- Start Date:
July 8, 2014
- Event Start Time:
11:30 AM
- Event End Time:
12:30 PM
- Organizers:
Gene Fiorini
- Seminar Series:
REU Seminar
- Presenter(s):
Kostas Bekris - Rutgers University
- Event Location:
DIMACS Seminar room
- Abstract:
Robots are becoming increasingly capable in terms of their mechanical, sensing and computational capabilities. In order to be effective, however, in solving many physical tasks like collision-free motion and manipulation of objects, they still need to address hard algorithmic problems. A key challenge, corresponds to motion planning, which is a prototypically computationally hard problem. It requires the computation of collision-free, feasible paths for moving bodies in continuous spaces. A variety of practical approaches have been developed for solving effectively many planning instances. A popular methodology samples valid configurations of a moving body to incrementally construct and connect the nodes of a graph on which path planning problems are then solved. These methods provide probabilistic instead of deterministic completeness. A recent development has been the identification of the conditions under which these methods also converge asymptotically to optimal paths in continuous configuration spaces. This talk will review the state-of-the-art in algorithmic motion planning and recent contributions in balancing performance guarantees, such as completeness and (near-)optimality, and computational requirements, such as time and memory requirements. Similar tradeoffs in the context of multi-robot motion planning will also be discussed.
