非線形システムに対するニューラルネットワークによるモデル規範制御

概要

論文の詳細を見る
This paper addresses the problem of model-reference control of non-linear systems. Model-reference control is an important class of non-linear scheme. A reference model is used to specify the ideal response of the control system to the external command. The controller should perform tracking control to this model. However, for non-linear plant with unknown structures, it may be difficult to ensure perfect model following. Given a reference model and a controller structures, such a gradient rule can be formulated for non-linear plants. However, the lack of a plant model structure is a major object in designing control schemes. In recent years, the artificial neural network (ANN) has come to be an important element in describing non-linear functions. It has been shown that a feed forward multilayered neural network can approximate a continuous function arbitrarily well. In this paper, model reference control of non-linear plants using the neural networks has been considered. Three approaches are studied for the 2-axies polar coordinate robot, the position control of a magnet suspended above an electro magnet and the antenna servo system. It is shown that our new method is only successfully applied to many systems.
久留米工業大学の論文
2006-06-20