Iterative learning control - an identification oriented approach

Abstract

Introduces a new type of iterative learning control (ILC) which seeks the desired input in an appropriate finite dimensional input subspace. The method achieves perfect tracking for a class of uncertain systems without using time derivative of tracking error signals in the learning law. Its effectiveness is demonstrated by experiment using a two-link direct drive arm. One of the distinguish features of this method is that it is regarded as an identification oriented method rather than a pure I/O signal based (or model free) one.