The perception of autonomous driving is essential for vehicle localization and route planning. However, relying solely on vehicle sensors makes it challenging to achieve reliable and all-around perception. In this paper, we propose a method for constructing a scene situation map based on vehicle-road coordination. Firstly, we propose a panoramic camera and LiDAR fusion detection method located on the roadside to extract dynamic targets in the road environment. Then, on the vehicle side, we detect and track dynamic targets and associate them with the roadside by using dynamic target trajectory matching. Finally, we perform a graph optimization model between the vehicle, road, and HD map. By mapping the optimized dynamic targets onto a high-precision map, we generate a scene situation map. The effectiveness of the proposed method is validated through benchmark testing and a real-world field test.
Enhancing Autonomous Driving Through Collaborative Perception and Scene Situation Map Construction
2024-09-24
8119383 byte
Conference paper
Electronic Resource
English
Multi-agent Collaborative Perception for Autonomous Driving: Unsettled Aspects
SAE Technical Papers | 2023
|