the thin tile layer formed by a
void separation underneath will lead to the absorption of most of the kinetic energy of the
impacting sphere through the flexural vibration mode of the tile. For a solid tile-wall,
however, the loss of kinetic energy of the sphere is very small.
The strength of free vibrations of the sphere caused by impact indentation is also affected
by the vibration energy factor λ=Ef /E sum. As a result, the relative intensity of sound radiated
from the vibrating sphere and plate can indicate the integrity status of the tiled structure.
Define Rps as the ratio of sound intensities from the sphere and plate. Because the solid tile
wall is generally over 20 times thicker than the thin layer of debonded tiles, the ratio of the
sound intensities from the sphere and plate after impact Rps will appear significantly
536
different in the presence of debonding. Using this impact sound method, the need to use
earlier reported coupling agents or to apply high pressure on tile-walls, can be avoided.
(a) (b)
(c) (d)
Figure 3.3 Impact sound feedback spectrum from a solid tile wall (a),
from a tile wall with the debond size 160mm×114mm (b),
with a debond 120mm×114mm (c), and with a debond 80mm×114mm (d)
Figure 3.4 Theoretical and measured fundamental frequency versus debond size
537
The deviations between the theoretical (based on assumed geometry) and measured
values are caused by many factors. Background noise and microphone distortions are just
some of the disturbance effects. While the system therefore can provide only a rough
estimation of the void size under individual tiles, there is little difficulty in identifying
whether there is a void or a solid bond underneath.
538
4.1 Automatic Floor Cleaning (AFC) Robot
Adapted to public hospitals in Hong Kong where wet mopping and dry dusting are
carried out daily, the cleaning staff must bend to access the areas under the beds (Figure
4.1). Not only does this slow down the cleaning process and is the work physically more
demanding, but it also inconveniences the patients and puts the staff at risk of contracting
diseases during their unavoidably close proximity
to the patients.
Shown in Figure 4.2, the developed AFC robot can
navigate autonomously, move under beds, avoid
obstacles, vacuum dust, and spray disinfectant.
Since it is battery powered, its path planning had to
be carefully designed to avoid cleaning the same
area repeatedly. This task has therefore been
partitioned into the “Strategic” and “Reflex” layers
shown in Figure 4.3.
Figure 4.2 Automatic Floor Cleaning
Implemented on an embedded PC, the Strategic
Robot Prototype
layer is responsible for the path-planning, position
estimation, and all strategic decision-making such
as path correction. Implemented on a 68HC11-
Global path
based “Handy board”, the Reflex layer is planning, Pentium
Strategic layer
PC board
mapping &
responsible for all the low-level control and data control
acquisition of the robot, including the quick actions
to deal with dynamic obstacles and unexpected Obstacle
avoidance & 68HC11 Reflex layer
Handy board
events. Such actions are triggered by the sensor local control
feedback.
For obstacle detection, the robot is equipped with
Pumping &
six IR-based distance-measurement sensors (Sharp- Infrared Servo motors Contact cleaning
sensors switches
devices
GP2D12) on the sides of the robot. Two additional
sensors on top of the robot are for aligning the Figure 4.3 Control System
robot with the bed-edge when entering under beds
using a dedicated algorithm. A light bumper, Figure 4.4, is
also included. It operates two contact switches that can detect
any side of the vehicle subject to impact to avoid conceivable
obstacles.
For the path planning, the cleaning area is divided into 24 Figure 4.4 Bumper
squares illustrated in Figure 4.5. This simplifies the mapping
6 7 18 19
process and data communication between the Strategic and Reflex
layers. The robot starts at the square 1 and stops at square 24. 5 8 17 20
Should there be no obstacles, the robot would follow the path
shown in Figure 4.5 to finish the cleaning task. However, should it 4 9 16 21
539
path is blocked, and “D” when the obstacle is before a corner. In the first instance at this
elementary level, obstacles are assumed to be smaller than the squares in Figure 4.6.
Otherwise, they are treated as multiple obstacles, each one smaller than the square.
