Reinforcement learning for adaptive optimal control of continuous-time linear periodic systems
Reinforcement learning for adaptive optimal control of continuous-time linear periodic systems
复制标题
DOI：
10.1016/j.automatica.2020.109035
复制
发表时间：
2020-08
期刊：
影响因子：
--
通讯作者：
Bo Pang;Zhong-Ping Jiang;I. Mareels
中科院分区：
文献类型：
--
作者：
Bo Pang;Zhong-Ping Jiang;I. Mareels
研究了连续时间线性周期（CTLP）系统的无限时域自适应最优控制问题。通过对CTLP系统的策略迭代（PI），导出了基于策略和基于策略的自适应动态规划（ADP）算法，使得在不需要系统动力学精确知识的情况下，也能求解最优控制问题。从初始稳定控制器开始，所提出的基于PI的ADP算法在温和的条件下收敛到最优解。应用于有耗Mathieu方程的自适应最优控制，证明了所提出的基于学习的自适应最优控制算法的有效性。
This paper studies the infinite-horizon adaptive optimal control of continuous-time linear periodic (CTLP) systems, using reinforcement learning techniques. By means of policy iteration (PI) for CTLP systems, both on-policy and off-policy adaptive dynamic programming (ADP) algorithms are derived, such that the solution of the optimal control problem can be found without the exact knowledge of the system dynamics. Starting with initial stabilizing controllers, the proposed PI-based ADP algorithms converge to the optimal solutions under mild conditions. Application to the adaptive optimal control of the lossy Mathieu equation demonstrates the efficacy of the proposed learning-based adaptive optimal control algorithm.