Accurate map-relative localization is crucial when dealing with navigation or other driver assistance tasks. Global Navigation Satellite Systems (GNSS), such as GPS, GLONASS or Galileo, meet the requirements demanded from road-level navigation systems, but with lane-level assistance being the next step, more accurate localization technologies are needed. A lot of effort has been put in improving localization relative to lanes in scenarios with simple road geometry, whereas more complex scenarios such as intersections have received less attention, even though being a key factor in navigational tasks. In this paper, we focus on self-localization in the vicinity of intersections by using the trajectories of other cars as spatial cue. The geometry of the considered intersections as well as the ideal routes of cars traversing the intersection are derived from enhancements of freely available and crowdsourced map data. The trajectory of each car is compared to possible intersection routes in order to find the route which most likely corresponds to the actual trajectory. Subsequently, a geometrical transformation is computed to align the trajectory to the corresponding route and to estimate the position of the ego-vehicle with respect to the intersection. The feasibility of the approach is demonstrated on multiple scenes from the KITTI Benchmark Dataset [1]. We show that the presented approach is suitable for self-localization in intersection scenarios.
Behavior-based relative self-localization in intersection scenarios
2017-10-01
389020 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Straight feature based self-localization for urban scenarios
Tema Archiv | 2010
|Determination of functional scenarios for intersection collisions
Elsevier | 2023
|British Library Conference Proceedings | 2022
|SAE Technical Papers | 2022
|