Seminar on Reinforcement Learning Approach to Data-Driven Optimal Control of Dynamical Systems with Communication Uncertainties
Date: 8 December 2023, Friday
Time: 4:30 pm
Venue: YEUNG Y5-203, City University of Hong Kong
Zoom Meeting ID: 967 1968 0176
Password: 123456
Speaker: Dr Yi Jiang, City University of Hong Kong
Reinforcement Learning Approach to Data-Driven Optimal Control of Dynamical Systems with Communication Uncertainties
Dr Yi Jiang, City University of Hong Kong
Abstract: Data-driven optimal control problem is a focal issue in systems and control engineering, which have attracted considerable attention from the systems and control communities. Notably, in many practical applications, control systems suffer from various uncertainties from modeling, parameter estimation and the environment. In engineering applications, actuator, sensor, controller, and their communication channels are generally contaminated with noise. Moreover, system models and communication networks are subjected to uncertain constraints such as packet dropouts, transmission delays, data quantization errors, and data rate limitation. All such uncertainties typically render the underlying systems to be uncertain, which in turn adversely affects the equipped online algorithms designed and operated based on the advanced reinforcement learning and adaptive/approximate dynamic programming. In this talk, data-driven optimal control problems for both linear and nonlinear systems in presence of packet dropout are investigated. A modified Algebraic Riccati equation and a Bernoulli model-based Hamilton-Jacobi-Bellman equation are developed with known system dynamics and probability models of packet dropouts. Two reinforcement learning-based policy iteration and value iteration algorithms are further developed to obtain the solutions to them, and their convergence analysis is also provided.
Speaker’s Bio: Dr Yi Jiang received the B.Eng. degree in automation, M.S. and Ph.D. degrees in control theory and control engineering from Northeastern University, Shenyang, Liaoning, China in 2014, 2016 and 2020, respectively. From January to July 2017, he was a Visiting Scholar with the UTA Research Institute, University of Texas at Arlington, TX, USA, and from March 2018 to March 2019, he was a Research Assistant with the University of Alberta, Edmonton, Canada. Currently, he is a Postdoc Fellow with the City University of Hong Kong, China. His research interests include networked control systems, industrial process operational control, reinforcement learning and event-triggered control.
WEBINAR WEBSITE:
http://cccn.ee.cityu.edu.hk/webinar/