Event boundary detection and gathering in wireless sensor networks

Given a large number of sensor nodes distributed randomly over a 2D region, we address the problem of estimating the boundary of an area affected by an event. The boundaries are often irregular shaped and accurate estimation of such boundaries requires complex computation and data structures which the tiny inexpensive sensor nodes, in general, can not support. In this paper, given a random distribution of homogeneous sensor nodes, in case of an event, the affected nodes execute a simple O(d) distributed algorithm based on local information only, that can identify a small set of boundary nodes, where d is the maximum node degree in the network. Next, a nearly load balanced shortest path routing algorithm based on farthest-first technique is proposed for gathering the information of the boundary nodes to the sink node with low latency. Simulation results show that compared to the earlier algorithms, the proposed algorithm with less neighborhood information and simpler computing results better estimation of the covered area and gather the information at the sink node with nearly minimum latency.