Controller/Critic Neurocontrol Design Method Based on Neighboring Optimal Control
Dagli, Cihan H., 1949- and Akay, M. F. and Chen, C. L. P. and Fernandez, B.R. and Ghosh, J.
The reinforcement learning approach to intelligent control decomposes the control task into controller and critic subtasks. The controller provides control actions based on observed plant states and the critic assesses control actions. The critic assessment may then be used to adjust controller structure to improve long-term control system performance. We demonstrate a design method using optimal control theory for controller design and neighboring optimal control for critic design. The resulting control system is capable of robust, near optimal control of nonlinear, time-varying plants.
J. S. Dalton and S. N. Balakrishnan, "Controller/Critic Neurocontrol Design Method Based on Neighboring Optimal Control," Intelligent Engineering Systems Through Artificial Neural Networks, American Society of Mechanical Engineers (ASME), Jan 1995.
Intelligent Engineering Systems Through Artificial Neural Networks (1995, St.Louis, MO, USA)
Mechanical and Aerospace Engineering
Article - Conference proceedings
© 1995 American Society of Mechanical Engineers (ASME), All rights reserved.