EM-based distributed estimation in cluttered environment

Abstract

We consider a one-dimensional target estimation problem in a cluttered environment. A distributed estimator based on expectation-maximization (EM) algorithm and uncoded analog transmissions is developed. Through this experiment, one can gain some insights into methodology in the design of distributed estimator in sensor networks. As shown in our simulation, the proposed algorithm achieves significant estimation performance gain over existing distributed estimators under a communication constraint. Modified Cramer-Rao bound is also derived to provide a performance benchmark.