|
TITLE: Model-based Programming of Robust Agile Systems
PRESENTER: Brian C. Williams, http://people.csail.mit.edu/williams/Web%20site/williams.shtml, williams@mit.edu
AFFILIATION:Model-based Embedded and Robotic Systems, Massachusetts Institute of Technology, http://groups.csail.mit.edu/mers/mers.htm
DATE: Thursday 11th September 2008
TIME: 12:00:00
PLACE: CSE Seminar Room, Level 1, K17
ABSTRACT:
Autonomous, self-repairing explorers, such as deep space probes, have
successfully performed complex missions by employing model-based
executives that continuously monitor mission goals, diagnose failures and
plan repairs. These executives employ models encoded as probabilistic
constraint automata, in order to observe and control the hidden states of
the system. These executives have also been incorporated within model-based
programming languages that facilitate the creation of a wide range of fault
adaptive systems, including automobiles and naval ships.
Future explorers, such as autonomous air vehicles and walking robots,
will require far greater agility, in order to robustly achieve their
missions. For example, to avoid falling, a walking robot must quickly
detect a loss of balance, and replan its control trajectory appropriately.
This talk presents recent advances in model-based programming and
execution for agile systems. First, to reason about a system~Rs dynamics,
these executives employ probabilistic constraint automata that are
extended to hybrid discrete/continuous constraints. Second, to robustly
achieve missions, these executives employ planning methods that reason
about continuous, as well as discrete, state changes, and employ
compilation and model-predictive control methods in order to adapt on the
fly. Finally, these executives employ estimation methods for hybrid PHA
that detect subtle failures through active control. Model-based execution
is demonstrated both on a team of cooperative air vehicles and a biped
walking machine.
BIOGRAPHY OF SPEAKER:
Professor Williams leads the Model-based Embedded and Robotic Systems
<http://mers.csail.mit.edu/mers.htm> group, within the Computer Science
and Artificial Intelligence Laboratory
<http://www.csail.mit.edu/index.php>(CSAIL) at the Massachusetts Institute
of Technology <http://web.mit.edu/>. His research concentrates on
model-based autonomy <http://ic-www.arc.nasa.gov/ic/project/mba/index.html>
-- the creation of long-lived systems that explore autonomously, while
commanding, diagnosing and repairing themselves using fast, commonsense
reasoning. Current research focuses on model-based programming and
cooperative robotics: Model-based programming supports goal-directed
programming of robust explorers and everyday devices, by incorporating
model-based deductive capabilities within traditional embedded programming
languages. Cooperative robotics extends model-based autonomy to robotic
networks of cooperating space, air, land and undersea vehicles, on Earth
and on other planets.
Professor Williams received his S.B., S.M and Ph.D. in Computer Science and
Electrical Engineering at MIT, and worked at the Xerox Palo Alto Research
Center and NASA Ames Research Center, prior to joining the faculty at MIT.
He is a pioneer in the fields of qualitative reasoning, model-based
diagnosis and autonomous systems. He received a NASA Space Act Award for
Remote Agent, the first fully autonomous, self-repairing space explorer,
demonstrated onboard the NASA Deep Space One probe in May, 1999. He was a
member of the Tom Young Blue Ribbon Team in 2000, assessing future Mars
missions in light of the Mars Climate Orbiter and Polar Lander incidents,
and is currently a member of the Advisory Council of the NASA Jet
Propulsion Laboratory at Caltech. He has won four best paper prizes for
his research in diagnosis, qualitative algebras, propositional inference
and soft constraints. He is a fellow of AAAI, has served as guest editor
of the Artificial Intelligence Journal
<http://www.elsevier.nlinca/publications/store/5/0/5/6/0/1/> and has been
on the editorial boards of the Journal of Artificial Intelligence Research
<http://www.jair.org/>, and MIT Press.
Host:
Toby Walsh
Seminar Convenor:
Van Hai Ho
|
