UBC Publications

UBC Publications

UBC Publications

Focus 1998

Item Metadata


JSON: focus-1.0115133.json
JSON-LD: focus-1.0115133-ld.json
RDF/XML (Pretty): focus-1.0115133-rdf.xml
RDF/JSON: focus-1.0115133-rdf.json
Turtle: focus-1.0115133-turtle.txt
N-Triples: focus-1.0115133-rdf-ntriples.txt
Original Record: focus-1.0115133-source.json
Full Text

Full Text

Mesh Generation Program Opens
New Fields of Application
Dr. Carl Ollivier-Gooch makes modelling with unstructured meshes
easier with new software he developed for work in fluid dynamics.
What do the human heart and an
aircraft wing have in common?
They can both bear heavy loads and
stress, and present interesting
problems in fluid and solid dynamics which are hard to
experimentally test. And now they
can both be modelled with a
sophisticated new mesh-generation
program developed by CICSR's
Carl Ollivier-Gooch.
Ollivier-Gooch, a specialist in
computational fluid dynamics, is an
assistant professor of Mechanical
Engineering at UBC. He developed
and launched GRUMMP (Generation and Refinement of
Unstructured Mixed-Element
Meshes in Parallel), a program that
generates two- and three-dimensional unstructured meshes.
Modelling Complex Objects
In aerodynamics (one of Ollivier-
Gooch's areas of expertise) and
other areas of fluid and solid
dynamics, computer simulation is a
practical modelling alternative to
expensive experimental testing.
Computer simulations require
engineers and designers to model
complex objects, such as a heart or
wing, by decomposing them into
simpler shapes. This decomposition
produces a mesh.
Innovating with Meshes
GRUMMP'S main innovation is
its use of unstructured meshes—so
called because they decompose
objects into irregularly connected
shapes. Structured meshes, in
contrast, use a uniform grid of
bricks or rectangles to model
"Unstructured meshes will fit
around odd shapes a lot more easily
than square or rectangular meshes,"
says Ollivier-Gooch.
Unstructured meshing also opens
the possibility of using different
continued on page 2
Spring 1998 Vol. 9, No. 1
Better Methods for Debugging 3
Global Memory System: 4
Comparing Power Suppliers 5
Ending Master-Slave Power 6
ProgrammableChips 7
Passing Notes 8 I wish to welcome many new members
to CICSR, but first I'm pleased to
welcome Jim Little as the new associate
director of CICSR. Jim is an associate
professor in the Computer Science
Department, whose work in computational vision and artificial intelligence
was profiled in the Spring 97 issue of
FOCUS. With Jim on board, we will be
able to initiate new CICSR projects.
I am also pleased to announce that
Huaxiong Huang joined us in January
as our industrial facilitator to strengthen
our co-operation with industry and our
contacts with other units on campus.
I'd also like to welcome new CICSR
members Ed Casas and S. Sidney Fels
who joined us in January, and Babak
Hamidzadeh who joined us in July 97.
This issue profiles a number of other
new researchers who joined UBC and
CICSR within the last two years. The
cover story looks at Carl Ollivier-Gooch,
a Mechanical Engineer whose work in
unstructured mesh generation has
attracted some attention.
We also look at Michael Feeley's work
in cluster computing, Hua Jin's research
in power distribution, and Tak
Niimura's interest in comparative power
supplier analysis.
Alan Hu's debugging research and
Steve Wilton's work in Field Programmable Gate Arrays round out this issue.
One last piece of news which is
plainly visible with this issue of FOCUS—we've updated the newsletter's
design to help deliver our message about
current computer systems research here
at UBC. We hope you like it!
Rabab Ward, CICSR Director
Ollivier-Gooch continued from page 1
element types where appropriate—mixing
quadrilaterals and triangles in two dimensions, for example, or prisms and tetrahedra
in three dimensions. More information,
however, is needed to model an object
using the irregular pattern of an unstructured mesh. The identity of neighboring
points is unpredictable and must be stored
explicitly, unlike a structured mesh which
has predictable connections between
"The tradeoff is that an unstructured
mesh is less predictable, but more flexible.
And flexibility is a plus when it comes to
dealing with complex geometries."
The most obvious practical applications
of GRUMMP are in solving problems in
hydrodynamics and aerodynamics. Researchers at the National Research Council
plan to use Ollivier-Gooch's techniques in
Department of National Defense work
with the F-16 jet.
