An evolutionary optimisation approach to motor learning with first results of an application to robot manipulators

摘要

We review optimisation approaches to the problems of motor control, specifically the ill-posed problem of trajectory planning for a redundant planar manipulator, and evidence for the use of such principles in primate voluntary reaching movements. We then present the preliminary results of some experiments using Evolution Strategies (ES) to optimise reaching movements of a simulated 3 revolute joint (3R) planar manipulator. Empirical comparisons are made using different cost functionals, and qualitatively compared with previous optimisation-based models.