Bayesian controller versus traditional controllers 1

Adaptive control is a field with a long tradition sine the early 1950’s. Despite the fact that Bayesian networks offer attractive properties, proven in other domains like data mining, they are seldom used in adaptive control. This paper develops a new type of controller, based on Bayesian networks. It is shown that controllers, trained with impulse and sinus response, shows nearly the same performance as the analytically calculated Dead-Beat controller.