Supervisory Control of Chaotic Systems Using Online GA Tuning Neural Networks

In this paper, we present a controller for the supervisory backstepping control of a class of general nonlinear systems using online GA tuning neural networks (GNSB controller). The weights of the neural networks (NNs) approximator employed in the backstepping controller can successfully be turned via an online genetic algorithms (GAs) approach. The genetic algorithm has the capability of directed random search for global optimization. A simplified form of GA (SGA) approach is proposed to accelerate the search speed, and a new fitness function is established by the Lyapunov design method for the requirement of tuning the weights of the NNs online. A supervisory controller is used to guarantee the stability of the close-loop nonlinear system. Examples of Duffing chaotic system controlled by the presented controller are shown to illustrate the effectiveness of the proposed controller.