Speaker:    

张坤，北京航空航天大学
Inviter:      
Title:
    System Control Schemes based on Actor-Critic Learning Algorithms
Language:     Chinese
Time & Venue:
    2022.12.11 19:00-19:30 腾讯会议:627-604-965
Abstract:
    

System optimization and control have become the key design philosophies in modern control system theories, where reinforcement learning (RL) algorithm has drawn considerable attention in recent literature. In this discussion, the relationship between optimal control and RL algorithm is presented and the problems in conventional optimal tracking control (OTC) scheme are analyzed. Firstly, a critic-only architecture for OTC problem is developed in the work, where a plug-n-play event-triggered algorithm is designed. Secondly, the parallel tracking control for Markov systems is also researched, and two algorithms are used to solve the optimal solution with different learning ways. Both the stability of systems and convergence of parameter learning processes are proved by theorems. Finally, the applications in engineering systems demonstrate the effectiveness of these proposed methods.
--
FROM 223.104.102.*