Distributed Sensor-Informative Tracking of Targets

In this work a distributed tracking technique for multiple non-overlapping targets is developed such that it utilizes only sensors that acquire informative observations about the targets. A framework is designed where norm-one regularized factorization is employed to decompose the sensor data covariance matrix into sparse factors whose support facilitates recovery of the target-informative sensors. Then, extended Kalman filtering recursions are derived to perform target tracking using only the target-informative sensors. Different from existing alternatives, the novel algorithm can determine the informative parts of the network topology without relying on underlying model parameters and target trajectory estimates, can handle multiple non-overlapping targets and is less sensitive to noise. Numerical tests corroborate the effectiveness of the proposed approach.