부산대학교 도서관

Classical PID controllers usually rely on some prior knowledge to manually adjust the gains of the controller and determine them. However, when the mobile robot works in a complex and changeable environment, the fixed PID gains may be difficult to meet the needs of the robot trajectory tracking accuracy. Therefore, this paper proposes a Q-learning-based adaptive PID trajectory tracking algorithm. Firstly, we construct a trajectory tracking Q-PID controller based on the error model of mobile robot. Then, the Q-learning algorithm is used to adaptively adjust the gains of the PID controller online. Meanwhile, the incremental active learning exploration method is used to improve learning efficiency and adaptability of agent. Finally, we use simulation experiments to verify the high performance of our algorithm.