Stochastic Channel Allocation for Nonlinear Systems with Markovian Packet Dropout

This paper addresses a channel scheduling problem for group of dynamically decoupled nonlinear subsystems with actuators connected through digital communication channels and controlled by a centralized controller. Due to the limited communication capacity, only one channel can be activated and hence there is only one pair of sensor and actuator can communicate with the controller at each time instant. In addition, the communication channels are not reliable so Markovian packed dropout is introduced. A predictive control framework is adopted for controller/scheduler co-design to alleviate the performance loss caused by the limited communication capacity. Instead of sending a single control value, the controller sends a sequence of predicted control values to a selected actuator so that there are control input candidates which can be fed to the subsystem when the actuator does not communicate with the controller. A stochastic algorithm is proposed to schedule the usage of the communication medium and sufficient conditions on stochastic stability are given under some mild assumptions.

本文针对一组由数字通信信道连接执行器、并由集中式控制器管控的动态解耦非线性子系统的信道调度问题展开研究。由于通信容量受限，同一时刻仅能激活一条信道，因此每一时间步仅有一组传感器与执行器可与控制器进行通信。此外，通信信道并不可靠，故引入马尔可夫数据包丢包（Markovian packet dropout）模型。本文采用预测控制框架开展控制器与调度器的协同设计，以缓解通信容量受限所导致的性能损失。相较于仅发送单一控制值，控制器会向选中的执行器发送一系列预测得到的控制序列，如此一来，当执行器暂未与控制器建立通信时，便可将备选控制输入用于该子系统。本文提出一种用于调度通信介质使用的随机算法，并在若干弱假设条件下给出了随机稳定性的充分条件。