A real-time tracking controller for piezoelectric actuators based on reinforcement learning and inverse compensation

作     者：Qin, Shijie Cheng, Long 

作者机构：Chinese Acad Sci Inst Automat State Key Lab Control & Management Complex Syst Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Artificial Intelligence Beijing 100190 Peoples R China 

出 版 物：《SUSTAINABLE CITIES AND SOCIETY》 (Sustainable Cities Soc.)

年 卷 期：2021年第69卷

页      面：102822-102822页

核心收录：

学科分类：0820[工学-石油与天然气工程] 082803[工学-农业生物环境与能源工程] 08[工学] 0828[工学-农业工程] 0814[工学-土木工程] 0833[工学-城乡规划学] 

基　　金：National Natural Science Foundation of China [62025307, U1913209, 61873268] Beijing Municipal Natural Science Foundation, China [JQ19020] 

主　　题：Piezoelectric actuators Real-time tracking control Reinforcement learning Adaptive dynamic programming Hysteresis compensation 

摘      要：Nanotechnology is a promising technology and has been widely applied for sustainable smart cities. As the fundamental devices for nanotechnology, piezoelectric actuators (PEAs) have gained wide attention in precision manufacturing because of the advantages of rapid response, large mechanical force and high resolution. However, the inherent nonlinearities of PEAs hinder wide applications for nano-positioning and high-precision manipulation. To eliminate these nonlinearities, various control methods have been proposed, while the optimal control of PEAs is considered rarely. Inspired by the reinforcement learning, adaptive dynamic programming (ADP) is proposed to solve the optimal tracking control problem of PEAs. In this paper, a controller based on reinforcement learning and inverse compensation is designed for the tracking control of PEAs. The experiments on the PEA platform are designed to verify the effectiveness of the proposed method. Comparisons with some representative controllers have demonstrated that the proposed controller has a better control performance.