A Nonlinear Adaptive Observer Based on Neural Networks for Nonlinear Systems including Secure Communication Problems

This paper presents a nonlinear adaptive observer based on multilayer perceptron (MLP) neural networks, which is indeed a nonlinear-in-parameters neural network (NLPNN), for a class of nonlinear systems including chaotic systems in which all MLP weights are tuned on-line. These systems can be decomposed into linear and nonlinear parts. It is assumed that the linear part is known while the nonlinear part is represented as an unknown piece-wise continuous function. The ultimate boundedness of the observer is guaranteed through the Lyapunov stability analysis. These observers are suitable in nature for secure communication problems. As a case study, the proposed neuro-based observer is applied to the Chua’s system in a master-slave synchronization problem. Some numerical simulations are performed. The results are very promising and approve the efficiency and robustness of the proposed robust neuro-based adaptive observer.