Reinforcement Q-Learning and ILC with Self-Tuning Learning Rate for Contour Following Accuracy Improvement of Biaxial Motion Stage

Abstract—Biaxial motion stages are commonly used in precision motion applications. However, the contour following accuracy of a biaxial motion stage often suffers from system nonlinearities and external disturbances. To deal with the above-mentioned problem, a control scheme consisting of a reinforcement Q-learning controller with a self-tuning learning rate and two iterative learning controllers is proposed in this paper. In particular, the reinforcement Q-learning controller is used to compensate for friction and also cope with the problem of dynamics mismatch between different axes. In addition, one of the two iterative learning controllers is used to suppress periodic external disturbances, while the other one is employed to adjust the learning rate of the reinforcement Q-learning controller. Results of contour following experiments indicate that the proposed approach is feasible.