LeGO-LOAM

Classic lightweight, ground-optimized lidar odometry and mapping baseline (IROS 2018)

License

BSD-3-Clause

Overview

Lightweight and Ground-Optimized Lidar Odometry and Mapping — real-time 6-DOF pose estimation and mapping for ground vehicles using a 3D lidar. Introduced at IROS 2018, it exploits the ground plane during segmentation and splits pose optimization into two lighter steps, so the full SLAM pipeline runs in real time on low-power embedded computers.

The pipeline segments each scan into ground and non-ground clusters, extracts planar and edge features from them, and estimates motion in two steps: ground features constrain z, roll and pitch, then the remaining dimensions are solved against edge features. A keyframe-based pose graph with loop closure produces the global map. Ships as a ROS package tuned for a Velodyne VLP-16 on a ground vehicle.

Why it matters

LeGO-LOAM remains one of the canonical lidar SLAM baselines: new lidar odometry papers still benchmark against it, and its ground-optimized segmentation ideas carried directly into LIO-SAM and many production ground-robot stacks.

Problem it solves

Full lidar SLAM pipelines were too heavy for the embedded computers on ground robots. LeGO-LOAM restructures LOAM around ground-plane segmentation, point-cloud clustering and a two-step Levenberg-Marquardt optimization so real-time 6-DOF odometry and mapping fits on Jetson-class hardware.

Key takeaways

• Ground-plane segmentation cuts compute while improving robustness on ground vehicles
• Two-step optimization solves the 6-DOF pose with far fewer iterations than monolithic LOAM
• Runs in real time on embedded ARM boards, not just desktops
• Direct ancestor of LIO-SAM — same author and design lineage
• Battle-tested ROS baseline still used for benchmarking a decade on

Who it is for

• Ground-robot developers needing proven lidar odometry
• SLAM researchers benchmarking new methods
• Students learning lidar SLAM from a readable classic codebase

Limitations

• Tuned for ground vehicles with a horizontally mounted spinning lidar — not suited to handheld, aerial or solid-state setups
• No tight IMU coupling, so aggressive motion degrades odometry — LIO-SAM supersedes it there
• Upstream maintenance is minimal and ROS 1 focused (last activity 2024)