Multiple Target Tracking With Uncertain Sensor State Applied To Autonomous Vehicle Data

摘要: In a conventional multitarget tracking (MTT) scenario, the sensor position is assumed known. When MTT sensor, e.g., an automotive radar, mounted to moving vehicle with uncertain state, it becomes necessary relax this assumption and model unknown explicitly. paper, we compare recently proposed filter that models state [1], two versions of track-oriented marginal MeMBer/Poisson (TOMB/P) filter: first does not uncertainty; second approximately by artificially increasing measurement variance. The results, using real data, show in terms performance, can outperform TOMB/P without uncertainty, comparable increased