Intelligent Systems Research Finds Its Center
by David Pescovitz
Stuart
Russell and Jitendra Malik's RoadWatch system is designed
to provide real-time, intelligent traffic surveillance for
the nation's freeways and cities using pole-mounted cameras
and computer processing.
Photo courtesy CIS
|
In the 1950s,
scientists set out to build computer systems that could, in some
sense or another, think. Artificial intelligence was a grand vision
and an even grander challenge, one that by its very complexity demanded
a disciplinary divide-and-conquer approach. But according to a group
of UC Berkeley researchers, the time has come to reunify fifty years
of fragmented research to solve large-scale societal problems.
"Enough progress has been made that we can now glue the disciplines
back together again" says computer science professor Stuart Russell,
director of the Center.
Launched in August as part of the Center for Information Technology
Research in the Interest of Society (CITRIS), the Center for Intelligent
Systems' (CIS) goal is to foster cross-pollination between researchers
from more than a dozen disciplines including artificial intelligence,
computer vision, robotics, neuroscience, psychology, and computational
statistics. Building on advances in the disparate disciplines, CIS
researchers are collaborating on the design of intelligent systems
for autonomous vehicles, assistive robotics, smart buildings, and
the tracking and prevention of terrorism.
"There are so
many interesting problems that none of us can solve individually,"
Russell says. "But if we learn each other's technical languages,
we can talk about the problems and come up with solutions."
Ron
Fearing's robotic fly, seen here as a mock-up, could someday
be used for search and rescue, monitoring, and reconnaissance.
(Click for larger image.)
Photo courtesy Ron Fearing
|
Previously, Russell
explains, intelligent systems researchers employed completely different
mathematics in their research. For example, artificial intelligence
problems revolved around high-level decision making, necessitating
a fluency in the mathematics of logic and symbolic representation.
Meanwhile, engineers working on the autonomous control of physical
systems immersed themselves in differential equations.
"The result is that nobody could understand each other," Russell
says. "Recently though, all the disciplines started to think about
probability theory and uncertainty. And that drove us toward a common
language."
For example,
one group of CIS faculty researchers and students is focused on
the development of intelligent systems to alleviate traffic jams
in cities. The system uses computer vision to spot cars and determine
how long it takes for a particular vehicle to travel from point
A to point B. Probability comes into play to determine if the computer
is seeing the same vehicle from camera to camera.
"We've
devised algorithms that can solve that problem in real-time for
thousands of vehicles," Russell says. "Now we can look at the control
problems such as how to improve traffic flow on a long-term basis
by changing sign locations and adding or subtracting lanes, or on
a short-term basis by controlling ramp metering lights."
Another CIS priority
is the design of computerized systems to track, predict, and prevent
terrorism. Quite simply, how do you sift through a haystack of spy
reports, email, sightings, telephone conversations, and the mass
amounts of other intelligence to find a single needle of relevant
information?
"You have to connect the dots to know what's going on," Russell
says. "From an intelligent systems perspective, that takes natural
language processing, knowledge representation, and statistical inference."
The CIS is structured to facilitate academic and industry partnership
so the Berkeley researchers can benefit from industrial experience
while industry can rapidly transfer new technologies resulting from
the research. Among other benefits, member companies participate
in research reviews and conferences and are invited to send employees
to work at the Center as visiting researchers. Additional support
may come from grants provided by the US Government's intelligence
community.
"We're all working on the same problem which is the optimal transduction
of process sequences into action," Russell says. "In other words,
that means building computer systems that do the right thing."
The Center for Intelligent
Systems at UC Berkeley
CITRIS
Stuart Russell's
Home Page
Robotic
Fly Gets Its Buzz
Lab Notes is published online by the Public Affairs Office of the UC Berkeley College of Engineering. The Lab Notes mission is to illuminate groundbreaking
research underway today at the College of Engineering that will dramatically change our lives tomorrow.
Editor, Director of Public Affairs: Teresa Moore
Writer, Researcher: David Pescovitz
Designer: Robyn Altman
Subscribe or send comments to the Engineering Public Affairs Office: lab-notes@coe.berkeley.edu.
© 2002 UC Regents.
Updated 8/28/02.
|
