Neural Network Augmented Intelligent Iterative Learning Control for a Nonlinear System

An iterative learning controller (ILC) is an online method which exploits the information of past trials to improve the performance of the system. For a system controlled by ILC, the state, error, and ILC time histories for varying operating conditions can be recorded. This paper proposes an offline learning method using a neural network which exploits this dataset to approximate the converged ILC for a nonlinear system. The proposed method provides an approximate ILC for the first iteration based on the data collected thereby achieving a faster convergence. The efficiency of the method is tested for a nonlinear problem and results are presented.