More Accurate Fluid Simulations
Ollivier-Gooch is also working to
improve algorithms for solving fluid
dynamics problems on unstructured
meshes. A recent application was in the Heavy Lift
competition, a North
American undergraduate
engineering contest which
challenges teams to design
a model aircraft that, with
a given engine and
maximum size, can take
off with the heaviest
weight from a 200-foot
runway. The UBC team
came to Ollivier-Gooch
for help with the computations necessary to
redesign an 8.5 foot wing that weighs less
than 2 pounds.
Existing unstructured mesh techniques
generate increasingly accurate results as the
number of mesh cells is increased and the
cell size reduced. Typically, solution error is
reduced by a factor of four when mesh cells
are reduced to half their former size.
Ollivier-Gooch is developing techniques
which can reduce solution error by a factor
of eight or sixteen under the same circumstances, allowing the use of fewer mesh
cells of greater size. The end result may be
more accurate results for the same computational cost in memory and CPU time.
"We don't know whether these new
techniques will be more efficient in solving
a given problem to a given accuracy,
because no one has yet been able to
investigate the question," says Ollivier-
Gooch. "I'm near the point of being able to
make meaningful comparisons of these
Broad Range of Applications
Ollivier-Gooch believes that his techniques can be used to tackle problems in
other areas such as solid mechanics. The
range of GRUMMP's possible applications
was brought home to Ollivier-Gooch when
he made the program available on the Web.
Within a week, 184 people in 25 countries
had downloaded the program.
"I wasn't surprised to hear from people
who wanted to give it a try," says Ollivier-
Gooch. "But what did surprise me was the
range of applications people had in mind
for it, from medical research to
Ultimately, Ollivier-Gooch wants to
develop his software to the point where it
requires little knowledge
of mesh generation to
"As the popularity of
unstructured mesh
techniques grows, we're
seeing a lot more people
using them who aren't
experts in mesh generation. As developers, we
need to accommodate
these people by writing
software that they can use
without having to know all the grubby
details. They want to say 'Here's my
geometry. Give me a mesh that I can trust.'"
"I'm striving to make the process as
automatic as possible."
Until the ultimate, user-friendly, unstructured mesh generation program is
available at Future Shop, you can download
a copy of GRUMMP from Carl Ollivier-
Gooch's page at Mechanical Engineering's
web site. His page can be at found at
Carl Ollivier-Gooch can be reached at (604)
822-1854 and cfog@mech.ubc.ca.
FOCUS Modern Methods for Better Debugging
Dr. Alan Hu is developing debugging software that can work
on a broad range of applications with the push of a button.
A few years ago, chip manufacturer Intel
introduced the long-awaited and much-
hyped Pentium processor, only to find that
it had a bug in it. An American math
professor discovered the bug when he
found errors in his results after running a
series of calculations.
The company's failure to catch the bug,
which was eliminated in a later version, was
a public relations fiasco and cost Intel
hundreds of millions of dollars.
Formal Verification
Alan Hu, an assistant professor of
Computer Science at UBC and a CICSR
member, is developing bug-catching
software to ensure other companies
do not have to repeat Intel's costly
experience. The software uses a
process called formal verification
and allows designers to quickly
check their work-in-progress for bugs.
"The best time to catch a bug is when
the idea is still in your head, and the worst
time to catch it is after you've shipped a
million units to your customers and they're
really angry." Hu says.
Hu's powerful software tools can check a
system's accuracy without actually running
the millions of different calculations the
system might be called on to perform. And
the debugging can be performed at the
push of button during any stage of the
development process.
Hu's methods are an improvement on
previous debugging techniques. Computer
simulations, for example, provide an
accurate test of a system's performance
with specific data, but millions of
simulations would be required to determine if the system works for all possible
"An engineering rule of thumb says that
the later in the design process you catch a
bug, the more it will cost you to fix it."
approach," Hu says. "The combination is a
little less powerful mathematically and
theoretically than some of the previous
work, but the big payoff is that you can
apply it by basically pushing a button."
Broad Application
Because many computer applications are
based on underlying mathematical similarities, Hu's bug-detecting methods can be
used on a broad range of applications
including complex systems such as microprocessors, asynchronous transfer mode
(ATM) switching circuits, or large computer systems consisting of several subsystems.
"The idea of saying the goal is better
debugging, rather than to prove
mathematically that a design is
correct, makes far more sense if
you actually want to achieve
something useful."
