Tracking control for robot arm using neural network with simultaneous perturbation learning rule

Abstract

We report tracking control for a robot arm using a neuro-controller. We adopt the simultaneous perturbation learning rule for the neuro-controller. The learning rule requires only two values of an error function. The twice operation yields modifying quantities of the weights in the network. Thus the neuro-controller can learn an inverse of robot kinematics. Some simulation results are shown.