ALIO-SAM: Lightweight Adaptation in LiDAR-Inertial Odometry for Mobile Robots in Complex Environments
In robotic operation scenarios, LiDAR-Inertial SLAM systems based on factor graph optimization often lack sufficient adaptability. Common issues include backend optimization latency leading to odometry state divergence, and performance degradation in the scan-to-map matching mechanism due to local map bloat when operating in fixed areas for extended periods. To address these challenges, this paper proposes a lightweight LiDAR-Inertial SLAM framework, ALIO-SAM, tailored for mobile robots. ALIO-SAM adaptively switches between LIO and LO modes according to the backend latency of the backend factor graph optimization to ensure stable system operation. To tackle the issue of varying constraints caused by LiDAR point distances, we devise a distance-weighted strategy to enhance scan-to-map matching accuracy. Furthermore, ALIO-SAM introduces a dynamic keyframe selection strategy based on Scan Context[Formula: see text] for local map construction, preventing map redundancy under specific operational conditions. The proposed method is validated on real-world scenes and the public M2DGR dataset. Experimental results demonstrate that ALIO-SAM achieves positioning accuracy comparable to or exceeding open-source frameworks such as LeGO-LOAM, LIO-SAM, FAST-LIO2, and Light-LOAM, and it simultaneously reduces the computational time of frontend scan-to-map matching.