Disturbance rejection using error estimation in neural network controller design

Abstract

Disturbance rejection is an important factor in evaluating the performance of a control system. By using error estimations, we expand a virtual area among actual error points in the error space which is composed of runtime errors and their derivatives. Rather than driving our neural network controller (NNC) with actual error signals, we utilize virtual error signals under different expanding parameters. Simulations have successfully shown that out method could resist unexpected disturbance in many cases.