Automated Multi-Target Tracking in Public Traffic in the Presence of Data Association Uncertainty
Andinet Negash Hunde, Beshah Ayalew · 2018
Automated driving systems of emerging and future vehicles need to resolve the behaviors of other traffic participants or targets in order to safely navigate in public traffic. In this paper, a comprehensive multi-target tracking system is outlined that addresses target birth/appearance and death/disappearance processes in the presence of measurement data association uncertainties. The tracking system is based on the joint integrated probabilistic data association filter, which is adapted to specifically include algorithms that handle track initiation and termination, clutter density estimation and track maintenance. The workings of the proposed algorithms are demonstrated via multiple traffic scenario simulations that show how tracks can be initialized and terminated autonomously, and how data association uncertainties can affect tracking performance if not handled correctly.