A Design Method of Learning Control Systems

Abstract

Recently there have been several results in the field of learning control or repetitive control. These are intended to control certain mechanical systems such as robot manipulators. But most of them are restricted to specific systems or systems of almost known parameters. This is because of the schemes of the control systems, which are a kind of servocompensator or are constructed with the simple P-D controller.This paper deals with the problem of constructing learning control for unknown systems. The problems stated before, are solved by introducing parameter adaptation schemes.First we describe repetitive control problems as a kind of Attasi-type two-dimensional systems. Then we design model reference adaptive control systems for these two-dimensional systems and construct learning control systems simultaneously. However these are different from those stated before in the point that adaptive parameters are adjusted by use of the overall state variables on each time interval. In this sense the proposed learning control systems are a kind of hybrid adaptive control systems.