Student-t Mixture GLMB Filter with Heavy-tailed Noises

Abstract

The generalized labeled multi-Bernoulli (GLMB) filter acts as a prospective solution in multi-target tracking (MTT) applications. Nevertheless, considering the heavy-tailed process together with measurement noises, the emerged noise outliers can seriously deteriorate the tracking performance exhibited by the GLMB filter. In order to solve this challenging issue, the study develops a Student-t mixture GLMB (STM-GLMB) filter, which employs multivariate St models for adapting the heavy-tailed noises (HTNs), deriving the closed-form implementation regarding the GLMB filter for propagating the parameters of the STM models considering the multi-target St distributions. The filter becomes tractable relying on the introduction of approximations. According to simulation results, the STM-GLMB multi-target tracking algorithm is valid and stable in heavy-tailed process and measurement noises.