Iterative Learning Control for Sampled-Data Systems: From Theory to Practice

Abstract

This paper aims to present a framework for the design and performance analysis of iterative learning control (ILC) for sampled-data systems. The analysis is presented in both time and frequency domains. Monotonic convergence criteria are derived in both time and frequency domains and coined in ILC designs. In particular, the causes or conditions that lead to the poor transient responses in the time domain are explored and disclosed. Four ILC designs associated with different learning functions and filters are considered, namely, the P-type, D-type, D2-type, and general filters. The criteria for the selection of each type are presented. In addition, a relationship is shown between sampling-time selection and ILC convergence. Theoretical work concludes with a guideline for the ILC designs. Simulation results are shown to support the theoretical analysis in the time and frequency domains. Furthermore, based on the frequency-domain design tools, a successful experimental implementation on an electric piezomotor is demonstrated.