A fast implementation of the generalized labeled multi-Bernoulli filter with joint prediction and update

Abstract

This paper proposes a new implementation for the delta generalized labeled multi-Bernoulli (d-GLMB) filter by combining prediction and update into a single step. In contrast to the original implementation which requires different truncation procedures for each component in the prediction and update, the joint strategy involves only one truncation per component in the filtering density, thus drastically reduces the number of computations. Performance comparison with the original implementation is presented through numerical studies.