Broad Area Colloquium for Artificial Intelligence,
Geometry, Graphics, Robotics and Vision
Programming Machines That Work
Daniel Koditschek
University of Michigan
Monday, May 20th, 2002, 4:45PM
Gates B01 http://robotics.stanford.edu/ba-colloquium/
Abstract
Robotics is a fledgling discipline concerned with programming work: that
is,
specifying and controlling the exchange of energy between a machine and
its
environment. Because our understanding of how to do this is still quite
rudimentary, the best progress in the field has come from a mix of
inspired
building and formal analysis. For more than a decade, my students and I
have pursued
such an agenda, building robots whose controllers drive the coupled
robot-environment state toward a goal set and away from obstacles. The
talk
reviews our progress to date: what sort of "programs" do we know to
build,
with what theoretical guarantees, and with what empirical
success?
Because animals exhibit so many of the capabilities we would wish to
imbue
in our robots, it seems plausible that useful inspiration may be gained
from biology. A second theme of the talk concerns
the benefits we have enjoyed from a close collaboration with
biomechanists
over the last few years.
In particular, our hexapod, RHex, built to embody the essential
biomechanical principles of
animal runners, exhibits mobility superior to any previously documented
autonomous programmable machine. The example of RHex serves well to
illustrate some of the important concepts from biology that hold
significant
promise for the future of robotics, among these being the confluence of
form
and function, the tradeoffs between central and peripheral and between
feedback and feedforward control implementations; and the allure and
present
limitations of "evolutionary" thinking.
About the Speaker
Koditschek received his Ph.D. in 1983 from Yale University in nonlinear
control, and has been applying principles of dynamical systems theory to
intelligent machine design for his entire career. After a number of
years on
the faculty at Yale, he moved to the University of Michigan in 1993,
where
he is presently a Professor of Electrical Engineering and Computer
Science.
Koditschek's archival papers have been published in a broad spectrum of
journals ranging from the Transactions of the American Mathematical
Society
through the Journal of Experimental Biology, with a concentration in
several
of the IEEE and ASME Transactions. Various aspects of this work have
received mention in
general scientific publications such as Scientific American and Science
as
well as in the popular and lay press such as The New York Times and
Discover
Magazine.
Contact: bac-coordinators@cs.stanford.edu
