Abstract
This paper considers the stabilization of networked linear systems with spatially deployed sensors and actuators. Each local agent of sensor/actuator suffers from lack of observability/controllability, but the collective measurements of all sensors are observable and the joint effects of all actuators render controllability of the plant. To compensate observability and controllability inadequacies of individual agent, a recent effort is the distributed scheme by using consensus-based communication with neighboring agents. Considering the fact that inter-agent communications are usually more expensive than local computations by agents, this paper abandons the communication based scheme and adopts another strategy where the plant at each time selects exactly one sensor/actuator in a switching way. Unlike existing sophisticated sensor/actuator scheduling algorithms, the switched scheduling design used in this paper exhibits remarkable simplification since it allows each senso/actuator to be arbitrarily picked up, with only one requirement that there exists contiguous time intervals not exceeding a certain bound such that on each time interval all sensors/actuators are selected without missing, leaving the switching instants and the dwell times of each sensor/actuator totally random. The explicit calculation of this bound is presented by utilizing an averaging method. The method is illustrated with detailed numerical experiments.
