Abstract :

We consider the problem of binary distributed detection when the
target position is unknown in the context of largescale, dense
sensor networks. We propose to divide the area where the target
could be present into smaller ones, performing a loglikelihood
ratio fusion rule in each one. We derive the Bayesian and NP fusion
rules under using a model of probability of detection that makes no
assumptions on the local decision rule. The performances of both
tests is analyzed using large deviation bounds on the error
probability and a parametric approximation to the probability of
detection function. The main conclusions of the analysis of these
bounds are that, for designing efficient tests in terms of energy
consumption, 1) the exploration area for each test must cover the
area in which the target could be present extended by a distance
that is less or equal to the range of the local sensors, depending
on the type (Bayes or NP) test, and, 2) the pattern for dividing the
large area into smaller ones is the area inside the range of a local
sensor.

