Multiple Model Poisson Multi-Bernoulli Mixture Filter for Maneuvering Targets
This work addresses the problem of tracking multiple maneuvering targets for researchers in multi-target tracking, but it is an incremental extension of the existing PMBM filter to multiple motion models.
The authors propose a multiple model Poisson multi-Bernoulli mixture (MM-PMBM) filter for tracking maneuvering targets with randomly time-varying dynamics. The proposed filter outperforms the MM-MB filter in tracking accuracy, especially under low detection probability scenarios.
The Poisson multi-Bernoulli mixture (PMBM) filter is conjugate prior composed of the union of a Poisson point process (PPP) and a multi-Bernoulli mixture (MBM). In this paper, a new PMBM filter for tracking multiple targets with randomly time-varying dynamics under multiple model (MM) is considered. The proposed MM-PMBM filter uses extends the single-model PMBM filter recursion to multiple motion models by using the jump Markov system (JMS). The performance of the proposed algorithm is examined and compared with the MM-MB filter. The simulation results demonstrate that the proposed MM-PMBM filter outperforms the MM-MB filter in terms of the tracking accuracy, including the target states and cardinality, especially for the scenerio with low detection probability. Moreover, the comparisons for the variations of detection probability and standard derivation of measurement noise are also tested via simulation experiments.