Techniques for Effective Contact Modeling in Dynamic Robotic Simulation

Evan Drumwright

Abstract

Many of the most vexing problems for roboticists that attempt to simulate robots performing manipulation tasks are related to contact modeling. Contact modeling determines forces that prevent bodies from interpenetrating, approximates frictional forces, and applies impulses to impacting bodies. As well understood as undergraduate physics makes the problem of contact modeling appear to be, it is a nonsmooth problem that is both computationally hard and difficult to solve robustly. These issues are compounded by difficulties with collision detection. I will present recent work that addresses all of these problems in a tractable and robust manner. Examples from dynamic simulation, including robotic grasping and locomotion, will illustrate our approach.

Bio

Evan Drumwright received B.S. degrees in Mathematics and Computer Science from the University of Memphis in 1999 and the Ph. D. in Computer Science (Robotics) from the University of Southern California in 2007. He is currently visiting faculty in Computer Science at the University of Memphis. Evan's research interests are physical simulation and humanoid robotics; he currently collaborates with Honda Research Institute using the Asimo humanoid platform and UC Merced on a HOAP3 robot. Evan develops the Physsim multibody dynamics simulator (http://physsim.sourceforge.net), which is targeted to simulating manipulator and humanoid robots