An Efficient Implementation of the Generalized Labeled Multi-Bernoulli Filter

摘要

This paper proposes an efficient implementation of the generalized labeledmulti-Bernoulli (GLMB) filter by combining the prediction and update into asingle step. In contrast to an earlier implementation that involves separatetruncations in the prediction and update steps, the proposed implementationrequires only one truncation procedure for each iteration. Furthermore, wepropose an efficient algorithm for truncating the GLMB filtering density basedon Gibbs sampling. The resulting implementation has a linear complexity in thenumber of measurements and quadratic in the number of hypothesized objects.