Data-driven Reinforcement Learning for Linear Quadratic Control of Unknown Systems over Lossy Channels
- Resource Type
- Conference
- Authors
- Zhang, Liangyin; Xiao, Lanqi; Wang, Jiepeng; Chen, Michael Z. Q.
- Source
- 2022 IEEE 5th International Conference on Automation, Electronics and Electrical Engineering (AUTEEE) Automation, Electronics and Electrical Engineering (AUTEEE), 2022 IEEE 5th International Conference on. :191-195 Nov, 2022
- Subject
- Communication, Networking and Broadcast Technologies
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Electrical engineering
Protocols
Q-learning
System dynamics
Optimal control
Packet loss
Data models
networked systems
linear quadratic control
reinforcement learning
packet dropout
- Language
- ISSN
- 2831-4549
This research deals with the linear quadratic (LQ) optimal control and suboptimal control of networked systems with packet losses and unknown system dynamics. Three system models including two kinds of Transmission Control protocol (TCP) and one User Datagram protocol (UDP) are discussed. The LQ control problems corresponding to these cases are solved respectively by the model-free methods. In the infinite time domain optimal regulation problem, the model-free method only uses the input and output data to design the controller. A numerical example illustrates the validity of the theoretical results.