High-order iterative learning control : convergence, robustness and applications

Abstract

Several new aspects of Iterative Learning Control (ILC) have been addressed and investigated for a better understanding of the fact that improved control performance can be achieved from system's repetitive operations. Firstly, a high-order ILC scheme is proposed and explained in the iteration number direction. By considering the dynamics along the ILC iteration number direction, the high-order scheme offers additional po-tentials in the improvement of ILC convergence property compared to the conventional first-order scheme which is merely an integral controller. Secondly, the use of current iter-ation tracking error information is shown to be helpful in tuning the tracking error bound and the ILC convergence rate. Thirdly, ILC for uncertain discrete-time nonlinear sys-tems has been studied systematically with the consideration of actuator saturation. The ILC scheme with a feedback controller and the ILC scheme utilizing the current iteration tracking error are investigated respectively and their differences are discussed explicitly. Fourthly, the ILC scheme with an iterative initial state learning method is shown to be effective to remove a commonly used re-initialization assumption in the conventional ILC methods. The unknown desired states can be identified through the ILC process. Fifthly, the terminal ILC is proposed when only the terminal output tracking error is measurable at the end of each run. This new point-to-point control method is applied to a rapid thermal processing chemical vapor deposition (RTPCVD) thickness control problem in wafer fab industry.