NONLINEAR CONTROLLER OF ACTIVE MAGNETIC BEARING

NONLINEAR CONTROLLER OF ACTIVE MAGNETIC BEARING

Chapter in a book

Two possible approaches to control of nonlinear systems, particularly active magnetic bearing, are described in this contribution. First approach is nonlinear compensator learned by control error. Compensator is constituted by automatic approximator based on one of the Locally Weighted Learning methods. Second approach is control by auxiliary linearizing controller with automatic parallel compensation. Coefficient (parameter) values come from actions of Continuous Action Reinforcement Learning Automata. Described concepts of control are proved by simulation study.

Two possible approaches to control of nonlinear systems, particularly active magnetic bearing, are described in this contribution. First approach is nonlinear compensator learned by control error. Compensator is constituted by automatic approximator based on one of the Locally Weighted Learning methods. Second approach is control by auxiliary linearizing controller with automatic parallel compensation. Coefficient (parameter) values come from actions of Continuous Action Reinforcement Learning Automata. Described concepts of control are proved by simulation study.

Nonlinear controller, active magnetic bearing, Locally Weighted Learning, Continuous Action Reinforcement Learning Automata

Nonlinear controller, active magnetic bearing, Locally Weighted Learning, Continuous Action Reinforcement Learning Automata

BŘEZINA, T.; TUREK, M.; PULCHART, J.

24.12.2005

Praha

80-214-3144-X

Simulation modeling of mechatronic systems I

Mechatronika

129

136

18