A robust scheme to track moving targets in sensor nets using amorphous clustering and Kalman filtering

This paper presents a Kalman-filter-based, distributed processing scheme to track a moving target in a sensor net. There are two major new features about the proposed scheme. First, distributed processing clusters are formed amorphously. That is: no explicit control message exchanges are needed to form a cluster and elect a leader; rather, data packets are routed to a place where leaders are spontaneously generated if they pass certain threshold condition. This leads to great savings in cost associated with cluster and leader management. Second, the proposed scheme loosens the strict requirement in traditional tracking schemes that there exists exactly one cluster leader at a particular step; rather, multiple leaders can exist at the same step. This leads to more robust operation where leaders can fail due to either device or communications links reasons. Further, the creation of multiple leaders is leveraged on the one and same packet flow and no extra cost is incurred. The proposed scheme is simulated in an example tracking application and the effects on error performance of various parameters are studied