T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A

Spearheading System-on-a-Chip ................... 3

Cutting Through Contradiction ........................ 4

Orchestrating Harmony ................................... 5

Capturing our Attention ................................... 6

Embracing Embodiment .................................. 7

Going Against the Flow .................................... 8

Computer Visionaries ...................................... 9

Geometry of Shadows and Light .................. 10

continued on page 6

When we sit down at our computer in the morning, it is difficult toimagine that less than twenty years ago PCs and laptops didn’t exist.The fax machine was revolutionary. Newspapers were still typeset byhand. Like any relationship, the one between humans and computershas had its ups and downs. Computers and related technology havetransformed the way we work, communicate, create, learn, and play.They have increased productivity, put a wealth of information at ourfingertips and connected us across the globe. They have also increasedworkloads and learning curves, caused repetitive strain injuries, andled to isolation and new addictions. For better or worse, they are nowan integral and increasingly intrinsic part of our lives. During the pasttwo decades, humans have been scrambling to catch up, keep up andadapt to new technologies.

To transform the way we use, design and interact with machinesand technologies requires a major conceptual shift—one that focuseson studying, understanding and communicating human experience.“ICICS will foster collaborations between researchers from all disciplines,

ICICS—Exploring Human Experience to Create Better Technology

Some of the key ICICS investigators left to right, front row: Rabab Ward, Jim

Varah, Sid Fels, KD Srivastava. Back row (l-r): Nick Pippenger, Raymond Ng, Alan

Mackworth, Dale Cherchas, Kelly Booth, Clarence de Silva, and Victor Leung.

particularly those studying health, emotions, social interactions, andour relationship with the environment,” says principal investigatorRabab Ward. It is a vision that is strongly supported. The ICICSproposal received over $22 million in funding: $8.85 million from theCanadian Foundation for Innovation (CFI), matched by $8.85 millionfrom BC’s Knowledge Innovation Fund, and $4.5 million from UBC’sBlusson Research Fund.

ICICS spans all disciplines

The process of writing the proposal took over a year and a half,and involved more than 80 faculty members. Jim Varah, professor ofComputer Science, and KD Srivastava, vice-president emeritus andformer ECE head, both helped author the proposal (which had 10senior applicants and 120 co-applicants). Not only was the initiativelauded for its innovation and vision, the breadth of disciplinesinvolved will make ICICS one of the most comprehensive and

$22 million in funding is transforming CICSR into ICICS, UBC’s Institute for Computing, Information & Cognitive Systems

Spring 2001 Vol. 12, No. 1

C E N T R E F O R I N T E G R A T E D C O M P U T E R S Y S T E M S R E S E A R C H

2 FOCUS

Computer Science professor and CICSRmember Anne Condon (profiled inFocus, Spring 2000) is currently theActing Director of CICSR while I am onsabbatical. Anne has agreed to take onthese extra responsibilities during anexciting time. In fact, we have so muchnews that we have expanded this issueof Focus to twelve pages.

The biggest story is CICSR’stransformation into ICICS. With $22million in funding from CFI, theProvince and UBC, our research willnow involve collaboration with disci-plines from across campus. The coverarticle and inside spread describes theprocess, the initiative, the infrastructure,and some of the groundbreaking workthat is being conducted under theICICS umbrella.

The other major story is aboutpeople—the talented new facultymembers that have recently joinedCICSR. Most have come to UBC fromacademic or industry positions inEurope and the US. This issue featuresIan Frigaard (non-Newtonian fluiddynamics), Wolfgang Heidrich (interac-tive computer graphics), Holger Hoos(artificial intelligence, computationalmusicology, bioinformatics), Tim Menzies(artificial intelligence), Ron Rensink(computational studies and psychophys-ics), Resve Saleh (System-on-a-Chip), andnew ME head Nimal Rajapakse (smartmaterials). You will also read about aCICSR/UBC spin-off success story, PointGrey Research. Their work in computervision technology is capturing theattention of companies worldwide.

Rabab Ward, CICSR Director

CICSRDirector’s Corner

When he speaks of “active control,” Nimal Rajapakse is referring to an essential feature of smartstructures. But the term could also apply to his philosophy of work and life. UBC’s new headof Mechanical Engineering is committed to the core values of his job: excellence in teachingand research, innovation and collegiality. In 2000, the CanadianSociety for Civil Engineering awarded him the Horst LeipholzMedal for his outstanding contributions to engineering mechanicsresearch.

Rajapakse came to UBC from the University of Manitoba, wherehe was also department head. “This is a dynamic department with agreat group of colleagues, excellent students and very helpful staff,”he says. “Research today is increasingly interdisciplinary and ICICS’sinfrastructure really supports this.” Rajapakse’s research—modellingof adaptive (smart) materials—involves continuum mechanics,material science, physics, and mathematics, as well as extensivecomputer-aided simulation.

Smart structures sense and adapt to change through a system of embedded sensors andactuators made out of adaptive (smart) materials. These structures are of particular interest tothe defence and aerospace industries, which are committing considerable funding to this areaof research, particularly in Europe, Japan and the US. NSERC and Manitoba Hydro havebeen funding Rajapakse’s smart structures research since 1995.

