报告题目:Reinforcement Learning for Stochastic LQ Control
报 告人: Minyue Fu
时 间:2025年2月28日下午15:00
地 点:海山楼A1101地
校内邀请人:隋天举，教授 金沙app(简体中文)版本
Abstract:
Reinforcement learning (Rl) has been an active research topic for model-freecontrol design. For linear systems without a given system model, RL design focuses ordirect learning of the desired controller from available experimental data, bypassingthe need for system identification. Most existing results in the literature are for thedeterministic systems without noises. In this talk, we present preliminary results on theRL design for the stochastic linear quadratic (LQ) control where the system model issubject to a process noise. For model-based design, it is well-known that the presenceof the process noise does not affect the optimal controller, But in the model-free casethe process noise has profound influence on the direct learning. We provide a modifiecRL algorithm for stochastic LO design and demonstrate the effectiveness of thealgorithm via theoretical analysis and simulations.
Biography:
Minyue Fu received his bachelor’s degree in electrical engineering fromthe University of Science and Technology of China, Hefei, China, in 1982, and Masteland Ph.D. degrees in Electrical Engineering from University of Wisconsin-Madison, USAin 1983 and 1987, respectively. From 1987 to 1989, he served as an Assistant Professorin the Department of Electrical and Computer Engineering, Wayne State UniversityUSA. He joined the University of Newcastle, Australia, in 1989 and became a Professolin Electrical Engineering in 2003. From 2023, he became a Chair Professor at SouthernUniversity of Science and Technology, Shenzhen, China. His current research interestsinclude networked control systems, distributed estimation and control, high-precisioncontrol, and reinforcement learning, He has been an Associate Editor for the lEEETransactions on Automatic Control, lEEE Transactions on Siqnal Processing, Automaticaand journal of Optimization and Engineering. He is a Fellow of lEEE, Fellow of IFACFellow of Engineers Australia, and Fellow of Chinese Association of Automation.
