Neuro-adaptive motion controller with velocity observer for operational space formulation

Abstract

An operational space controller that employs a three-layer neural network (NN) adaptive controller with the velocity observer is presented in this paper. This incorporates the versatility of NN based adaptive controller with the performance and effective formulation of the operational space framework, which accommodates unified force/motion control as well as highly redundant mechanisms postures. In this paper, it is shown that the trajectory tracking errors, the estimation tracking errors, and the NN weight errors are bounded even when the actual and accurate velocity feedbacks are not available, such as often the case in a physical robot. Consequently, the controller with velocity observer..