|
|
Probabilistic Adaptive Computing
Overview
On October 8, 2005, a robotic vehicle designed by a Stanford University team
stood up to the challenge posed by DARPA, the US defence funding agency and
won a prize of two million US dollars. What did it do? It drove autonomously,
without any human control, for 132 miles (211 kilometers) over dusty unpaved
roads, mountain passes, and flat lake beds (see
video on PBS Nova and
New York Times story). On that day, the history of
driving turned to a new page.
Much closer to home in Singapore, a similar challenge has been posed to
build an autonomous robot that can navigate in outdoor and indoor
environments, clear obstacles, climb stairs, and search an HDB building.
To succeed in these challenges, robots must have the capabilities of
acting intelligently despite imperfect sensor information, e.g.,
noisy camera images, blocked GPS signals, etc. Our goal is to develop
the algorithmic foundations and practical technologies that enable
robots to operate reliably under such uncertainty. We will develop
probabilistic representations and associated modeling and learning
methods for robots to perceive, abstract, and learn in the physical
world. We will also develop probabilistic reasoning methods to support
effective adaptation in an automatic and systematic way. We will focus
on specific tasks, such as target tracking and activity recognition.
People
Home
»
Research
|