Rajapakse’s particular area of expertise is in piezoelectric/ferroelectric solids, crystallinematerials that exhibit unique behaviour under mechanical loading (pressure) or the applicationof electricity. When subjected to pressure they create electricity (this is the concept behindsensors). Conversely, subjecting these solids to an electric field generates a strain in the material(the concept behind actuators). Since they can be used as both sensors and actuators, piezo-electric solids are very important to the design of smart structures. They are also used in thinfilm applications involving microelectromechanical systems.

One example of these applications is helicopter blades with embedded piezoelectricactuators for adaptive shape (camber) and vibration control. Another is the active control ofvibrations in space structures. Smart materials also have important applications in civil engi-neering, particularly in structural problems related to earthquakes. They could also provide amore comfortable driving experience. “A smart suspension system in a car could sense theprofile of the road and provide better control and comfort,” says Rajapakse, referring to arecent application of piezoelectric ceramics by a leading automaker. Smart skis can even takethe edge off of moguls. Nimal Rajapakse is at rajapakse@mech.ubc.ca or at (604) 822-0497

New ME Head Pioneers Smart Structures

Recent books by CICSR members

Yusuf Altintas, Manufacturing Automation: Metal CuttingMechanics, Machine Tool Vibrations, and CNC Design(Cambridge University Press, 2000).

Clarence de Silva , Intelligent Machines—Myths and Realities(CRC Press, 2000); and Vibration—Fundamentals andPractice (CRC Press, 2000).

Anne Condon (left)and Rabab Ward

“This is a dynamic department with a great group of colleagues,excellent students and very helpful staff.”

Spring 2001 3

continued on back page

The widening gap between technology andproductivity is fuelling a revolution in siliconchip design. Advances in deep submicrontechnology (DSM) have escalated theamount of logic, circuitry and functions thatcan be designed into chips. Currently,hundreds of millions of transistors can be puton a single chip, and this is predicted toincrease to one billion before the end of thedecade. “The problem is the number oftransistors that designers can build in a day isnot increasing significantly,” says Electrical andComputer Engineering professor Resve Saleh.He believes we need to change the way chipsare constructed to solve this productivity gap.

Challenges to SoC

The concept of System-on-a-Chip (SoC)is simple. A computer chip is made up ofseveral blocks, many of which performstandard functions. Instead of redesigningthe blocks that control functions such asmemory, interface blocks and processingunits, they would be reused. However, thereare several hurdles to overcome before SoC isfeasible. Since technology changes so rapidly,what works on today’s design may not workon subsequent ones. And the concept ofreuse is not built into chip design because ofthe cost, effort and the time it takes to get tomarket.

Another challenge to SoC is the technol-ogy itself. Designers have to work at a higherlevel of abstraction. Putting transistors closertogether and driving more current throughthem results in more coupling and substratenoise and other undesired electrical effects.

“The next generation of designers mustunderstand what’s going on both at thesystem level, and at the silicon level,” saysSaleh. In order to train designers to work inSoC, his current course in DSM integratedcircuit design will become a permanent partof the curriculum.

Perhaps SoC’s biggest challenge is integrat-ing digital and analog components on a single

Spearheading System-on-a-ChipResve Saleh’s mission is to create a world-leading research centre for System-on-a-Chip design,testing and verification.

chip. More analog blocks arerequired for wireless, LAN andpacket processing applications,and analog requires more voltageand different processing optimiza-tion than digital circuitry. It alsoworks best in quiet environments.Digital circuitry, which constantlyswitches from low to highvoltage, is very noisy. “David Pulfrey of theSoC group is exploring these issues in hisresearch,” notes Saleh.

QoS and programmable IP

Designing reusable intellectual propertyblocks involves issues of legality, licensing,

management, storage, andretrieval. And standardsmust still be developed sothat third party IP will becompatible with propri-etary design blocks. Qualityof Service (QoS) prioritiesare also difficult to establishwhen industry standards

are constantly changing.Along with Steve Wilton and Andre

Ivanov, Saleh is working to develop program-mable IP blocks with on-chip testingcapability, and Computer Science colleagueAlan Hu is investigating issues of high-levelverification of SoC design.

Resve Saleh is a member of the International Roadmap of Semiconductors Committee.

“Just putting things on a

chip and hooking them

together doesn’t

guarantee that it will

work. In fact, right now it

probably guarantees

that it won’t work.”

4 FOCUS

The concept of using randomized search forsoftware testing still raises indignation—andeyebrows—among many software engineers,who prefer to use established methods toensure that software products are reliable andpredictable. However, they face an increas-ingly daunting challenge. As softwarerequirements increase, so does complexity,and testing involves solving extremelydifficult NP-hard problems (computationallyintractable problems for which there are noknown efficient algorithms) or trying to findthe best possible solution within a web ofvariables and possibilities. In the dog-eat-dogdot.com world, software developers simplydon’t have the time or money for traditionaltesting.

Studying ants to understand software

“Pragmatically, we know that 60 to 80percent of the requirements of any systemappears after the first version of software hasbeen fielded,” says cognitive scientist TimMenzies. “Whatever picture you have at thestart is at best 20 to 40 percent correct.”Menzies explores how quirks in humancognition affect the process of software andknowledge engineering. Of all the variables insoftware design, the greatest are humanfactors such as having different designers withtheir own approaches and concerns workingon the same product. “Yet somehow, reason-ably good software is produced most of thetime,” he says.

