Trajectory tracking control by an adaptive iterative learning control with artificial neural networks

Abstract

An iterative learning control (ILC) is a kind of the control algorithm which is capable of tracking a desired trajectory perfectly in a period of time. The conventional algorithm, however, have some drawbacks where some nominal parameters are required. In this paper, we propose to combine an adaptive control with artificial neural networks (ANNs) and an adaptive iterative learning control algorithm to overcome the problem. In the parameter updating of the ANNs, two cases are compared with respect to their performance: 1) only the weights are updated, and 2) both the weights and the center of radial basis functions are updated . The efficiency of the proposed methods are examined by experiments of a golf-swing robot.