Multi-target tracking pipeline for MIMO-FMCW radars based on modified GM-PHD

Multitarget tracking (MTT) with radar is challenging due to the radar's low-resolution noisy clutter-prone observation data. Classical approaches employ data association filters to deal with clutter and interference, but they are highly complex. This paper presents an improved Gaussian mixture probability hypothesis density (GM-PHD) multi-target tracking pipeline. The proposed GM-PHD solution considers (i) radar bounding-box- based measurements, (ii) a clutter model for radar observations, and (iii) robust target identification. Using both simulated and experimental scenarios, these contributions are proven to improve the tracking performance in terms of robustness against noise and clutter, as well as target identification. The significance of the presented MTT scheme lies in its low complexity since it does not need any data association. This makes it applicable for intense high-level applications for which MTT is used.