He and Computer Science colleagueHolger Hoos are part of a group of researcherswho are studying the self-organizing behav-iour of ant colonies to try and unravel theshape of software and the mystery of why, asa rule, it works so well. “If we try to describesoftware as pathways, are they expressways ortangled spaghetti?” Menzies asks. The answerseems to be a bit of both, and the emergentbehaviour of ants provides a clue to theinherent order in software’s randomness.Consider the apparent chaos of an ant colony.When forager ants search for food they laydown a trail of pheromones. The other ants

Cutting Through Contradiction in New Software Models

Tim Menzies uses AI tools to find key testing variables in the complex chaos of software code.

follow the scent of the forager that returnsfirst, and this becomes the expressway to thefood source.

Now, imagine software as ants running ina maze, or trying to get from the head to thefeet of a skeleton. There are things the antsdon’t know about the space they are explor-ing, and they have differences of opinionas to the best way to get from point Ato point B. Should they go down onearm, or another, or down the spine?Eventually, they will choose the spine.Menzies is interested in settling these“arguments” in software design byhelping developers sift through redun-dant options to find the pathway of leastresistance. His “funnel theory” shows howthis narrow “software backbone” containsvariables that appear in every solution,producing an expressway out of the spaghettiof code. Simple machine learners can exploresoftware with funnels to automatically findorder within the apparent chaos.

Tarzan: finding pathways in a jungle of code

“The surprising observation is that if youtroll the space of possibilities, you will usually

find a small number of master variables thatcontrol the rest of the show,” says Menzies.He is currently working with Marcus Featherand NASA’s Jet Propulsion Lab on require-ment engineering for deep space missions.Flight software has to be highly reliable andthere are always early life-cycle decisions

where people disagree.Menzies’ early life-cycle assessment

tool, called Tarzan, can sift throughmillions of combinations to determinewhich ones offer the most directpathway to the best solution. He notesthat Tarzan’s ability “to swing through

decision trees” has many other softwareapplications.

Menzies believes that most software codeinherently generates these expressways. “Ifyou think about how complex the universeis, and how many options we face each dayand how we manage to muddle on through,either we have some wonderful control overthe universe, or there is more order in itsrandomness than we thought. My money ison the universe.”

Tim Menzies can be reached at

timm@ece.ubc.ca or at (604) 822-3381

Assistant Professor of

Electrical and Computer

Engineering, Tim Menzies is

also co-founder and organizer

of WISE, the international

Workshop on Intelligent

Software Engineering.

Spring 2001 5

Computer scientist, musician, theorist, scubadiver—Holger Hoos is truly a man of thepost-modern renaissance. His ability to thrivein both research and creative environmentshas led to a unique cross-pollination ofresearch interests. A bassoonist who playedregularly with several ensembles in Germanybefore coming to UBC, Hoos admits thatmusic has influenced much of his research.“The department here is fantastic. I thinkICICS will provide great opportunities for meto tie all of these interests together,” he says.

Hard problems & stochastic search

Hoos’ initial research was on thesatisfiability problem (SAT), a prototypicalhard (NP-complete) combinatorial problem,which involves a number of logical variablesconnected with “and” and “or.” SAT plays acentral role in three key areas of computerscience: theory and computational complex-ity, artificial intelligence and automatedreasoning, and hardware design andverification. As software systems and theirunderlying algorithms become more complex,rigorous theoretical analysis is increasinglydifficult. Although it is possible to see and

analyze each line of code, it is often next toimpossible to predict the global behaviour ofalgorithms, and this is critical in being able toapply or improve software systems.

Along with Electrical and ComputerEngineering colleague Tim Menzies, Hoos is apioneer in the emergent field of empiricalcomputer science. “Tim and I tend to viewcertain algorithms and systems as physical orbiological phenomena, where you don’t reallyknow, or care, what happens at the mostdetailed level. Yet you can observe, hypoth-esize and model the system’s behaviour, andthen test it through further experiments,” hesays. For example, stochastic search algo-rithms use randomized decisions that canactually increase the efficiency of the search.However, the resulting behaviour is verycomplex and can often only be analyzedempirically. He and Menzies draw onbiological models such as Ant ColonyOptimization to try to build a theory on howstochastic search techniques can solve NP-hard problems faster and more efficiently.

Hoos’ interest in biology dates back to hisundergrad studies at TU Darmstadt inGermany. Along with Computer Science

colleagues Anne Condon, Nick Pippengerand David Kirkpatrick, Hoos recentlyfounded the Bioinformatic, Empirical andTheoretical Algorithms Laboratory (ß-Lab) atUBC. Bioinformatics is a new field of studythat combines research in molecular biology,computer databases and algorithms in orderto facilitate biological research, particularly inthe field of genomics. Hoos, Condon and ß-Lab colleagues are working on projects suchas inverse RNA folding, DNA word design,and phylogenetic analysis for environmentalecology applications.

Computational musicology

Hoos’ work in computational musicologyinvolves collaborations with colleagues acrosscampus, including Keith Hamel from theSchool of Music. Hoos developed the guidoMusic Notation Format to represent music(and the parameters of duration, tone, pitch,and instrumentation) at a logical level in anelectronic, humanly-readable form. Themotivation for GUIDO came from his earlywork on SALIERI, a software environmentthat supports the computer-assisted composi-tion, manipulation and analysis of music. Inaddition to being able to quickly transpose,edit, listen to music, and see the score, theuser/composer can create computer musicfrom mathematical concepts or even biologi-cal sequences. “One exciting example is youcan render DNA sequences musically, andthis can complement our visual perception ofthis kind of data in a most enthralling way.”

