LOAM (LiDAR Odometry and Mapping)

A classic method that splits lidar SLAM into a high-frequency odometry step and a low-frequency mapping step, plus its lightweight variants.

LOAM is a lidar odometry and mapping method proposed by Ji Zhang and Sanjiv Singh at Carnegie Mellon University at RSS 2014. Its core idea is splitting the problem into two parts running at frequencies about an order of magnitude apart: odometry estimates the lidar’s motion roughly, at high frequency, while mapping registers the point cloud finely into the map at low frequency. It picks out edge points and planar points from each scan based on local curvature and matches only these features, keeping computation light; it achieves low drift with no IMU needed, reaching accuracy close to offline batch methods on the KITTI odometry benchmark. LeGO-LOAM, published by Tixiao Shan and Brendan Englot at IROS in 2018, lightened this for ground vehicles: it segments out ground points before extracting features, solves for the 6-DoF pose in two steps, and adds ICP-based loop closure. The later LIO-SAM then tightly coupled in the IMU.

ExampleLeGO-LOAM’s original configuration targets the Clearpath Jackal ground robot: a horizontally mounted Velodyne VLP-16 lidar plus an optional IMU, outputting 6-DoF pose in real time; the README also warns that its simple ICP loop closure often fails when odometry drift gets too large.

Also called

LeGO-LOAM, A-LOAM

Related

LiDAR SLAM · LIO-SAM · LiDAR-Inertial Odometry · Iterative Closest Point · Loop Closure Detection · FAST-LIO2

Sources

LOAM: Lidar Odometry and Mapping in Real-time (Zhang & Singh, RSS 2014)
GitHub: RobustFieldAutonomyLab/LeGO-LOAM