Proposal of Betterment Process: A Learning Control Method for Dynamical Systems

摘要

Given a desired output for a dynamical system whose paraneters are unknown, it is not easy in general to find out the input which generates the desired output, To overcome this difficulty, we propose a learning control scheme for such a dynamical system when desired outputs are given over a finite time interval [0, T]. In this control scheme, the input at present operation is modefied by the error at previous operation which is the difference between the desired output and the system response. Repeating these operations, the input which can eventually generate the desired output is constructed if some conditions for the dynamical system and the modification for the input are satisfied. It is theoretically shown that this learning control scheme can be applied to a class of nonlinear multi-input and multi-output systems. Also this scheme is practically applied to a robot manipulator which has three degrees of freedom, and its effectiveness is assured by several experimental results.