A decentralized tracking technique of a target in clutter is discussed. The coordinator must reconstruct the trajectory of a moving object based on the return histories of N geographically distributed sensors. The coordinator can only access the (local) statistics produced by local processing of the sensors' return histories, not the sensors' returns themselves. The local processing models may be different from the coordinator's models of the distributed observation dynamics as long as these local models meet algebraic constraints. To account for the uncertainties in the origin of the sensors' returns, the probabilistic data association technique is used.
展开▼