Low Dimensional Representations for Motion Planning and Control

Speaker: Professor Daniel D. Lee 

Host Faculty: Professor Farshad Khorrami

Abstract

Planning and controlling robots with many degrees of freedom in complex environments is very challenging with real-time computational constraints. I will show how low-dimensional reductions can be used to solve high-dimensional motion trajectory optimizations. Our algorithm learns symmetries in the high-dimensional cost function to find an appropriate low-dimensional value function, in accordance with Noether's theorem in Lagrangian mechanics. This allows us to compute optimal trajectories with complexity less than the dimensionality of the naive configuration space. Applications of these methods will be shown on highly articulated arms and humanoid robots.

About the Speakers

Daniel D. Lee is currently the Evan Thompson Term Chair, Raymond S. Markowitz Faculty Fellow, and Professor in the School of Engineering and Applied Science at the University of Pennsylvania. He received his B.A. in Physics from Harvard University in 1990, and his Ph.D. in Condensed Matter Physics from the Massachusetts Institute of Technology in 1995. Before coming to Penn, he was a researcher at Bell Laboratories, Lucent Technologies, in the Theoretical Physics and Biological Computation departments. He has received the NSF Career award and the University Lindback award for distinguished teaching; he was a fellow of the Hebrew University Institute of Advanced Studies in Jerusalem, an affiliate of the Korea Advanced Institute of Science and Technology, and has helped organize the US-Japan National Academy of Engineering Frontiers of Engineering symposium. At the GRASP Lab and as director of the University Transportation Center at Penn, his group focuses on understanding general computational principles in biological systems, and on applying that knowledge to build autonomous systems.