A design of hybrid neural net-based GMVC and its application

In this paper, a design method for a nonlinear system is proposed, using a linear control scheme for the approximated linear model of the controlled object and a NN compensating the modeling error. In this scheme, the linear controller is designed based on the approximated linear model. And the augmented controlled object which consists of the controlled object and the NN behave same as the approximated linear model, after the NN is trained enough. The effectiveness of the proposed method is evaluated on a temperature control of a water bath which has a nonlinear property.