Obstacle
“A”“B” “C” “D”
3 10 6 7 6 7 6 7
Robot
Robot 2 11 5 8 5 8 5 8
path
1 12 4 9 4 9 4 9
Beacon Wireless
LAN
access
Wireless LAN point
transceiver
Door RF
transceiver
540
The MARCH robot navigation combines the wall-following strategy, local path planning,
IR beacon-based landmarks, and central monitoring. The destination and the preferred
route are issued from central control room via the wireless LAN. To locate the position of
the robot, beacon-based landmarks are adopted and are normally fixed at corners,
doorways or places where beacon information is needed. Each beacon contains a unique
ID code transmittable to the vehicle through a short-range IR communication device. The
vehicle would be able to identify its position in the hospital map based on these ID codes
and send it back to the centre via the wireless LAN for monitoring purposes.
A three-input (front/left/right proximity sensor) and two-output (speed, direction) fuzzy
logic controller has been designed for collision avoidance. Two schemes of operating lifts
are being investigated. One is based on a Cartesian robot manipulator for operating lift
buttons and requires a simple vision system for the recognition of lift buttons, which slows
its operation but requires no modifications to the lifts. The other scheme relies on the RF
interface with the lift control system, which operates faster but requires prior installation
work on lifts.
5. Baggage-Carrying AGV
The intended purpose of the service AGV (automated guided vehicle) is to carry
passenger luggage at airports, shopping malls, museums or theme parks for people with
special needs. As the host passenger moves along corridors (structured or otherwise) on
foot or in a wheelchair, the AGV loaded with baggage will follow him/her a small
distance behind. Since the human user is within the operation loop, the AGV is not
required to perform highly complex tasks such as comprehensive understanding of the
environment or precise path planning. A low-cost, simple sensor and control system
would suffice for this type of application. Important functions for the AGV are to identify
the direction where the host passenger is, and avoid hard collisions. Within reason,
accurate maintaining of the distance between the host and the AGV is not fundamental.
Hence, the AGV is mainly concerned with two distinct task-achieving behaviours: (i)
passenger following and (ii) collision avoidance.
The maximum signal intensity received during the sensor revolution, Amax, is checked. If it
is greater than a preset threshold, the AGV is then deemed to be too close to the passenger
and is stopped to wait for the passenger to move further.
542
provide both functions simultaneously, resulting in tactile sensing with a range of over 15
cm. This range is two orders of magnitude larger than the range of traditional tactile
sensors. Throughout their 15 cm tactility range, curvature gauge loops generate negligible
reaction forces with the impacting body. (Djordjevich et al. 2000)
The range mentioned is on top of the one provided by the elastic bumper on which the
optical fibres are mounted (Figure 5.3). Configuration of such an elastic bumper is also
monitored by a separate curvature gauge taped along it, making it sensitive to deflection
within its entire range of deflection. Instead a strip, this elastic bumper can be a sheet wide
enough to cover the entire front surface of the AGV. Importantly, such elastic sheet would
be sensitive to its own curvature. (Djordjevich et al. 2000)
Figure 5.3 The elastic bumper and fibre ribbons are sensitive to curvature
When an unexpected obstacle is detected between the passenger and the AGV (points B in
Figures 5.4a and 5.4b), the guidance algorithm is such that the vehicle backs by a
predetermined distance (to point A in Figure 5.4) and attempts to go around the obstacle
Passenger
following
Passenger
direction
Passenger
direction
C
E A C
A E
D Passenger
B D F
following
B
G
(a)(b)
Figure 5.4 Left-hand (a) and Right-hand (b) Obstacle Avoidance Schemes
with the objective of getting back onto the track behind it and continuing further along.
Figure 5.4a shows such a manoeuvre in its left-hand version triggered by the deflection of
sensors on the right half of the AGV. Deflecting other frontal sensors, whether alone or in
a combination with those on the right-hand side of the vehicle, initiates, as the default
case, a manoeuvre illustrated in Figure 5.4b. Depending on the passenger’s location with
543
respect to the AGV principal direction (the sign and magnitude of angle θ), the two
avoidance schemes may be swapped.