Although Hu has not yet reached
his goal of commercializing his software, he
is optimistic that, through his collaboration
data. These simulations have also become
less effective as computer systems have
become more complex. The other approach     with Fujitsu Laboratories of America, HAL
is theoretical, and although it is considered
to be the ultimate proof of a system, it is
time-consuming and laborious.
Streamlined Approach
Hu's approach streamlines and updates
Computer Systems and other experts in the
area of verification, he'll see the demand for
his software tools grow.
Hu and international colleagues will get
together to explore advances in verification
when UBC hosts the International Confer-
these two methods. Using a combination of     ence of Computer-Aided Verification, June
model checking and binary decision 28 to July 2, 1998.
diagrams (BDD), Hu is able to check Hu, the conference co-chair, can be reached at
whether a circuit or system generates the (604) 822-6667 or ajh@cs.ubc.ca.
same result as the mathematical function it
is expected to perform. And, instead of
being forced to run an enormous number
of simulations, a BDD can be used to
represent the set of all similar tests to run,
thereby making it possible to effectively run
more tests while doing relatively less work.
"The idea of combining BDD and model
checking was the watershed event in this
Spring 98 Global Memory System Boosts
Power of Networked Computers
Dr. Michael Feeley's GMS allows workstations on
a network to share RAM and increase productivity
Changes in network technology are
providing opportunities for computer users,
particularly those working in local area
network (LAN) environments, to boost
computing power to near supercomputer
levels at a fraction of the cost.
Dr. Michael Feeley, a professor in UBC's
Computer Science department, is exploring
these opportunities through a global
memory system (GMS) he developed while
completing his doctoral work at the
University of Washington with funding
from the American National Science
Foundation and Digital Equipment Corp.
Sharing Memory
Feeley's system allows workstations or
PCs on a network to share physical
memory (i.e., RAM) to dramatically
increase computing speed and efficiency.
"The recent improvements in processor
and network performance have been
astounding," says Feeley. "In the very near
future, a conventional workstation or PC
network will be able to achieve or even
exceed the performance levels of special-
purpose supercomputers that people are
buying for millions of dollars."
To achieve this goal, Feeley is changing
the way that operating systems manage
memory. In a traditional system, a PC's
memory is managed locally for programs
running on that PC.
Memory Cooperation
Modern networks, however, are so fast
that a PC's "local" memory is valuable to
programs running on every PC in the
network. To harness this power every PC's
operating system must cooperate to manage
their "local" memories globally.
The major advantage of global memory
management, says Feeley, is that the
rapidly accessed physical memory available
is multiplied by the number of
workstations in a cluster.
Improved Performance
A user therefore has automatic access
through the operating system to not only
the memory of his or her own PC, but to
the combined memory of all those within
the network. And, since the retrieval speed
"High-speed super
computing is no longer the
sole domain of researchers
with custom-designed
of data from physical memory is much
faster than retrieval from, the speed of
an application can be greatly improved.
"If you look at the performance
characteristics of next-generation
networks as a whole, and recognize their
potential to operate as a multiprocessor,
one of the first things you have to
consider is the managing of resources,
hardware and abstract, globally," Feeley
Widely used networks are relaying
information at speeds of ioo megabits-per-
second, already io-times faster than when
the popular Ethernet was invented.
Network speeds of greater than a gigabit-
per-second are also within the reach of even
the relatively cheap and readily accessible
commodity networks such as Ethernet.
"High-speed super computing is no
longer the sole domain of researchers with
custom-designed systems," says Feeley.
And although clusters of PCs will never
gain quite the speed of a super computer,
the computing power is there and the
speed is still impressive."
Feeley and colleagues in CICSR, including Alan Wagner and Norm Hutchinson,
are putting GMS through the paces using a
16-node Pentium-based cluster, with a
collective RAM of two gigabytes, connected
by a 1.25-gigabits-per-second Myrinet
network in the CICSR building.
The GMS project is a collaboration with
researchers at UBC, University of Washington, and Duke University.
Michael Feeley can be reached at
feeley@cs.ubc.ca and at (604) 822-8179.
Michael Feeley: "We're still trying to fully understand
the range of possible commercial and research
applications for this type of computing."
W Apples and Oranges:
Fuzzy Logic Draws
Comparisons between
Power Suppliers
Dr.Tak Niimura sees a future of
deregulated power markets and the
need for a measuring stick to
compare competitors.
