Distributed Tracking for Mobile Sensor Networks with Information-Driven Mobility

In this paper, we address distributed target tracking for mobile sensor networks using the extension of a distributed Kalman filtering (DKF) algorithm introduced by the author in [11]. It is shown that improvement of the quality of tracking by mobile sensors (or agents) leads to the emergence of flocking behavior. We discuss the benefits of a flocking-based mobility model for distributed Kalman filtering over mobile networks. This mobility model uses author\´s flocking algorithm with a natural choice of a moving rendezvous point that is the target itself. As the agents "flock" towards the target, the information value of their sensor measurements improves in time. During this process, smaller flocks merge and form larger flocks and eventually a single flock with a connected topology emerges. This allows the agents to perform cooperative filtering using the DKF algorithm which considerably improves their tracking performance. We show that this flocking algorithm is in fact an information-driven mobility that acts as a cooperative control strategy that enhances the aggregate information value of all sensor measurements. A metric for information value is given that has close connections to Fisher information. Simulation results are provided for a group of UAVs with embedded sensors tracking a mobile target using cooperative filtering.