A maze of crisscrossing
ventilation ducts is a rather
common sight in modern high rise
commercial and industrial
buildings. Their unsightliness
apart, a major problem with
ventilation ducts is that they
enclose often cool and condensate
laden environment favourable for
the development and spreading of
mildew. Mildew irritants
represent a serious health hazard
for sensitive occupants in the
buildings who are left with little Figure 6.1 Ventilation-duct Inspection Vehicle
choice on how to avoid the
exposure or protect themselves. A tele-steerable tracked vehicle capable of navigating
through common types of ventilation ducts, whether horizontal, inclined, vertical, or with
sharp corners, is shown in Figure 6.1. The main purpose for its development is inspection,
although it could easily be adapted to spray fungicides or multi component sealing or
rust-proofing agents, or to perform duct cleaning and mechanical repair.
Equipped with a camera, the vehicle relays back to its operator outside the visual internal
information about the duct. To allow track adhesion to duct walls while climbing, moving
on the sealing, or floor/wall/sealing transitioning, the vehicle has many powerful rare-
earth magnets incrusted in its tracks. The tracks are made of flexible plastics for firmer
grip on metal sheets. Because the vehicle during the operation is enveloped by the steel
sheet metal of the duct acting like the Faraday cage that hampers wireless communication,
the vehicle drags its umbilical cord that must be coiled back when reversing the vehicle at
the end of the mission.
545
discovered appliances as logical
appliance objects to the Common Presentation
Event Logging
Intelligent
manager control modules
Service Access Layer and the user
Application Layer
applications can access all appliances
via the Common Service Access Layer.
The User Application Layer provides
the client applications, such as Web Common service access layer
BT
Users connection
InfraRed
RF Power
manager plug
LonWork Native
Cgate
UPnP
The system is implemented on the for
CBUS
OSGi
Bundle Device
Home Gateway
implementation of OSGi architecture.
Under the OSGi framework, most of Fig ure 7.2 System Architecture
services of the Home gateway are
implemented as OSGi ‘bundles’. The OSGi framework provides several advantages
important for the Home/SOHO gateway, including:
• platform independence;
• service discovery and dependency resolution;
• dynamic service update; and
• sharing of service.
All applications are currently implemented as OSGi bundles under JES, which allows the
just-in-time service delivery to the end-user.
User-friendliness is important. Since the end users will not normally be knowledgeable
enough to set up a complicated network system for all of the home or office appliances,
the whole system should be plug-and-play requiring only simple set up procedures.
Hence, the Universal Plug and Play (UPnP) standard has been implemented. It does the
features of automatic service discovery, remote access and event notification, which are
essential for developing the plug-and-play capability for networking intelligent ad hoc
devices. Since UPnP uses standard Internet protocols like TCP/IP, HTTP, XML, SOAP etc,
it makes interoperability an inherent feature of the system. For legacy devices, such as IR
appliances, additional UPnP bridging software modules are implemented on top of the
device drivers in order to make the devices UPnP compliant with the UPnP framework.
Wireless communication is important for the Home/SOHO system because it enables
users to add/remove devices from the network dynamically. Bluetooth has certainly
attracted worldwide support and a number of home appliance manufacturers, such as
Sony, intend to add Bluetooth communication to their future products. Hence, Bluetooth
has been adopted as the main wireless communication network. Among the various
profiles supported by Bluetooth, Audio/Video Remote Control Profile (AVRCP) is most
546
suitable for the current applications. In this profile, the controller translates the detected
user action to the A/V control signal, and then transmits it to a remote Bluetooth device.
Two different roles are defined for devices in this profile, a controller (CT) and a target
(TG). The CT, such as a personal computer, a PDA, a mobile phone or a remote controller,
sends a command to a TG, such as a TV, a headphone or a video player/recorder. Upon
receiving a command, the TG responds back to the CT as shown in Figure 7.3.
