Estimation of Target Location Via Likelihood Approximation in Sensor Networks

A fully decentralized sensor network, without fusion center, is deployed to estimate the position of a target. Taking advantage of the limited communication range of the nodes, and exploiting their (unknown) location inside the surveyed area, the likelihood profile is approximately reconstructed. A distributed ML-like estimator is, therefore, proposed and its asymptotic performance is investigated analytically, while computer experiments assess the behavior of the estimator in nonasymptotic regimes. The differences between one- and two-dimensional scenarios are also discussed.