An intelligent PID controller with a neural supervisor

It is difficult to control nonlinear systems using a PID controller with fixed gains. Many types of PID controllers with variable gains have been proposed based on advanced control techniques. In this paper, an advanced PID controller for nonlinear systems is proposed. Since the nonlinear input/output characteristics can be explained using a set of linear input/output characteristics, we consider that a set of linear system self-tuning PID controllers are used through switching based on signals from the control system. The switching is carried by a neural network based supervisor. This supervisor selects the linearized model which is most accurate for the next step. The effectiveness of this method is investigated through two numerical simulations.