Bluetooth Connection
Manager
TV (TG)
command
respons
Bluetooth Repeater
Unlike the traditional IR remote controller, this one can support several A/V devices
implemented with AVRCP. The communication between CT and various TGs is
organised through the Bluetooth’s piconet mechanism. In the present system, a Bluetooth
connection manager has been developed and implemented with AVRCP. All the A/V
devices within the coverage area are visible to this manager. It can control all the devices
from the user input and the knowledge captured based on the user habits. Moreover, a
Bluetooth repeater with AVRCP has also been developed to extend the coverage area of
the whole system. This repeater keeps a list of all the A/V devices and sends the list to the
connection manager periodically. The limited coverage area can be expanded by attaching
more Bluetooth repeaters to cover a larger area as shown in Figure 7.3. The advantages of
using this profile are that all the Bluetooth A/V devices have a standardised model, which
makes the devices inherently interoperable, and also future A/V devices with AVRCP
could be operated without major modifications.
The AVRCP communication among all devices and TCs is inherently bidirectional. This
two-way communication network constitutes the fundamental requirement for
development of an intelligent control system. To illustrate this application, an intelligent
control module will be described in next section.
548
a) regular web browsers on PCs using HTML (HyperText Markup Language) –provide a
very common means of controlling devices remotely from anywhere in the world;
b) web browsers on PDA using HTML and web enabled cellular phones using WML
(Wireless Markup Language) – provide mobile solutions for users to control devices
with limited functions;
c) Java client for systems with installed JVM (Java Virtual Machine) – requires that users
pre-install a client application on the client machine (Java native client applications can
generally provide an interface with faster response, and enable that voice technology
be embedded in the Java client applications);
d) voice enabled bundle in server – all devices can be controlled through a microphone
from the server.
Overall, the multi-layer software architecture applied and Java based implementation
provides a flexible platform for developing ad hoc heterogeneous network system for
intelligent devices. The two-way communication feature among all components (TC or
TG) constitutes the infrastructure for developing an intelligent control module. Moreover,
it allows the development of the user-friendly and reliable interface so that the status of all
devices could be monitored by the user even remotely.
Acknowledgements
The MARCH robot, baggage carrying robot, Cleanbot III and safety support vehicle were
financially supported by the City University of Hong Kong grant for CIDAM research unit
(CityU #9360077), an RGC grant #9040407 and an RGC Central Allocation grant (CUHK
3/98C). Cleanbot I, Cleanbot II and WIC robots were funded by RGC grants #9040407 and
#9040711, an ISF grant (AF/76/98) and an ITF grant (ITS/33/01) respectively. The home
automation project was supported by an ITF grant (UIM-55). The telemedicine project
was funded by an ITF grant (ITS/99/02).
The authors would like to express their sincere gratitude to the following persons:
• Prof. Y.T. Zhang and his research team, Mr. D. Lam, Mr. A. Choy and Dr. Y.H. Fung
for their important collaboration and advice on the development of the telemedicine
project;
• Dr. Y.H. Fung, Mr. K.L. Lee, Prof. P. Sooraksa and his research team for their help in
some stages of the MARCH project;
• Prof. G.H. Zong and his research team in Beijing University of Aeronautics and
Astronautics, Dr. W.L. Xu, Dr. Y.H. Fung, Dr. K.L. Fan, Dr. H. Shao, Dr Y.D. Zhang,
Mr. W.L. Chow, Mr. J. Zhu, Mr. X.C. Meng, Dr. D.W. Tang, Dr. X.J. He and Mr. G.
Cheung for their contributions on the Cleanbot I/II/III development;
• Mr. A. Ng, Dr. S.R. Yan and Prof. Y.S. Xu for their contributions and suggestions on the
development of the safety supporting vehicle;
• Mr. A. Kar, Mr. V. Pjevalica and Dr. Y.H. Fung for their help in developing the baggage
carrying robot;
• Dr. Y.H. Fung, Dr. K.L. Fan, Dr. L. Han, Dr. F. Tong and Mr. G. Cheung for his their
contributions on the WIC robot development;
• Mr. D. Lam, Mr. A. Choy and Dr Y.H. Fung for their help and contributions in
developing the home automation project.
• Dr. J. Zhang and Dr. K. Wong for their contributions to the development of the
extended tactility.
