Learning of inverse-dynamics for SCARA robot

Abstract

In this paper, we describe a positioning control for a SCARA robot using a recurrent neural network. The simultaneous perturbation optimization method is used for the learning rule of the recurrent neural network. Then the recurrent neural network learns inverse dynamics of the SCARA robot. We present details of the control scheme using the simultaneous perturbation. Moreover, we consider an example for two target positions using an actual SCARA robot. The result is shown.