Variational PMB filter via coordinate descent Kullback-Leibler divergence minimisation

TL;DR

This paper introduces a new variational derivation of the V-PMB filter for multi-target tracking, using Kullback-Leibler divergence minimisation on an augmented space, and compares its performance with other PMB filters.

Contribution

It provides a novel derivation of the V-PMB filter based on KL divergence minimisation, maintaining the posterior's PHD and enhancing multi-target tracking accuracy.

Findings

V-PMB filter performs better when targets are close and then separate.

The derivation uses an augmented space including target states and hypotheses.

Comparison shows V-PMB outperforms other PMB variants in certain scenarios.

Abstract

This paper presents a new derivation of the variational Poisson multi-Bernoulli (V-PMB) filter for multi-target estimation proposed in [#Williams15]. The proposed derivation is based on considering an augmented space that includes the set of target states with their track indices and the global hypothesis variable. Then, we show that the V-PMB projection performs a coordinate descent Kullback-Leibler divergence (KLD) minimisation on this augmented space to fit the best possible PMB density to the Poisson multi-Bernoulli mixture (PMBM) posterior. We also show that this V-PMB projection keeps the probability hypothesis density of the posterior. The paper also includes a comparison with the PMBM filter and other PMB filter variants, including a track-oriented Murty-based implementation, a track-oriented loopy belief propagation implementation and a global nearest neighbour implementation, showing the benefits of the V-PMB filter compared to the other PMB filters when targets get in close proximity and then separate.