Low-Drift LiDAR-Inertial Odometry With Reliable Feature Matching for Unmanned Ground Vehicles

LiDAR-inertial odometry (LIO) is a key technique for autonomous navigation of unmanned ground vehicles (UGVs), but LiDAR’s limited vertical field of view (FoV) causes significant vertical pose drift. Existing methods introduce ground observation constraints to mitigate this drift phenomenon, but these are often unreliable. To address this issue, this paper proposes a low-drift LIO method that works without ground constraints. It leverages point cloud position uncertainty to build reliable horizontal and vertical matching features separately, compensating for the vertical FoV limitation. Validated via experiments on diverse datasets, the method significantly reduces pose drift and maintains reliable performance.