A comparison of methods for the control of redundantly-actuated robotic systems

Abstract

Mechanical hands, legged vehicles and cooperating manipulators are robotic systems containing closed kinematic chains which are typically driven by more actuators than required. As a result, the force distribution existing in these systems cannot be determined simply from the governing rigid-body statics or dynamics equations since these equations are underdetermined. Techniques have been proposed to overcome this obstacle — the most common being to formulate an optimization problem whose solution will be a force distribution which is optimal in a prescribed sense. A second approach which has been suggested is one in which the elasticity of the constituent bodies is considered in order to render the force-distribution problem determinate, in a manner analogous to the techniques typically used in structural mechanics to analyze hyperstatic structures. A third approach would be to deactivate certain actuators in order to reduce the number of unknowns so that the problem becomes determinate. In the present paper, these methods are compared and the first is shown to yield the best results.