Distributed double threshold spatial detection algorithms in wireless sensor networks

In this paper we propose two alternative event-driven double threshold detection algorithms to be used in decentralized wireless sensor networks. The proposed approach assumes that a sensor may decide about the presence of an event of interest either directly or asking for additional data from nearby nodes. The proposed methods aim at minimizing the network energy consumption associated to the detection process. The problem is formulated associating a cost proportional to the (average) number of nodes involved in the decision. After a first activation phase, initiated by a single node, we examine two alternative approaches: a fixed sample size and a sequential detector. We show that there is a need of including an activation threshold when there is a stringent constraint on the power consumption or when the SNR on each sensor is quite low. We compare the performance of the proposed approaches showing that, also in this double threshold setup, sequential detection algorithms involve smaller average number of sensors to guarantee the same performance metrics.