Distributed Kalman filter for relative sensing networks

This paper deals with the distributed estimation problem in a relative sensing network. Each node is governed by a homogeneous dynamic model and has the measurements of relative states between itself and its neighbors. A subset of nodes in the network, called anchor nodes, can additionally have the measurements of their own absolute states. The relative sensing network is modeled by a bidirectional graph. Information about the state and covariance is exchanged locally to implement a collaborative estimation scheme. A centralized optimal estimator is constructed and three distributed suboptimal estimators based on the Kalman filtering technique are then designed. The distributed estimators require local communication only and are applicable in large scale systems. Their performances are compared and discussed through simulations.