549
8. References
CIDAM (2004). Promotional Brochure, City University of Hong Kong.
Djordjevich, A.; & Boskovic, M. (1996). Curvature Gauge, Sensors and Actuators - Physical,
Vol. 51, pp. 193–198, 1996.
Djordjevich, A.; Tso, S.K.; & Zhang J. (2000). Extended range tactility in material handling,
International Journal of Production Research, Vol. 38, No 17, pp. 4357-4367, Nov. 2000.
Engelberger, J.F. (1989). Robotics in Service, Cambridge, MA, MIT Press.
Huttenrauch, H.; & Eklundh, K.S. (2002). Fetch-and-carry with CERO: observations from
a long-term user study with a service robot, Proceedings of 11th IEEE International
Workshop on Robot and Human Interactive Communication, pp. 158-163, Sept 2002.
Ishii S.; Hiramatsu, F.; Tanake, S.; Amari, Y.; & Masuda, I. (1991). A meal assistance robot
system for handicapped people's welfare, Proceedings of Conference on Robots and
Mechatronics, Japan Society of Mechanical Engineers, 1991.
Jackson, R.D. (1993). Robotics and its role in helping disabled people, Engineering Science
and Educational Journal, December 1993.
Kwee, H.H. (1998). Integrated control of MANUS manipulator and wheelchair enhanced
by environmental docking, Robotica, vol.16, no.5, pp 491-498, 1998.
Paul, H.A.; Mittlestdt, B.; Bargar, W.L.; Musits, B.; Taylor, R.H.; Kazanzides, P.; Zuhars J.;
Wlliamson, B.; & Hanson, W. (1992). A surgical robot for total hip replacement
surgery, Proceedings of IEEE International Conference on Robotics and Automation, Nice,
France, vol.1, pp 606-611, May 1992.
Rossing, T.D.; & Fletcher, N.H. (1994). Principles of vibration and sound, Springer-Verlag,
New York, 1994.
Schraft, R.D.; Schaffer, C.; & May, T. (1998). Care-o-bot: The concept of a system for
assisting elderly or disabled persons in home environments”, Proceedings of the 24th
Annual Conference of the IEEE Industrial Electronics Society, vol.4, pp 2476-2481, 1998.
Schraft, R.D. (1994). Mechatronics and robotics for service applications, IEEE Robotics and
Automation Magazine, pp 31- 35, December 1994.
Song, W.K.; Lee, H.Y.; Kim, J.S.; Yoon, Y.S.; & Bien, Z. (1998). KARES: Intelligent
rehabilitation robotic system for the disabled and the elderly, Proceedings of the 20th
Annual International Conference of the IEEE Engineering in Medicine and Biology Society,
vol. 5, pp.2682-2685, 1998.
Tso S.K.; Fung Y.H.; Chow W.L.; Zong G.H.; & Liu R. (2000). Design and implementation
of a glasswall cleaning robot for high-rise buildings, Proceedings of the World
Automation Congress Eighth International Symposium on Robotics with Applications,
Maui, Hawaii, paper ID: ISORA123, June 2000.
Tso, S.K. (a); Luk, S.L.; Choy, W.H.; Liu, K.P.; Chow, C.S.; Leung, K.F.; Lee, G.; Yau, F.;
& Lam, C.W. (2003). An intelligent networking and automation system for home and
SOHO environments, The Fourth International Conference on Control and Automation
(ICCA), Montreal, Canada, 2003.
Tso, S.K. (b); & Tong,F. (2003). Wall Inspection Robot for Improved Maintenance of High
Rise Buildings, 20th International Symposium on Automation and Robotics in Construction
(ISARC2003), Delft, Holland, pp. 449-455, Sept 2003.
Whittaker, M. (1992). HANDY 1 robotic aid to eating: A study in social impact, Proceedings
of RESNA International'92, pp 589-594, June 1992.
Ylonen, S.J.; & Halme, A.J. (2002). WorkPartner - Centaur like service robot, Proceedings of
IEEE/RSJ International Conference on Intelligent Robots and System, Vol. 1, pp. 727-732,
Sept/Oct 2002.
550