Lane-level road maps are necessary for Advanced Driving Assistance Systems (ADAS) and autonomous vehicles, providing prior contextual information on the road network. At this level of detail, map-matching becomes complex: much ambiguity can arise due to the high number of likely hypotheses (especially in dense urban areas) and due to the varying performance of GNSS positioning systems. This paper presents a method measuring the confidence in the information provided by the positioning system using a metric called Map Aided Horizontal Uncertainty Level (MA-HUL). It relies on a map-matching algorithm designed to keep all likely hypotheses that are coherent with the vehicle odometry. Additionally, exteroceptive information is compared to the map data to remove ambiguities (i.e. lane marking detection from a camera). Some hypotheses are then trimmed if a mismatch is detected with sufficient confidence, keeping their number to a minimum. The use of a reference frame linked to the road is also studied to represent uncertainty independently in the along-track and cross-track axes. The performance of the proposed approach is illustrated using data recorded from an experimental vehicle operating on public roads.
Estimating localization uncertainty using multi-hypothesis map-matching on high-definition road maps
2017-10-01
300784 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Road Tracking for Multi-Hypothesis Localization on Navigable Maps
British Library Conference Proceedings | 2008
|HIGH-DEFINITION MAPS AND LOCALIZATION FOR ROAD VEHICLES
Europäisches Patentamt | 2022
|High-definition maps and localization for road vehicles
Europäisches Patentamt | 2024
|