Effective Computational Methods for High-dimensional Motion Planning

Erion Plaku

Motion planning plays a fundamental role in robotics and has applications in many diverse areas. Motion planning is in general necessary in order for robotic systems to carry out assigned tasks and objectives, such as exploration of unknown environments, autonomous or assisted driving, search-and-rescue operations, and many others. The motion planning problem consists of finding a trajectory for a given system from an initial state to a goal state such that certain conditions are satisfied at each state of the trajectory. As an example, motion planning for a car requires the computation of a trajectory that not only connects the starting and destination places but also avoids collision and respects legal speed limits. Theoretical results provide strong evidence that the motion planning problem especially for high-dimensional robotic systems is computationally challenging.

This talk focuses on novel computational methods developed during my research that have effectively solved motion planning problems with hundreds of dimensions, which previously could not be dealt with in a practical amount of time. In many cases, the computational cost has been reduced by several orders of magnitude.

Biography

Erion Plaku is a Ph.D. candidate in computer science at Rice University. He is studying algorithmic robotics under Prof. Lydia E. Kavraki. His research interests include high-dimensional motion planning for continuous and hybrid systems, large scale distribution of motion planning algorithms, proximity algorithms, and dimensionality reduction. He received the B.S. degree in computer science from the State University of New York, Fredonia, NY, in 2000 and the M.S. degree in computer science from Clarkson University, Potsdam, NY, in 2002.

When: 4pm, November 3, 2006

Where: Fenton Hall, Room 105