Self-tuning neurocontrol of nonlinear systems using localized polynomial networks with CLI cells

This paper presents a novel self-tuning neurocontrol scheme for nonlinear systems. It consists of the online nonlinear system identification and the online self-tuning controller design based on the certainty equivalence principle. The d-step ahead prediction input-output models of sampled-data nonlinear systems are identified using the localized polynomial networks with competitive lateral inhibitory (CLI) cells. Several self-tuning adaptive neurocontrol laws are then derived based on these models. The global stability of these self-tuning neurocontrol systems is guaranteed. The proposed scheme works for both minimum and non-minimum phase nonlinear systems. Simulation results confirmed the above theory.