Q-Learning for Finite-Horizon H∞ Tracking Control of Unknown Linear Discrete-Time Systems

Mingxiang Liu1; Qianqian Cai1; Wei Meng1; Dandan Li1; Minyue Fu2; Zhimin Han3

10.23919/CCC58697.2023.10241116

2023

1934-1768

979-8-3503-4259-8

2023 42nd Chinese Control Conference (CCC)

2023-July

2353-2357

24-26 July 2023

Tianjin, China

In this paper, we present a Q-learning algorithm for finite-horizon H∞ tracking control of unknown linear discrete-time systems. Finite horizon control is challenging due to its correspondence with a time-varying Riccati equation. The proposed algorithm can determine the solution of the time-varying Riccati equation with unknown system dynamics. We first establish the time-varying Q-function, then propose a corresponding Q-learning algorithm and prove its convergence. Finally, the proposed algorithm is proved to be effective through numerical examples.

Finite-horizon zero-sum games tracking control Q-learning

其他

EI

20234515016057

Digital control systems ; Discrete time control systems ; Game theory ; Learning algorithms ; Navigation ; Reinforcement learning ; Time varying control systems

Artificial Intelligence:723.4 ; Machine Learning:723.4.2 ; Control Systems:731.1 ; Calculus:921.2 ; Probability Theory:922.1

IEEE

1.School of Automation, Guangdong University of Technology, Guangzhou, China
2.School of System Design and Intelligent Manufacturing, Southern University of Science and Technology, Shenzhen, China
3.School of Automation, Hangzhou Dianzi University, Hangzhou, China

Mingxiang Liu,Qianqian Cai,Wei Meng,et al. Q-Learning for Finite-Horizon H∞ Tracking Control of Unknown Linear Discrete-Time Systems[C],2023:2353-2357.