Likelihood Consensus 2.0: Reducing Interagent Communication in Distributed Bayesian Target Tracking

Abstract

We propose a communication-efficient scheme for distributed Bayesian target tracking (distributed particle filtering) in possibly nonlinear and non-Gaussian state-space models. The scheme is a sparsity-promoting evolution of the likelihood consensus (LC) that uses the orthogonal matching pursuit (OMP), a B-spline dictionary, a distributed adaptive determination of the relevant state-space region, and an efficient binary representation of the LC expansion coefficients. Our simulation results show that a reduction of interagent communication by a factor of about 190 can be obtained without compromising the tracking performance.