CT-ICP: Real-time Elastic LiDAR Odometry with Loop Closure

Abstract

Multi-beam LiDAR sensors are increasingly used in robotics, particularly for autonomous cars for localization and perception tasks. However, perception is closely linked to the localization task and the robot's ability to build a fine map of its environment. For this, we propose a new real-time LiDAR odometry method called CT-ICP, as well as a complete SLAM with loop closure. The principle of CT-ICP is to use an elastic formulation of the trajectory, with a continuity of poses intrascan and discontinuity between scans, to be more robust to high frequencies in the movements of the sensor. The registration is based on scan-to-map with a dense point cloud as map structured in sparse voxels to operate in real time. At the same time, a fast method of loop closure detection using elevation images and an optimization of poses by graph allows to obtain a complete SLAM purely on LiDAR. To show the robustness of the method, we tested it on seven datasets: KITTI, KITTIraw, KITTI-360, KITTI-CARLA, ParisLuco, Newer College, and NCLT in driving and high-frequency motion scenarios. The CT-ICP odometry is implemented in C++ and available online. The loop detection and pose graph optimization is in the framework pyLiDAR-SLAM in Python and also available online. CT-ICP is currently first, among those giving access to a public code, on the KITTI odometry leaderboard, with an average Relative Translation Error (RTE) of 0.59% and an average time per scan of 60ms on a CPU with a single thread.

A Practical Survey on Visual Odometry for Autonomous Driving in Challenging Scenarios and Conditions

“…This algorithm is run multiple iterations until a local minimum is found. The basic concepts of ICP are used in a variety of state-of-the-art approaches, for example, CT-ICP [42], which is one of the best performing odometry systems on the KITTI autonomous driving benchmark. CT-ICP adapts the ICP algorithm to work in real-time, taking into consideration the distortion of point clouds caused by sensor motion, as it happens in the autonomous driving scenario.…”

Effective Solid State LiDAR Odometry Using Continuous-time Filter Registration

“…These drawbacks prohibit it from the realtime applications. Recently, Dellenbach et al [10] compensate the distortions in conventional spinning LiDAR by linearly interpolating the beginning pose with end one within a scan, where the discontinuity between consecutive scans is penalized by the additional position and velocity constraints.…”

“…where the first state s 0 is set to identity. The previous method [10] only takes into account of the translation consistency between the consecutive scans in selfdriving cars. However, the movement of handheld devices is more complicated and irregular, where the orientation change needs to be considered as well.…”

“…To tackle the above critical challenges, we present an effective Continuous Time LiDAR Odometry (ECTLO) method for SSLs with the non-repetitive scanning patterns in this letter. Inspired by the recent progress on multi-line spinning LiDAR [10], we try to eliminate the motion distortions within a scan through a continuous-time motion model. To account for the sparsity of a single scan, we take advantage of a pointto-plane Gaussian Mixture Model scheme [11] to effectively align the scans.…”

Semantic localization on BIM-generated maps using a 3D LiDAR sensor

“…Specifically, we use lidar odometry and mapping (LOAM) [6] and direct lidar odometry (DLO) [51] to generate 3D point cloud maps with only a mobile LiDAR sensor. To address the problem, some SLAM systems [52,53] introduce some more complex modules, e.g., graph optimization with loop closings, but these methods requires fine-tuned parameters to achieve drift reduction. Besides that, to generate a complete map of a multi-storey building, robot traveling trajectory needs to cover all rooms and halls, bringing potential difficulties for applications.…”