Controller design using parametric neural networks

Abstract

A neural network (NN) of a more flexible internal structure than usual is used to design a better controller. A parametric NN (PNN) can represent both linear and nonlinear relationships explicitly and simultaneously by setting its parameters appropriately. In many cases we have some information about the system which enable us to build a linear controller for it. But of course this is not enough for treating nonlinear plants. Using PNN we could make a complimentary linearized controller and then, after starting the learning, in an online manner it will be extended to a nonlinear dominant controller.