A New Iterative Learning Control Algorithm for Final Error Reduction*

Abstract

Iterative learning control (ILC) is a powerful technique for tracking performance improvement by adding the iteratively updated learning signal to the feedforward loop of a repetitively operated system. To increase the robustness to uncertainties and noises, usually a low-pass filter is needed in the learning law which can result in a final steady-state error as a trade-off. This study proposes a new ILC algorithm to further reduce the final error associated with the traditional ILC design. This new algorithm can be easily applied to most ILC systems. In the proposed learning algorithm, when the learning converges and a final error is obtained, the learning law is updated by adding a compensating term which consists of a compensatory filter and the obtained final error. The to-be-designed compensatory filter introduces no more design trade-offs and uncertainties. By updating the learning law, the reached learning convergence can be broken and the final error is further reduced. The compensatory filter design guideline along with four different designs are provided and analyzed. Numerical studies with the four designs have been conducted, and the results are compared to that with the traditional ILC design. It shows that the proposed learning algorithm is able to further reduce the final error effectively.