关键词:移动机器人; 动力学模型; 运动控制; 非完整系统; 增强学习; 策略迭代
Learning controller design for class of two-wheeled mobile robots
ZHANG Hong-yu, XU Xin, ZHANG Peng-cheng, LIU Chun-ming, SONG Jin-ze
(Institute of Automation, College of Mechatronics Engineering & Automation, National University of Defense Technology, Changsha 410073, China)
This paper proposed a novel self-learning path-following control method based on reinforcement learning for a class of two-wheeled mobile robots. The path-following control problem of autonomous vehicles was modelled as a Markov decision process (MDP) and by using the kernel least-squares policy iteration (KLSPI) algorithm, the lateral control performance of the two-wheeled mobile robot could be optimized in a self-learning style. Unlike traditional table-based reinforcement learning (RL) and RL based on neural networks, KLSPI used kernel methods with automatic feature selection and value function approximation in policy evaluation so that better generalization performance and learning efficiency could be obtained. Simulation results show that the proposed method can obtain an optimized path-following control policy only in a few iterations, which will be very practical for real applications. ......