Electric power is not a standard consumer
commodity. People generally buy their
power from a monopoly utility and that
regulated relationship makes comparison
shopping impossible.
But an international trend toward
deregulation of electric power systems and
the emergence of competitive power
suppliers may change the way we shop for
power. Dr. Tak Niimura has anticipated
that change and the growing needs of
power consumers and suppliers for comparison information.
Niimura is an assistant professor of
Electrical and Computer Engineering at
UBC. He is also a CICSR member who,
with the support of the Japan Institute of
Systems Research, has developed software
which offers a reliable method of ranking
various aspects of a power supplier's
Identify Strengths & Weaknesses
Using mathematical models and computer simulations, Niimura's software allows
power suppliers to identify their strengths
and weaknesses in relation to their competitors: it also provides power consumers
with the knowledge they need to choose a
supplier who best meets their requirements. The comparisons are made and
weighted using a range of factors including
environmental impact, security of power
supply, and voltage consistency.
Competition in the power sector has
brought a new sort of player into the
market, says Niimura. Companies with an
eye to short-term profits are likely to
construct thermal generating stations,
which can be built cheaper and faster than
hydro-electric systems.
While easier to bring into operation,
thermal power stations emit nitrogen
oxides, sulphur oxides and carbon dioxides.
Environmental Comparisons
"If we try to reduce cost by favouring a
cheap supplier, we may end up increasing
environmental emissions," says Niimura.
The capacity that has to be maintained in
transmission lines and transformers to cope
with varying loads and fluctuating voltages
is another variable that is factored into the
Fuzzy Membership Values
The value between zero and one, called a
fuzzy membership value, gives some idea of
how satisfactory a particular factor is,
whether that factor is voltage, cost, security
of supply, or environmental impact.
A value of 0.2 for environmental impact,
for example, suggests that the environmental performance of that supplier is not
particularly strong. On the other hand, the
same supplier might be expected to have a
much better rating of, perhaps, 0.7 for cost.
In this case, the tradeoff between the cost
and environmental impact is evident in the
values of each.
Japanese steel producers are exploring the
possibility of marketing their surplus
"If we try to reduce cost by favouring a cheap supplier, we
may end upincreasing environmental emissions."
"The system I'm developing allows
customers to determine how much power
to purchase from different producers if, for
example, the customer is environmentally
Niimura's technique uses fuzzy sets and
logic to make the factored comparisons.
Fuzzy sets provide a mathematical way of
expressing such vague terms as "satisfactory" and "unsatisfactory."
"By defining such fuzzy sets for security,
environmental impact or even for cost, we
can change everything into the graded
measurement of zero and one," says
power, says Niimura, and some are even
looking at shifting to power generation as a
primary business. And Niimura sees a
growing need in the North American
power systems sector for the sort of
evaluation he can provide.
Niimura hopes to develop his software
into a commercially-available product to
serve the consumer and businesss market.
Tak Niimura can be reached at (604) 822-2405
or at takn@ee.ubc.ca.
Spring 98 Power and the End of Master-Slave
Dr. Hua Jin is using digital signal processors
to develop flexible power supply systems.
A crusader for independence, Hua Jin
wants to put an end to the master-slave
relationship which has dominated the
operation of facilities such as hospitals,
industrial complexes and subway systems.
When Jin—a CICSR member and
assistant professor of Electrical and Computer Engineering at UBC—refers to
master-slave relationships, however, he is
not talking about the interpersonal kind.
He is referring to power supply where
control commands are distributed from a
central control block (the master) to other
dependent units (the slaves). This type of
system is commonly used in hospitals,
industry and other essential services which
require steady and reliable power supplies.
Power Units Function Independently
Jin is collaborating with Statpower
Technologies Corporation of Burnaby,
B.C., to develop an alternate power supply
system comprised of multiple power supply
units that operate in parallel. The advantage of this system is that each power
supply can function independently: if there
is a breakdown somewhere in the system,
the other units can still deliver power.
To free power units from being dependent on the master unit, Jin's aim is to give
each power unit the ability to independently monitor and assess power
requirements, and to adjust output in
response the load.
The key, says Jin, is a digital signal
processor (DSP) based control algorithm
which processes the input information and
controls the power supply. The algorithm is
designed such that the load can be shared
automatically by each power supply.
