PMBM-Based SLAM Filters in 5G mmWave Vehicular Networks

Radio-based vehicular simultaneous localization and mapping (SLAM) aims to localize vehicles while the landmarks in environment. We propose a sequence of three Poisson multi-Bernoulli mixture (PMBM) based SLAM filters, which handle entire problem theoretically optimal manner. The complexity proposed filters is progressively reduced sustaining high accuracy by deriving density approximation with marginalization nuisance parameters (either vehicle state or data association). Firstly, PMBM filter serves as foundation, for we provide first complete description on Rao-Blackwellized particle filter. Secondly, (PMB) standard reduction from PMB, but involves novel interpretation auxiliary variables relation Bethe free energy. Finally, using same variable argument, derive marginalized PMB filter, avoids particles instead implemented low-complexity cubature Kalman evaluate comparison probability hypothesis (PHD) 5G mmWave networks show computation-performance trade-off between them.