Using chaos to improve generalization in smart NN control design

In this paper, a smart NN control scheme is proposed. This scheme is designed such that the current control action can utilize the knowledge that the NN learned from the past control process. A chaotic signal is employed as the reference signal to improve the generalization ability of the NN in the training phase of the scheme, where the complex chaotic signal offers much more information for NN learning thereby significantly improving the efficiency of the NN generalization. Compared with most of the adaptive neural controllers, the smart neural controller (in the operational phase) is a static and low-order controller, and thus needs much less computational resources, and is more feasible in practical implementation. Simulation studies are included to demonstrate the effectiveness of the new control scheme.