An improved labeled multi-bernoulli filter for bearings-only multi-target tracking

Abstract

The Multi-Bernoulli filter propagates the multitarget posterior probability density forward under the precondition that targets are indistinguishable, whereas distinguishing target identities is important in real tracking systems. The recently proposed Labeled Multi-Bernoulli (LMB) filter addresses the target trajectories and track labels that makes it more suitable for practical applications. For tracking targets with bearing-only measurements, the classical nonlinear filters sometimes are not effective. In this paper, we propose an improved LMB filter which replaces the one Gaussian in the likelihood function with a Gaussian mixture. Simulation results show that the proposed method successfully reduces estimation error with tractable computational cost.