Neural Network and Observer-Based Compensation of Networked Control Systems with Packet Dropout

This paper investigates a new alternative approach to compensate for the effects of packet dropouts (data do not arrive at destination) on a dynamic networked control system (NCS). The approach uses an observer-based control computation using time-delay predicted values as the sampling times of the NCS. To obtain the predicted values of time delays we propose the use a one-step-ahead and multi-step prediction algorithms based on an adaptive time delay neural network (TDNN). The control gains are calculated using optimal control.