抄録
There have been many studies on the application of neural networks to servo controllers in order to design flexible controllers which have nonlinear mapping capabilities. Most of them are studies on series controllers which use the neural network output as the object plant input. Only M. Kawato has proposed a parallel controller which uses both the neural network output and the conventional control output as the object plant input. However, his proposed controller is analog rather than digital. The stability of a digital controller is different from that of an analog controller. We should investigate digital controller stability in order to construct a neural network servo controller which uses software.
Therefore, this paper proposes an adaptive type feedforward feedback controller which uses a neural network. The proposed controller is a parallel type controller. It is designed for digital control systems and its stability condition is investigated. Simulated results for a second order plant confirm both its effectiveness and characteristics.
