Neural Network-Based AILC for Nonlinear Discrete-Time System with Iteration Varying Initial Error and Reference Trajectory

Abstract

Based on the neural network approximator, an online adaptive iterative learning control (AILC) algorithm is proposed for a class of nonlinear discrete-time system. Here, the random iteration initial error and trajectory reference are considered. The nonlinear system is transformed to a predictor form, and the desired control signal is thus achieved by implicit function theorem. A neural network using only a norm adaptation law is utilized to approximate this desired control signal iteratively. By using Lyapunov analysis, it is proven that all the system signals are bounded and the system tracking error converges to a neighborhood around zero as iteration number goes to infinity. In contrast to the existing AILC results, the most advantage of the proposed neural network-based AILC is that the number of the adjustable parameters is highly reduced.