LiDAR-Inertial Odometry

Fuses lidar point clouds with IMU readings to estimate a robot’s pose in real time while building a point-cloud map.

LiDAR-inertial odometry fuses a lidar with an IMU (inertial measurement unit) to estimate its own motion, forming the core of lidar SLAM. The two are complementary: a single point-cloud scan takes a stretch of time to capture (commonly 10 Hz, or 0.1 second), and the robot’s motion during that time distorts the point cloud — the high-frequency IMU can estimate this motion to correct the distortion (deskewing) and provide an initial guess for registration; registering the point cloud against the map, in turn, corrects the drift that accumulates from integrating the IMU. Systems are categorized as loosely or tightly coupled by how they fuse the two, with tightly coupled now the mainstream approach. Landmark systems include Tixiao Shan and colleagues’ LIO-SAM at IROS 2020 (factor-graph optimization) and the University of Hong Kong MARS Lab’s FAST-LIO / FAST-LIO2 (iterated Kalman filter), the latter skipping feature extraction to register raw points directly, reaching up to 100 Hz and also working with solid-state lidars.

ExampleWalking a loop through a building holding a lidar with a built-in IMU, FAST-LIO2 can output the trajectory and point-cloud map in real time; in the paper’s tests, pose estimation stayed reliable even while the sensor was spun rapidly at angular rates up to about 1000°/s.

Also called

LIO, LiDAR-IMU Odometry

Related

LiDAR SLAM · FAST-LIO2 · LIO-SAM · IMU Preintegration · Tightly-Coupled vs. Loosely-Coupled Fusion · Visual-Inertial Odometry

Sources

FAST-LIO2: Fast Direct LiDAR-inertial Odometry (arXiv 2107.06829)
LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping (arXiv 2007.00258)