INTERNAL MODEL CONTROL USING NEURAL NETWORKS: APPLICATION TO SCARA ROBOT

Abstract

The paper describes an internal model control scheme, in which both forward model and inverse model are provided by neural networks. The controller is applied to the robot SCARA which is a multi-input multi-output nonlinear system. To improve the quality of the inverse model, a two-step procedure is applied: general training and specialized training. Simulation results show that the control system has good performance and robustness. Keyword: Internal Model Control, Neural Networks, Forward model identification, Inverse model identification, Multi-input Multi-output system.