Does Hoos still find time to play bassoon?“I really miss playing, but bassoonists arepretty rare so I’m confident that I will find anensemble to play with,” he says. “The onlything that’s not ideal about my situation hereis I am involved in too many interestingthings.”

Holger Hoos can be reached at

hoos@cs.ubc.ca or at (604) 822-1964

Orchestrating the Harmony of BitsHolger Hoos designs and analyzes algorithms for an eclecticrange of areas, from artificial intelligence and computationalmusicology to bioinformatics.

Bassoonist and Assistant Professor of Computer

Science, Holger Hoos is also co-organizer of the

IJCAI Workshops on Stochastic Search Algorithms

and Empirical Methods in Artificial Intelligence.

6 FOCUS

Capturing our Attention

unique interdisciplinary research institutes in the world, and bring together researchersfrom the faculties of Applied Science, Arts, Commerce, Dentistry, Education, Forestry,Medicine, Pharmacy, and Science.

The new ICICS facility is scheduled for completion in the 2002-2003 academic year. Itsspace and equipment will support research that consists of three overlapping components:human communications technologies, multi-agent systems and global information systems.The largest infrastructure component will be a fully-equipped human communicationstechnology laboratory that will include a human observation and measurement lab, avirtual reality room and a fully instrumented system demonstration lab. A robust, highbandwidth intra- and inter-building communication network—complete with manipula-tors and robots—will explore multi-agent technology. A parallel router/switching facility willsupport research in global information systems.

The research undertaken at ICICS will have applications in e-commerce, education,engineering, entertainment, linguistics, medicine, and psychology. “The breadth ofdisciplines that are represented is quite unique,” says Srivastava. For example, the largeamount of funding available for IT research in the US means that most of their academicinstitutions tend to specialize in certain areas. ICICS’s founders all agree that improvinghuman-computer interaction begins with improving interaction between people—and thismeans bringing researchers together under one well-equipped roof.

ICICS continued from page 1

The new 30,000 sq. ft. ICICS facility will include

4 Onyx II Infinite Reality Systems for real-time

image generation

Optotrak for high-speed tracking in the Human

Measurement Studio

Electromagnetic Articulograph (EMA) for tracking

the human face and tongue

6 degrees-of-freedom tracking devices, body suits,

cybergloves and stereo head-mounted displays for

virtual reality and interactive visualization projects

Staubli AX170 Unimation and CRS A465 Ind. Robotic

robots for intelligent multi-agent systems research

2 fully wired and switchable large video projection

rooms with high-quality audio

Research infrastructure for file services, online A/V

services, backups, and multimedia production

ICICS infrastructure

“We want to change the focus of our research so that human needs come first,”

How does a magician mesmerize? Whatprocesses are or aren’t at work that allow us tobe convinced by “now you see it, now youdon’t” sleight of hand? Why do we sometimessimply not see things that are right in front ofus—like a car bumper, for instance? Anassistant professor in Computer Science andPsychology, Ron Rensink’s work in attentiveand pre-attentive processes is unravelling someof the riddles of perception and consciousness.

Rensink came to this field of study by arather circuitous route. After completing hismaster’s degree in theoretical physics at ubche went on to study ai and computer vision.“I was inspired by Bob Woodham to talk toresearchers who study real vision,” saysRensink. He began working with AnneTreisman and later Jim Enns in the Depart-ment of Psychology and this led to a postdocat Harvard’s Vision Sciences Lab. Rensink’swork in computational studies and psycho-physics led him to discover that the processesinvolved in human sight—such as interpret-ing three-dimensional curvature, depth and

To Ron Rensink, seeing is not necessarily perceiving. He is one of several new CICSRmembers whose study of human experience is key to ICICS.

surface properties—are much more complexthan originally thought.

Looking without seeing

From Harvard, Rensink went to work forCambridge Basic Research (CBR), a collabo-ration between MIT, Harvard and Nissan.There he studied the issue of invisibility, orwhy so many drivers involved in trafficaccidents claim not to have seen what wasright in front of them. At first it was thoughtthe phenomenon was caused by poorvisibility. “It turns out it is attentional,” saysRensink. “You actually have to focus yourattention on the thing that is changing inorder to see it, otherwise you will effectivelybe blind to the change no matter how large itmay be.”

That means if a driver in front of you putson the brakes, and you are changing a CD, ortalking on a cell phone (as a recent ICBCstudy has confirmed), you probably won’tnotice until you actually focus your attentionon the brake lights. And with cars becoming continued on page 11

” says CICSR director Rabab Ward.

Embracing Embodiment

Computer

Scientist and

Psychophysicist

Ron Rensink is

helping to

design simpler,

more effective

interfaces for

increasingly

complex tasks.

Sid Fels: “I think an artist goes through a similar

process as an engineer in that both are involved

with bringing an idea into existence—into

something you can see, hear, use, or appreciate.”

