Stochastic Channel Allocation for Networked MPC Without Terminal Conditions

Yushen Long, Lihua Xie

Abstract: In this paper a channel schedule and controller co-design for networked control systems are proposed. The considered system consists of a group of dynamically decoupled nonlinear subsystems whose sensors and actuators are connected with a centralized controller through a digital communication channel. Due to the limited capacity, only one sensor and actuator can communicate with the controller at each time instant. Furthermore, Markovian packed dropout occurs because of the unreliability of the communication channels. We adopt a predictive control framework for controller/scheduler co-design to alleviate the negative effect caused by the limited communication capacity and the packet dropout. Instead of sending a single control value, the controller sends a sequence of predicted control values obtained by MPC without terminal conditions to a selected actuator so that there are control input candidates which can be fed to the subsystem when the actuator cannot receive new control input from the controller. A stochastic schedule is designed for the communication of sensors and actuators and we show that stochastic stability can be achieved under some mild assumptions and properly chosen design parameters. The upper bound of the closed-loop performance of each sub-system is provided in terms of system and controller parameters.