Foundations of Robotics Seminar, February 7, 2007
Time and Place | Seminar Abstract
| Speaker Biography | Presentation Slides
| Speaker Appointments
Improved Motion Planning Speed and Safety using
Regions of Inevitable Collision
Nicholas Chan
Matthew Zucker
James Kuffner
NSH 1507
Refreshments 4:15 pm
Talk 4:30 pm
Providing safety
guarantees for autonomous robots operating in environments with moving obstacles
is a difficult problem, particularly for underactuated
systems or systems with drift due to momentum.
The conventional approach to replanning in
dynamic environments typically computes partial plans within the allotted CPU
time and validates explored states through robot-obstacle collision
checks. However, this approach cannot
provide any safety guarantees for the robot beyond the finite planning
horizon.
In this talk, I
will present some new exploratory work involving the approximate computation of
"regions of inevitable collision" for state validation in a replanning framework for dynamic systems, and present
experimental results that demonstrate the effectiveness of this technique in
providing dramatically improved safety for partial plans in the domain of an underactuated dynamic vehicle.
This work was done
as part of Nicholas Chan's MS thesis at CMU.
For
appointments, please contact James Kuffner.
The Robotics
Institute is part of the School of
Computer Science, Carnegie Mellon University.