“Technology is driving a way of thinking that was not possiblebefore,” says Sid Fels. “And this is changing how we defineourselves and our tools.”

continued on back page

Electrical and computer engineer

and media performance artist, SidFels spends a lot of time thinkingabout the nature of technology andour relationship with it. One of theten principal researchers on theICICS proposal, Fels explores issuesof embodiment and human-computer interaction.

Forklift Ballet

Fels’ multimedia dance collabora-tion, Forklift Ballet, was recentlyperformed in Acqui Terme, Italy.In the piece, skilled drivers guideelectrical forklifts about the stagein a semi-choreographed “ballet.”An accompanying musicianperforms synthesized music by movingtwo wands through space. The balletexamines the way that people embodytechnology. To drive a forklift withproficiency and precision involves anintimate relationship between human andmachine, where the machine becomes anextension of the body and its use involvesboth physical skill and expression. Incontrast, the musical wands are difficult toembody, and the music they produce seemssomewhat devoid of expression beside thegraceful movements of the machines.

“The reason is that forklifts—andconventional musical instruments—give youforce-feedback, so they really do become partof your body,” says Fels. He believes thathuman aesthetic, physical and emotionalneeds must be integrated into engineeringand design in order to develop more adap-tive, intuitive and satisfying interfaces.

Modelling human speech

Fels has already created a system thatsupports embodiment—Glove Talk II—which allows users to talk with their hands

(see Focus Spring 1999). He and other ICICSresearchers, including Bryan Gick (PSYC) ,Dinesh Pai (CS) and Kathy Pichora-Fuller(AUDI), are working on articulatory speechsynthesis, which involves modelling humanspeech to produce more natural-soundingsynthesized speech.

“Think of your vocal tract as an irregularlyshaped tube, with a tongue and uvula andfleshy walls of muscle,” says Fels. “Current

2d models condense those shapes intosimple circular cross-sectional areas.” Tounderstand and model how our vocaltract really works first involvescollecting data from imaging toolssuch as x-rays, MRI, ultrasound, andmagnetic tracking of the tongue.Airflow is then simulated through

the vocal tract and articulatory phoneticsis used to determine how movements ofthe tongue, uvula, cheeks, and lipschange the shape of the vocal tract tomake different kinds of sound.Would an automated message be less

annoying if the voice sounded sincere?Would the voice of an AI tutor be morecaptivating if it sounded more human?Possibly. Certainly avatars such as Britain’sAnna Nova or the late Max Headroom wouldbe more life-like. Since articulatory speechsynthesis connects the face and vocal tractmodels, the facial movements would beperfectly synchronized with speech.

“We know this is a major contributor tothe understanding of speech,” says Fels. “Webelieve that these articulation patterns will bemuch easier for voice recognition systems touse.” And if it is easier for computers tounderstand us, it will be easier to designbetter voice-based interfaces.

Changing human-computer relationships

Whether developing AI applications orcreating new media performances, Fels

8 FOCUS

When we squeeze a tube of toothpaste,dollop whipped cream on a latté, or put gelin our hair, we likely don’t think of thesesubstances as “fluids” or consider the complexproperties that make them behave the waythey do. Most fluids are Newtonian, withpredictable behaviour and viscosity (orresistance to flow when subjected to stress).Non-Newtonian fluids (such as pastes andgels) display such a complicated array ofproperties that their errant behaviour stilleludes theorists. It is this very complexity thatfascinates new CICSR member Ian Frigaard.

“I am motivated by physically complexengineering problems in which analysisrequires a mix of different mathematicalmethods.” Frigaard’s concurrent academic/industry experience played a crucial part indirecting his research. He completed his PhDat Oxford while working for Alcan, and as apostdoctoral student at Cambridge heworked as an oil and gas research engineerwith Schlumberger. Both positions involvedthe study of non-Newtonian fluids.

“Schlumberger is one of the few globalindustrial companies with a dedicatedresearch facility for fluid mechanics,” saysFrigaard. As a rule, companies tend tocontract out this type of research, which isoften conducted with university partners. Hesays the challenge is for academic researchersto understand what companies want, and forindustry to realize that solutions to complexengineering problems require basic research.

Solving “mud problems”

One example is drilling for oil. Rotarydrilling involves controlling the complexinteraction between two non-Newtonianfluids: drilling mud and cement. Drillingmud is a weighted fluid that is circulated intothe well bore through the drill pipe. The

Going Againstthe Flow

mud’s hydrostatic pressure prevents waterand gas from seeping into the well andcausing blowouts or “gushers.” It also carriesrock cuttings to the surface and lubricatesand cools the drill bit. However, wells usuallydeliver oil, gas and water in continuallychanging proportions, and developingtreatment fluids with the right viscosity andweight to maintain the desired pressuredepends on continuously changing variables.Modelling this complex flow behaviour iscritical to developing the right drilling mudfor the job.

In deep drilling, a series of progressivelysmaller steel casings are inserted into the welland cemented to the outer circumference ofthe borehole to prevent the transfer of fluidsinto the well. This method allows wells to bedrilled to depths of more than 9,000 metersthrough rock formations that have fluidpressures greater than 1,400 kilograms persquare centimetre. The difficulty is gettingthe cement down into the well, which meansdisplacing one non-Newtonian fluid withanother. “The deeper you go, the lessflexibility you have on how fast you canpump fluids,” says Frigaard. “As the wellbecomes narrower and more complex, itschanging geometry alters the fluid behav-iour.” The very characteristics of the mud, thedensity and weight required for drilling,make this displacement extremely difficult.

