Learning Control of Redundant DOF Robots by Optimization of Parameterized Control Space

Abstract

A framework for the learning control of robots based on a parameterized control space is doscussed. Emphasis is put on how to utilize stored motion knowledge. The principles are applied to an example of redundancy resolution for a simple manipulator. Global sub-optimal solutions for feedforward control are achieved using a simple optimization algorithm. A time-integral performance criterion is used.