Gossip-Based Distributed Tracking in Networks of Heterogeneous Agents

We consider the distributed tracking problem in networks of heterogeneous agents with limited sensing and communication ranges. A gossip-based distributed Kalman filter (GDKF) is proposed, where an average consensus on predictions of different agents is achieved by randomized, asynchronous gossip algorithms in a totally distributed way. The error dynamics of GDKF is proved to be a globally asymptotically stable system and the error reduction rate is provided. To demonstrate the improved performance of GDKF, we compare it with an alternative distributed estimation strategy termed Kalman-Consensus Filter (KCF) by implementing them to track a maneuvering target collectively with heterogeneous agents.