Researchers at Hefei University of Technology and the University of North Texas have developed a map updating method that significantly accelerates session alignment in outdoor semi-static environments, such as factories and parking lots. The approach addresses the common problem of slow map updating algorithms that cannot run in real time, ensuring the map remains consistent with the environment.
Key takeaways
- The method introduces Gaussian curvature to sparsify point clouds, reducing data size and focusing on key points for registration.
- Sparse voxelized point cloud registration (SVGICP) replaces traditional ICP, improving speed while maintaining accuracy.
- Historical constraints are used to repair factor graphs when loop closure or registration fails, ensuring robust alignment.
- Evaluation on MulRan and LT-ParkingLot datasets shows session alignment frequency reaches 13 Hz, an 80% improvement over the original method, sufficient for real-time operation.
This work builds on the LT-mapper framework but overcomes its limitations by enabling real-time session alignment and handling challenging local scenarios. The method is validated for map updating in classic semi-static environments, offering a practical solution for long-term autonomous robot operation.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-01-13 · “室外半静态环境下基于快速会话对齐的地图更新方法”