Autonomous Vehicle Localization on Standard Definition Maps Based on Camera and LiDAR Sensor Fusion

Engineering notes

State-of-the-art approaches that rely on the Global Navigation Satellite System (GNSS) suffer from poor reliability in urban contexts. Our method is validated through a comparison with a state-of-the-art approach.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Localization is a crucial aspect of every autonomous driving vehicle, as determining its position in the navigation space enables the vehicle to safely plan its motion and interact with the environment. State-of-the-art approaches that rely on the Global Navigation Satellite System (GNSS) suffer from poor reliability in urban contexts. This issue can be overcome with Simultaneous Localization And Mapping (SLAM) methods. Generating and maintaining maps of large dimensions using these methods is a high-resource consuming task. This has motivated many to develop localization methods based on map databases provided by third-party sources. This paper presents a localization approach based on the OpenStreetMap (OSM) database. In particular, a local perception map generated from LiDAR and Camera observations is aligned to a graph representing an approximation of the road structure. Our method is validated through a comparison with a state-of-the-art approach.

Links and sources

• Official / arXiv page

Comments