GIT-GVU-00-23

Robot Catching

M. Riley
C.G. Atkeson

Our focus is on creating interesting, ralistic behaviors for humanoid robots and virtual charaters. Among the methods we use to create human-like movements are a template-based approach, where example movements from humans are stored for later use, and a dynamical systems approach where new movements are generated from motion primitives as they are needed. Here we discuss a dynamical systems approach for implementing ball catching for a 30-degree-of-freedom humanoid robot. In our catching task, we generate ball-hand impact predictions and human-like motion trajectories to move the hand to the impact position. We also discuss what is necessary to use a template-based approach to generate behaviors.

Humanoid robot, catching, primitives, trajectory generation

PDF

Postscript