Event-Driven Networked Predictive Control

In networked control systems, random transmission delay significantly degrades the control performance and can cause system instability. To address this problem, the method of networked predictive control (NPC) has been proposed, which takes advantage of the feature of the network that a packet of data can be transferred simultaneously. At the controller side, future control sequences for every possible time delay are generated, which are then packed into a single packet and transmitted to the plant side. The plant side receives the packet and chooses the proper control signal based on the time delay measurement. However, this scheme needs an accurate plant mathematical model and the measurement of time delay. In this paper, a new event-driven NPC method is considered where the control signal is selected according to the plant output rather than the time delay measurement. The new method does not need any time delay measurement and can significantly improve the system performance in the presence of model uncertainty. The stability of the system when the method is used is analyzed. To illustrate the improved performance using the proposed method, the results from both simulations and real Internet-based networked experiments for a servo system are presented