Distributed coestimation in heterogeneous sensor networks

Abstract

We propose a system-theoretical dynamic information fusion framework for heterogeneous sensor networks, where a sensor network with nonidentical node information roles and nonidentical node modalities is considered. Nonidentical node information roles allow nodes to be either active or passive in the sense that active nodes receive observations from a process, whereas passive nodes do not receive any information. Active and passive roles of nodes can be fixed or varying in time. Nonidentical node modalities allow active nodes to receive different classes of measurements from the process. For this class of sensor networks, we propose a distributed input and state coestimation architecture, where the time evolution of input and state updates of each node both depend on the local input and state information exchanges. Using Lyapunov theory and linear matrix inequalities, we establish stability and performance guarantees of the sensor network executing the proposed distributed coestimation architecture under local sufficient conditions.