Effect of Table Grid at Q-learning Process of Active Magnetic Bearing Controller

conference paper

English

Paper describes numerical experiments when Q-learning with implemented table grid is applied to the active magnetic bearing controller. The aim is to find the smallest grid of table of Q-learning architecture, which provide sufficient control of magnetic bearing. There are used linear and also nonlinear grids of the table. A simple single-mass model of the magnetic bearing is used. Q-learning using found values of the table was proved especially from a speed point of view. Presumably further experiment will lead to table improving learning process.

Q-learning, magnetic bearing controller, magnetic bearing

BŘEZINA, T.

2001

6. 6. 2001

Brno, ČR

80-214-1894-X

Mendel 2001

215

220

6