Frigaard is currently working with two UBCgrad students to help Schlumberger solvethese “mud problems.”

Capping oil wells

A major concern for Canadian oil compa-nies is pressure and leakage at the surface ofcapped wells. To stop surface casing vent flows,companies must first determine the depth ofthe leakage, make a hole in the casing and theninject it with cement. In addition, a series ofcement plugs must be inserted at variousdepths of the well to seal the casing hole. Sincecement is heavier than the fluid beneath it, theproperties of the mud must be such that thecement remains suspended on top and driesand seals without falling down the well.Frigaard’s research is helping companies likeSchlumberger avoid the extensive cost ofrepairing leaks by ensuring that their wells areenvironmentally sound at the outset.

“The challenge is not necessarily in themathematics, but in its application tocomplex problems, and this involves under-standing the dynamics of the systemsinvolved,” says Frigaard. “With the PacificInstitute of Mathematical Sciences (PIMS)and the Institute of Applied Mathematics(IAM), UBC is one of the strongest places inNorth America to pursue this kind of work.”

Ian Frigaard is at (604) 822-1316 and

frigaard@mech.ubc.ca

“I am motivated by physically complex

engineering problems in which analysis requires

a mix of different mathematical methods.”

Ian Frigaard models complexflow problems in non-Newtonianfluids for industry partners.

Spring 2001 9

continued on page 12

Imagine a computer that recognizes your face,greets you in the morning, and logs in all ofyour passwords automatically. Imagine beingable to search databases of 3d images that youcan rotate with simple hand and face gestures.Or, imagine machines that perform complextasks in hostile environments, such as under-ground mining and arctic mapping, withoutthe need for human intervention. Computervision is the core technology on which muchartificial intelligence (AI) research andapplication depends.

In four short years, Vancouver’s PointGrey Research (PGR) has become a leadingdeveloper of computer vision hardware andsoftware for an expanding array of applica-tions. The company was founded in 1997 byformer CICSR MSc students VladimirTucakov and Malcolm Steenburgh; DonMurray, currently a PhD student at ubc; andRod Barman and Stewart Kingdon, formertechnical staff of the Institute for Robotic andIntelligent Systems (IRIS). “Much of theresearch leading up to PGR’s spin-off wasfunded by IRIS’s Technology-GAP program,”says Tucakov.

Digiclops and Triclops

The core of PGR’s computer vision hard-ware is Digiclops, a digital stereo visioncamera that captures 3d imagery in real timeand on-line for applications such as objectand people tracking, virtual reality modellinglanguage (VRML), human-machine interfac-ing, and mobile robotics. The Digiclopscamera is an extremely sophisticated sensorwith a large field of view. Unlike laser sensors,which scan an object by beaming light at it,the Digiclops is passive, instantaneous, andoperates in dynamic and unstructuredenvironments. Three cameras are used tocapture 3d information, so high calibration isthe keystone of the Digiclops system. Allimages are synchronized internally, producingaccurate and dense 3d measurements of bothhorizontal and vertical features. Triclops is the

Point Grey’s Computer VisionariesUBC spin-off Point Grey Research is trailblazing computer vision technology for a host of emergingand yet-to-be-envisioned applications.

complementary software development kitthat enables accurate, high-speed 3dprocessing of digital stereo images.When the products were launchedin December 1999, the demand wasso great that the company sold out thefirst production run before the first unit waseven completed.

The company is working with high-profile players such as Microsoft, Intel,Hewlett-Packard, MIT, and NASA on severalhigh-cost, low-volume applications. Forexample, Microsoft uses PGR’s Digiclopscamera in the development of smart homesand intelligent environments. While workingwith these companies enhances PGR’scredibility in the marketplace, Tucakov

believes the real growth potential lies in high-volume, low-cost applications. “We can

customize the core technology to awide range of applications,” he

says. “The challenge now is tofind strategic partners and

identify their needs.”

Emerging markets and applications

Point Grey Research is currently develop-ing applications in the retail, safety andsecurity industries. The company’s Censys3dsoftware can recognize and track individualsin a crowd under varying lighting andenvironmental conditions. People tracking isused in retail environments to improve

Point Grey Research: (from left to right) Don Murray, Malcolm Steenburgh,

Rod Barman, Vladimir Tucakov, and (front) Stewart Kingdon.

10 FOCUS

Have you ever watched a televised weatherreport and noticed the peculiar way shadowsare cast on the set? Or wanted to buy an itemon-line, but didn’t have enough visual in-formation to make a decision? These are someof the problems that Wolfgang Heidrich ishoping to remedy. A new CICSR memberand assistant professor in Computer Science,Heidrich came to UBC from the Max PlanckInstitute for Computer Science in Germany,where he was working on interactive compu-ter graphics for use in industrial design.

Photorealistic rendering in industrial design

One of Heidrich’s projects in Germanywas the simulation of car interiors withBMW. The traditional process for mostindustrial design applications has been tobuild several physical prototypes. Whilecomputer modelling and simulation is nowchanging this process, the problem withtraditional computer graphics is that theimages tend to be so pared down that theyno longer have any physical meaning. “Itlooks moderately okay, but you have no wayof making anything that follows physical lawsand corresponds to something that exists inthe real world,” Heidrich says. While interac-tive computer graphics can simulate materialsin an immersive (virtual reality) environment,they still require more refinement beforeindustrial designers can use them to makemeaningful decisions.

