Adaptive neural network state predictor for nonlinear time-delay systems

A new adaptive nonlinear state predictor (ANSP) is presented for a class of nonlinear systems with input time-delay. High-order neural network (HONN) incorporating with a special identification model is first employed to identify the unknown nonlinear time-delay system. The predicted weight updating law of neural network is calculated based on the identification, which can be used to predict the future system states. With the predicted system states feedback, the resulting controller can compensate the time-delay in the overall close-loop system. Rigorous stability analyses for the identification and state predictor are all provided by means of Lyapunov stability criterion. Experiment results for a temperature control system with large time-delay are included to demonstrate the effectiveness of the proposed scheme.