Neural networks in design of iterative learning control for nonlinear systems

Abstract

An approach to control design for nonlinear system working under repetitive regime is presented. The general iterative learning control scheme is enhanced with a neural network controller to reduce the uncertainty of the model used for the control design. To achieve this goal an effective data-driven technique for training neural controller is developed. In result, in each process trial, both the control performance and process model can be substantially improved. Also, the stability issues of the neural controller are discussed. Finally, as an illustration of the proposed approach the application to nonlinear pneumatic servomechanism is given.