// the take
LeGO-LOAM is a ROS1 package for real-time lidar odometry and mapping on ground vehicles, built around a horizontally-mounted Velodyne VLP-16 and an optional 9-DOF IMU. It's aimed at robotics/SLAM researchers and UGV builders who need 6D pose estimation without the compute cost of full LOAM, and it comes out of a cited IROS 2018 paper rather than a weekend project.
Ground-plane segmentation before feature extraction cuts the point cloud down before the expensive optimization step, which is the actual reason it's lighter than LOAM. Two-stage Levenberg-Marquardt optimization plus GTSAM/iSAM for mapping is a solid, well-understood backend. The README gives concrete angular-resolution parameters for adapting to other lidar models (HDL-32e, VLP-32C) instead of leaving that as an exercise. There's a real dataset (Stevens VLP-16, 20K+ scans with loop closures) to validate against, not just a toy bag file.
Loop closure is admittedly a naive ICP method that the authors say fails once odometry drift gets large - they point you to a separate repo (SC-LeGO-LOAM) for anything production-grade. It's locked to ROS1 (indigo/kinetic/melodic) and an old GTSAM 4.0.0-alpha2 release that's a pain to build on anything current; there's no ROS2 port here. The algorithm hard-assumes a horizontal lidar with a ground plane in every scan, so it doesn't port to handheld, aerial, or non-ground-vehicle rigs without rewriting the projection logic yourself. Last push was August 2024 but the README itself directs users to LIO-SAM as the actual successor, so this is effectively the authors' own legacy project rather than something under active development.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →