研究目的
To study a SLAM method for vector-based road structure mapping using multi-beam LiDAR, focusing on the extraction and vectorization of road structures, efficient vector-based matching between frames, and loop closure and optimization based on the pose-graph.
研究成果
The proposed vector-based mapping method for road structures, especially road boundaries, directly generates the vectorized map desired by autonomous driving applications. The method demonstrated superior accuracy and efficiency in matching and concatenation of polylines, with an average global accuracy of 0.466m without GPS aid.
研究不足
The study focuses on a specific road structure, the road boundary, and does not include other road structures like lanes. The method's performance in scenes with less structural features is not as robust.
1：Experimental Design and Method Selection:
The study employs a SLAM method for vector-based road structure mapping using multi-beam LiDAR, focusing on polyline as the primary mapping element.
2：Sample Selection and Data Sources:
Data were captured at Jiading campus of Tongji University using the TiEV autonomous driving platform equipped with Velodyne HDL-64, among other sensors.
3：List of Experimental Equipment and Materials:
Velodyne HDL-64 LiDAR, IBEOlux8, SICK lms511, vision sensors, and RTK GPS+IMU.
4：Experimental Procedures and Operational Workflow:
The process includes road boundary segmentation, vectorization, matching between vector-based local maps, and optimization using pose-graph.
5：Data Analysis Methods:
The accuracy of matching was evaluated based on the differences between the vehicle’s pose calculated by the proposed method and the ground truth.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
