A Design of Learning Control System for Linear Multivariable Systems

摘要

Systems, which are just like industrial robots, must track the same predetermined trajectory precisely and as fast as possible. But it is very difficult for the systems to track the trajectory completely because of the dynamics of the servo mechanism and uncertainty of the parameters. Recently, in the case, where the same motion is repeated, the method to determine the input reducing the error between the desired and actual outputs in the new trial has been studied by many researchers. But those conventional methods have the limitation of class of applicable systems or the property that the final error cannot be avoided.In this paper we propose a new method that is derived from the method of steepest descent in Hilbert spaces. As an integral norm is taken as a criterion in this algorithm, mapy advantages can be achieved and the modification of a criterion becomes easy. With our method the synthesis of the compensator is consistent and the determination of parameters for that is easier. So it can be considered that this method is applicable to wider industrial problems effectively.