Robotics Institute: Foundations of Robotics Seminar

Foundations of Robotics Seminar, January 29, 2008
Time and Place | Seminar Abstract | Bio

Geometric Methods for Dynamically Complex Manipulation Planning and Control

Todd Murphey – University of Colorado at Boulder

Time and Place

NSH 1507
Talk 4:30 pm

Abstract

Mechanical contact in grasping and manipulation is typically modeled using complex friction models and some representation of the underlying dynamic system. One typically chooses a friction model and then couples that model with force balance equations arising from rigid body interconnections. However, for purposes of understanding how much information is actually necessary to sense or compute in order to complete a manipulation task, reduced-order modeling can have many benefits. These include decreased computational complexity, increased robustness, and decreased sensing requirements.

I will discuss recent results on geometric modeling of frictional contact and how they relate to the analysis of various types of manipulation that involve mechanical contact. These results also give insight into how to systematically couple the dynamics of the bodies engaged in mechanical contact with the model of contact itself using techniques from algebraic graph theory and differential geometry. In particular, direct variational calculations provide a way to simulate and optimize systems with computational complexity similar to force balance techniques while avoiding numerical instability. Throughout the talk I will use examples including reduced-order grasp planning, vehicle contact-state estimation, and humanoid motion in the form of mechanical marionettes.

Bio:

Todd D. Murphey received the B.S. degree in Mathematics from the University of Arizona in Tucson, AZ, in 1997 and received a Ph.D. in Control and Dynamical Systems from the California Institute of Technology in Pasadena, CA, in 2002. After a postdoctoral position at Northwestern, he worked at the Aerospace Corporation in the Electro-Mechanical Systems Department.

He has been an Assistant Professor with the Department of Electrical and Computer Engineering at the University of Colorado at Boulder since 2004.

He is the recipient of a National Science Foundation NSF CAREER award. His current research interests include robotics, symbolic dynamics, the role of uncertainty in cooperating systems, and friction-dominated mechanical systems.

The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.