A Cognitive Control Approach for Networked Control Systems with Random Packet Losses

Abstract

This paper presents a novel hybrid cognitive control method inspired by human brain mechanism for networked control systems (NCSs) with random packet dropouts. The random packet losses are assumed to occur in both channels from sensor to controller and from controller to actuator. Different from most of the existing studies, the information entropy is utilized to characterize the uncertainties of the random packet dropouts. A hybrid cognitive control model is designed for NCSs subject to random packet losses. More specifically, the proposed hybrid cognitive control algorithm is constructed by using the Q-learning and backstepping control simultaneously. That is, the Q-learning is employed for the cognitive controller to regulate the backstepping controller and meantime act on the control object. In addition, the information gap is defined based on the information entropy, and used for depicting the uncertainties of the random packet dropouts. Simulation results are performed to validate the designed hybrid cognitive control method, demonstrating that it can eliminate the effects of packet losses and thus performs better than the traditional backstepping or proportional-integralderivative (PID) controllers.