Jin is using DSP to perform the functions which have traditionally been
implemented using analog circuitry. The
advantage of a DSP over analog circuitry is
the flexibility it provides through software
programmability. In addition, a DSP can
deal with complex algorithms which would
be difficult to implement using analog
"We've used software tools for simulation, analysis and verification of the
algorithm," he says. "The simulations have
gone well and we're now at the stage of
implementing the algorithm."
System Flexibility
Another advantage in using multiple,
independent power supplies is that they
can operate in settings that are physically
remote from one another. In a master-slave
system, the need for communication
between the master and slave units limits
the range, and the communication itself is
a source of noise that can compromise
reliability. System flexibility and
expandability are other key features.
"If the load demand picks up, you have
the option of either buying a bigger unit,
or buying several separate units and
connecting them in parallel. We're trying
to make systems more flexible and reliable
so that if one unit fails, others can pick up
the load."
In addition to his research involving
parallel power supply systems, Jin is also
investigating the use of DSP techniques in
other applications.
Xantrex, a manufacturer of power supply
equipment, is working to incorporate DSP
control techniques in its power supply line
in order to gain an edge on the competition. Xantrex is developing a six-kilowatt
DC power supply which will use DSP
algorithms to achieve functions such as
adaptive controller parameters, power
factor correction, and soft start.
Jin has also written circuit simulation
software specifically for the power electronics industry. His software is available
commercially and is being used by universities and industries in 17 countries.
In addition to funding from Statpower
and Xantrex, Jin also receives funding from
the BC Advanced Systems Institute,
Nortel, and Super Power Technologies Inc.
Hua Jin can be reached at (604) 822-1695 and
FOCUS Programmable Chips Offer
Time-to-Market Advantages
Field Programmable Gate Arrays are flexible, and
Dr. Steve Wilton is working to make them faster.
The faster a company can get a product
to market, the greater the chance of
success. The stakes are even higher in
computer-chip manufacturing where the
design and production of a single dedicated
chip can take many months and cost
hundreds of thousands of dollars.
Companies that want to get a chip to
market quickly, and can sacrifice chip
speed, have an alternative in the
Field Programmable Gate Array
(FPGA), a programmable chip
which is the specialty of CICSR
member Dr. Steve Wilton.
Readily Reconfigured
Unlike their faster dedicated
cousins, FPGAs can be readily
reconfigured to perform different functions
in the event of mistakes or new requirements. Although FPGAs cost more per
chip than mass-produced dedicated chips,
they are proving invaluable to companies
who don't anticipate massive sales volumes
for their particular product.
"For many companies time-to-market is
everything," says Wilton, an assistant
professor in UBC's Department of Electrical and Computer Engineering. "One
design mistake caught after fabrication can,
from a business perspective, be disastrous.
All that money and all that time goes down
the drain, and your competitive edge can
go out the window."
At present, FPGAs are widely used in the
communications industry, particularly for
applications such as digital signal processing. Wilton hopes that by changing FPGA
architecture, the chip will offer advantages
to an even greater range of applications.
"The goal of my research is to reduce
problems with the speed and density of
FPGAs," says Wilton, who is collaborating
to that end with programmable logic
structures. The more flexible the memories,
the greater the number of applications that
will be able to use them, but the slower
supplier Cypress Semiconductor. "If we can     they will be."
do that, companies that currently can't use
them for certain applications will gain a huge
advantage in terms of cost-
savings and time-to-market
The route Wilton has
taken on his quest for a
faster FPGA involves modifying
hip's architecture to provide a portion
dedicated to storage or memory. Many
applications must retrieve rapidly from
memory. By implementing this memory on
the FPGA itself, rather than on a separate
chip, the overall performance of these
applications can be improved dramatically.
Enhance Performance
"In the past FPGAs have not been well
suited to implementing storage," Wilton
says. "By optimizing a small part of the
FPGA to implement storage we can greatly
enhance FPGA performance."
The challenge is not actually in implementing storage on an FPGA. Rather, it is
finding the best compromise between
flexibility and overhead for the memory
Wilton is also working with Cadence
Design Systems to develop new computer-
aided design (CAD) tools to enable circuit
designers to take full advantage of a new
FPGA architecture.
FPGA Architecture
"A new architectural design component
is wasted if CAD tools don't support it," he
says. "Given the proper design tools,
improved programmability of FPGAs and
faster speeds, manufacturers will be able to
use FPGAs for applications that may not be
attractive right now."