Global illumination

Typical problems with current renderingmethods involve how light interacts with thesurface of objects. Simple graphic renderingsuse local illumination, which model a singlelight source interacting with an object’ssurface. But most light sources are not direct,and light bouncing off walls and othersurfaces affects how we perceive shadows,colour, texture, and depth. Global illumina-tion takes all of this light activity intoaccount. This is particularly important inphotorealistic rendering and in applications

such as lighting design—for example, whereyou put a light source in a car console. “If youare driving at night, and wearing a whiteshirt, you don’t want a light source thatshines onto your arm, because you will not beable to see anything outside,” says Heidrich.

Image-based rendering and microgeometry

Image-based rendering uses calibratedcameras and light sources to capture geometryand measure the appearance of materials sothat they can be more easily modelled andgraphically rendered. This works well forsimple smooth-surfaced objects and materialsthat are not too shiny. However, photo-realistic simulation of complex materials suchas fabric and carpet requires not only model-ling the geometry of the fibres at a micro-scopic level, but also modelling the micro-scopic reflection of light bouncing off eachfibre.

Heidrich has some remarkable examples ofthese techniques. In one simulation of aroom, an image of sheer curtains has the effectof silk taffeta, where the warp and weft

threads are different colours. The fabricshimmers and changes colour with the angleof the light. In another example, a picture ofclouds is illuminated by a single light source.“Here, you only see the light that comes offthe clouds directly. But when you alsosimulate how the light bounces aroundbetween the different water particles withinthe cloud, you get a much more realisticpicture,” Heidrich says. Indeed, the effect isstartling. He and colleagues have developed aglobal illumination algorithm to enable moreefficient and faster modelling of micro-geometry and global illumination effects.

Applications for interactive computergraphics include virtual and augmented reality,e-shopping catalogues, computer games, andspecial effects for the movie industry. “Withthe research expertise at UBC, and companieslike Electronic Arts, Radical Entertainment,and MainFrame here in Vancouver, this isdefinitely one of the best places for me topursue my work,” says Heidrich.

Wolfgang Heidrich can be reached at

heidrich@cs.ubc.ca or at (604) 822-4326

By studying how light shapes and shades objects in the natural world, Wolfgang Heidrichgenerates computer images that are more photorealistic.

Spring 2001 11

Dr. Alain Fournier

(1943-2000)

Alain, a CICSRmember andprofessor in theDepartment of

Computer Science, passed away onAugust 14, 2000. He is fondlyremembered by his colleagues, hisstudents, his family and his manyfriends.

An endowment fund has beenestablished to honour Alain and hiscontributions to the field of computergraphics. The endowment will fund ascholarship to a PhD student incomputer graphics at a Canadianuniversity.

Alain’s influence as a scientist, ateacher, a mentor and a friend wentfar beyond the boundaries of UBC.He is remembered with respect,gratitude and abiding affection. Thescope of his research profoundlyaffected the field of computer graphicsat many levels. It is therefore fittingthat the scholarship should benational in character, and that itshould be adjudicated by Alain’sformer doctoral students, in conjunc-tion with the Department of Compu-ter Science at UBC.

We invite friends, colleagues,alumnae—indeed, anyone who feltAlain’s influence—to contribute to theAlain Fournier Scholarship. It is ourhope and expectation that Alain’smemory will live on for generations tocome through this endowment. Pleasesend your contributions to

The Alain Fournier Memorial Fund

Department of Computer ScienceUniversity of British Columbia201-2366 Main Mall,Vancouver, bc, v6t 1z4canada

CICSR Passing NotesIn the last issue of Focus we promised a profileof Rob Rohling (ECE). Rob arrived only inJanuary, so we had to postpone his story. Kris

de Volder (CS)and Will Evans (CS) alsojoined us in January so watch for profiles ofthese new CICSR members in the future.

Vinod Modi (ME) receivedthe Control AuthorityAward for his entry entitled“The Moving SurfaceBoundary-Layer Control,”at the Fluids 2000 confer-ence of the American

Institute of Aeronautics and Astronautics.

Nick Jaeger (ECE) has received the NSERCUniversity-Industry Syngery R&D Partner-ship Award, in conjunction with Farnoosh

Rahmatian of Nxtphase Corporation, Brent

Sauder of BC Advanced Systems Institute(ASI), Greg Polovick of BC Hydro, and Vern

Buchholz of Powertech Labs. The awardrecognizes the collaboration that led to thetransfer of photonics technology from Nick’slab to the power measurement marketplace.

David Pulfrey (ECE), anew CICSR member, hasbeen elected a Fellow ofIEEE “for contributions tothe modelling of hetero-junction bipolar semicon-ductor devices.” (More on

David in a future issue.)

Recent winners of ASI Research Fellowshipgrants include Cristina Conati (CS), Ian

Frigaard (ME), Wolfgang Heidrich (CS),and Robert Rohling (ECE). Victor Leung,Babak Hamidzadeh and Matt Yedlin (all ofECE), in partnership with Telus, havereceived funding from the ASI StrategicResearch Program.

smarter and dashboards more like cockpit controls, drivers have much more to pay attentionto. Nissan is looking at possible applications of Rensink’s research on attention processes inbrake light design. They found that having lights flash on two or three times, instead of onlyonce, produces far better detection of brake light onsets than standard brake light systems.“Many problems that are interesting from an academic perspective are also of great interest toindustry,” says Rensink. “This is just one example.” He is also working on applications incomputer vision and human-computer interfaces.

