This paper presents the probability hypothesis density (PHD) filter for sets of trajectories. The resulting filter, which is referred to as trajectory probability density filter (TPHD), is capable of estimating trajectories in a principled way without requiring to evaluate all measurement-to-target association hypotheses. As the PHD filter, the TPHD filter is based on recursively obtaining the best Poisson approximation to the multitrajectory filtering density in the sense of minimising the Kullback-Leibler divergence. We also propose a Gaussian mixture implementation of the TPHD recursion, the Gaussian mixture TPHD (GMTPHD), and a computationally efficient implementation, the L-scan GMTPHD, which only updates the PDF of the trajectory states of the last L time steps.
View on arXiv