Predictive iterative learning control

Abstract

In this paper, we develop a predictive iterative learning controller (PILC) for nonlinear discrete systems. First, a neural-network-based technique is used for developing nonlinear dynamic models from empirical data, Secondly, a local linear model is extracted from this neural network model on-line. Thirdly, as a major benefit of the linearization, linear ILC based on a generalized predictive model is presented. Finally, simulations results are used to illustrate that the proposed control can yield a good set-point tracking performance.