Robust and Flexible Tracking of Vehicles Exploiting Soft Map-Matching and Data Fusion

摘要

Accurate positioning of vehicles and pedestrians is crucial for enhancing road safety. In this paper, we propose and compare two implementations based on Unscented Kalman Filter (UKF) and Particle Filter (PF) to perform trajectory estimation with sensor fusion. For the latter, a novel soft map-matching technique is applied on top of a PF. The main benefit of our method is the possibility of detecting reliably critical situations, like vehicles skidding off the road. Moreover, we can reduce the positioning error by 45% w.r.t. prior art approaches. Our solution can be implemented as a cloud service in the 5G mobile radio network.