Decentralized Multi-Bernoulli Multitarget Tracking Using Multistatic Doppler-Only Measurements

Abstract

This paper proposes a decentralized multi-Bernoulli filter for multitarget tracking over a network of separately located Doppler sensors, in which each sensor exchanges its received measurements and posterior estimates with partially selected neighbors via one-iteration-only diffusion. The selection of an optimal neighbor subset for each sensor is formulated as a partially observable Markov decision process with the probability ratio of nonexistence to existence of targets being the cost function. In particular, the locally collected measurements and posteriors are handled by the particle-based iterated-corrector multi-Bernoulli filter and Gaussian-mixture-based arithmetic average fusion, respectively. The validity of the proposed filter is verified via computer simulations.