In context
In 2023, ground unmanned platforms in industrial settings faced persistent challenges in precise localization and environmental perception, especially in areas with occlusions or complex structures. Single-view LiDAR SLAM systems often suffered from accumulated drift and perceptual blind spots, limiting their safe autonomous operation. The integration of aerial views, typically from UAVs or satellites, was emerging as a promising solution to enhance ground robot navigation.
What was reported
Researchers from Beijing University of Technology proposed a laser SLAM system that fuses ground and aerial view information. The system uses an aerial point cloud map, constructed by a UAV flying at 100 m altitude, as prior information. A registration network aligns aerial submaps to ground local maps, and a multi-view factor graph optimization framework integrates aerial priors with ground perception.
Experiments on a 1000 m road at a construction site showed that the proposed system reduced average translation error by 5.87 m and average rotation error by 1.67° compared to classic single-view laser SLAM. The system also filled perceptual blind areas caused by intersections and obstacle occlusions through map fusion.
Key technical innovations include a cluster-boundary-scale-based validation method for inter-frame feature matching, a preprocessing pipeline for ground-air submap registration, and a multi-view factor graph optimization algorithm that incorporates ground-air constraints into backend optimization.
Why it mattered
This work demonstrated that fusing high-altitude aerial priors with ground LiDAR data can significantly improve localization accuracy and perception robustness for autonomous ground vehicles. It offered a practical approach to overcome the limitations of single-view SLAM in complex industrial environments, potentially reducing reliance on costly high-definition maps and enhancing safety in automated logistics and construction operations.
The results show that the proposed method effectively improves the localization accuracy of the ground unmanned platform.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2023-09-16 · “一种基于地空视角信息融合的激光SLAM系统”