Learning-Enhanced PI and Fuzzy-Sliding-Mode Tracking of a Piezo-Actuated System

Abstract

In this study, simple PI and fuzzy-sliding-mode controllers incorporating a memory-based learning control scheme are proposed to resolve tracking problem of a piezo-actuated system. Unlike traditional model-based approaches, which require precise inverse model in order to cancel out undesirable hysteresis effects, the methods proposed here are model-free. The proposed schemes are denoted as the learning-enhanced fuzzy sliding-mode control (LEFSMC) and learning-enhanced PI control (LEPIC) because a memory-based predictor is adopted to compensate for the hysteresis-induced tracking error. The learning scheme predicts the tracking error of the present cycle based on the knowledge of the previous tracking errors. The predicted tracking error is then used to update the control action and to serve as a learning enhancement to the traditional PI and fuzzy sliding-mode controller (FSMC). Experimental investigations focusing on different types of reference inputs show that the proposed control schemes may suppress the tracking error to within 5% full span range (FSR) of the actuator by using the LEFSMC and to within 6% FSR using the LEPIC.