RBF NN-based observer design for a class of chaotic systems

摘要

The study on observers of chaotic systems is weak comparatively, and there exist two main problems: a) no systematic design methods; b) difficult to deal with uncertainties. To solve these problems, a robust observer based on RBF Neural network (NN) is proposed. This design divides the system into linear and nonlinear parts and describes them using the addition of nominal part and uncertain part. The RBF NN estimates the uncertainties, the observer gain is designed to satisfy the Lyapunov equation, and then using robust control method improves the robust performance of observer. The proposed method avoids the strictly positive real (SPR) condition. Building on above theoretical basis, the finite-time robust observer is put forward. The applications on Rossler chaotic system and Rikitake oscillator show these methods are effective.