Data-Based Approach to Robust Predictive Iterative Learning Control via Admissible Behaviors

This article is dedicated to developing a robust data-based predictive iterative learning control (PILC) framework for linear time-varying (LTV) systems via a behavioral approach. By investigating the properties of the admissible behaviors of LTV systems, an input/output representation is constructed from data, based upon which a data-based trackability criterion is developed for iterative learning control (ILC) systems. Moreover, in the presence of measurement noises, a robust PILC framework is constructed from noisy data through adopting a slack-variable-based strategy. Consequently, even in the absence of model information, ILC systems can achieve robust tracking performance with a faster convergence speed of tracking errors. To validate the effectiveness of the proposed PILC framework, simulation tests are performed on a permanent magnet synchronous motor (PMSM).