Controlling the chaos of free flight

If today’s drivers are overwhelmed with information, imagine how air traffic controllers mustfeel. “With so many people flying, the density and capacity of traffic control systems isstretched to the maximum,” says Rensink. As a result, the current “highways in the sky” systemwill eventually change to free flight, where flight patterns will be more random and controllerswill have to do more manual tracking. Obviously, this involves a greater risk of human error.Rensink is working with Kelly Booth, Karon MacLean and Brian Fisher from ComputerScience, and Jim Enns from Psychology, to develop new displays for air traffic control systems.

“If we know what kinds of errors humans make, we can design our systems to compensatefor them.” Rensink notes that our attentional processes account for only part of what we areactually aware of, therefore our brains are able to process much greater amounts of informationthan we realize. He wants to replace some of the conscious “thinking” processes, such as havingto interpret coded instrumentation, with “intuitive,” non-attentional processes of other sensorysystems, such as simple visual and auditory recognition. The trick is understanding—andlearning how to trust—these more intuitive processes. “We need to download some of theconscious work and let our unconscious mind do the walking,” says Rensink.

Ron Rensink can be reached at (604) 822-0598 or at rensink@cs.ubc.ca

Rensink continued from page 6

CICSR Centre for Integrated Computer Systems Research www.cicsr.ubc.ca

Return Address:

CICSR, University of British Columbia289-2366 Main Mall, Vancouver, BC, V6T 1Z4CANADA

Writer: Mari-Louise Rowley,Pro-Textual Communications

Design: William Knight, wilyum creative

Photos: Janis Franklin, Greg MortonUBC Media Group

Office: University of British Columbia289-2366 Main MallVancouver, BC, Canada, V6T 1Z4

Tel: (604) 822-6894

Fax: (604) 822-9013

E-mail: cicsrinfo@cicsr.ubc.ca

Contact: Linda Sewell, Publications Coordinator,CICSR Office

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 80 CICSR faculty members who direct over 350

graduate students and collaborate with dozens of industrial firms in areas such as robotics,

artificial intelligence, communications, VLSI design, multimedia, and industrial automation.

“I believe the greatest challenges also present the greatest researchopportunities,” says Saleh. He notes Canada has taken a very strategicview of SoC. The Canadian Microelectronics Corporation (CMC)has been funded to the level of $40 million to build the infrastruc-ture, acquire the intellectual property and install the managementfacilities to allow Canadian researchers to access IP through a virtualprivate network.

This was a large part of Saleh’s motivation to move back toacademia from industry positions at Simplex, Nortel, Tektronix,

Toshiba, and Mitel. He chose Canada because of the CMC infra-structure and the high quality of expertise at UBC. He also creditsindustry partner PMC-Sierra for major funding and support. Hebelieves all of the ingredients are in place to create a research centre ofexcellence in SoC/IP. “We have innovative experts in the field, high-visibility projects, state-of-the-art equipment, lots of funding fromindustry and government, and the ability to attract key researchers.”

Resve Saleh can be reached at res@ece.ubc.ca or phone

(604) 822-3702

Point Grey Research continued from page 9

Fels continued from page 7

Saleh continued from page 3

“We can customize the core

technology to a wide range of

applications. The challenge now is

to find strategic partners and

identify their needs.”

doesn’t distinguish between the practical andartistic applications of his work. He creditsJapan’s Advanced TelecommunicationsResearch Laboratories (ATR) for much of hisfunding.

“If someone creates a new technology forbungee jumping, it is really no different than

together to change the way we think aboutand design technology—not merely assomething that gets a job done, but assomething from which we get pleasure andsatisfaction.”

Sid Fels is at ssfels@ece.ubc.ca and

(604) 822-5338

if I create some weird interface for makingmusic,” he says. And whether for aestheticenjoyment, thrill seeking or increased produc-tivity, the beauty and merits of any technol-ogy—like art—rest with the viewer/user.

“ICICS will provide an opportunity forscientists, engineers and artists to work

shopper-to-cashier ratios, labour schedulingand product placement. It also helps withmerchandising decisions by monitoring howmany shoppers stop at product displays andfor how long. For security and safety systems,the advantage of this technology over videomonitoring is that someone doesn’t have tocontinuously watch a screen in order toidentify suspicious behaviour, unauthorizedindividuals, or personnel entering hazardousareas. The camera’s 3d sensors are not affectedby shadows or sunlight and can accurately

pick individuals out of a crowd. PGR is alsoresearching applications in virtual reality, com-puter interfaces, interactive entertainment,medicine, manufacturing, and the automotiveindustry.

“What we are really undertaking is atechnology push, which is very difficultbecause the applications are still emerging,”says Tucakov. But he is confident that oncecompanies understand what PGR’s productscan do, the applications will follow. “Astechnology becomes more affordable,computer vision applications will becomemore commonplace.”

For further information on Point Grey

Research, contact Vladimir Tucakov at

tucakov@ptgrey.com