Thanks to the flexibility it provides, the
FPGA, says Wilton, has the potential to
play an ever greater role as its speed and
performance improve. A better FPGA, with
its existing advantages of flexibility and ease
of use, is bound to become an increasingly
valuable player in the chip marketplace.
Steve Wilton can be reached at (604) 822-1263
or at stevew@ee.ubc.ca.
"By optimizing
a small part of the
FPGA to implement storage we
can greatly enhance FPGA performance."
Spring 98 CICSR Passing Notes
New CICSR members
CICSR welcomes four new members from
Electrical and Computer Engineering
(ECE), all assistant professors:
Ed Casas is currently working with
wireless LAN protocols and low-cost implementation of LMDS systems. He can be
reached via http://www.ee.ubc.ca:8o/
home/comlab 1 /edc/etc/www/.
S. Sidney Fels is working with two
CICSR research groups: Computational
Intelligence: Perception, Reasoning and
Action, and Graphics, Visualization,
Auralization, Image/Video Processing, and
Human-Computer Interaction. His web
page is at http://www.ee.ubc.ca:8o/home/
Babak Hamidzadeh's research is in
computer systems and engineering. For
more information, see his web page at
CICSR Members honoured
Clarence de Silva (me) was Elected Fellow
of the IEEE December 1997 and will
receive the award at the 1998 IEEE Industry Applications Conference in St. Louis.
Mohamed Gadala (see FOCUS, Fall 97)
(me) received the 1997 NSERC University-
Industry Partnership Award with Tristar
Maria Klawe (cs) was awarded the BC
regional NSERC Chair for Women in
Science and Engineering in summer of 97.
Charles Laszlo (ece) receieved the
Distinguished Service Award from the BC
Association of Speech Language Pathologists and Audiologists in November 1997.
Takis Mathiopoulos (ece) was appointed
to the European Commission Advisory
Visionary Research Panel on Advanced
Wireless and Mobile Telecommunications
Jack Snoeyink (cs) received the Charles
McDowell Award for Excellence in Research in March 1998.
Rabab Ward (ece) was awarded a UBC
Killam Research Prize in February 1998,
and was elected Fellow of the Engineering
Institute of Canada (EIC) in December of
last year.
APEC contribution
CICSR members also contributed to the
Asia Pacific Economic Cooperation (apec)
conference held in Vancouver from
November 17 to 19, 1997.
Laboratory for Computational Intelligence members, headed by Computer
Sciences' Jim Little and Alan Mackworth,
demonstrated their robot "Jose." Jose is
capable of automatic surveillance, as well as
delivery, repetitive inspection and hazardous environment tasks.
Associate member Peter Cahoon (magic) ,
shared an interactive 3-D workspace with
Dalhousie's maxiofacial surgery unit as
well. The presentation was given with UBC
Assistant Professor of Medicine Barry
Wiggs. It was the first of its kind in
Canada, showing off the Dragon hierarchical surface editor developed by CICSR
member David Forsey.
ISO conference, 1999
CICSR and Image Power, Inc., a leading
developer of still-image compression
applications, will host the July 1999
meetings of Subcommittee 29 (SC29)
working groups of the International
Standardization Organization (iso). SC29 is
responsible for the international standardization of the coded representation of
images (wgi or jbig/jpeg), moving pictures
and audio (wgii or mpeg), as well as
multimedia and hypermedia information
(wgi2 or mheg).
The three working groups are composed
of hundreds of international experts in the
field of multimedia content representation,
and over 400 invited delegates are expected
to attend the meetings.
CICSR Centre for Integrated Computer Systems Research
The UBC Centre for Integrated Computer Systems Research (CICSR) is an interdepartmental
research organization made up of computer-related research faculty members in the Departments of Computer Science, Electrical and Computer Engineering and Mechanical Engineering.
Currently there are more than 70 CICSR researchers who direct over 300 graduate students and
collaborate with dozens of industrial firms in areas such as robotics, artificial intelligence, communications, VLSI design, imaging, and industrial automation.
Return Address
289-2366 Main Mall
University of British Columbia
Vancouver, BC, V6T 1Z4
Editors:   Will Knight, Linda Sewell
Writers:   Stephen Forgacs, Will Knight
Design:   wilyum creative
Photos:   Janis Franklin
Office:   289-2366 Main Mall
University of British Columbia
Vancouver, BC, Canada V6T 1Z4
Tel:   (604) 822-6894 Fax: (604) 822-9013
E-mail:   cicsrinfo@cicsr.ubc.ca


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items