Particle filtering for lane-level map-matching at road bifurcations

Abstract

Estimating the map-matched position of a vehicle on a digital map is a key technology for modern navigation and advanced driver assistance systems (ADAS). Inaccurate sensor measurements and a simplified road network representation cause uncertainties in the map-matched position. This paper describes a particle filter method for finding a robust and stable solution for the map-matching problem at the lane-level in ambiguous situations. The proposed approach fuses low-cost sensor data including camera detections of the lane markings with commercial map data. A new spatio-temporal filtering algorithm is introduced for estimating the hypotheses for the map-matched positions together with a confidence measure from the set of weighted particles. The results of tests at road bifurcations indicate that this method produces stable decisions for the correct road segment. It was found that the integration of lane based features improves the performance of the map-matcher at the road level.