Section-LIO: A High Accuracy LiDAR-Inertial Odometry Using Undistorted Sectional Point

Abstract

Simultaneous localization and mapping has become one of the core modules of unmanned platforms. High-precision localization and mapping play an important role in the collision avoidance and planning of robots in complex environments. One of the key technologies for improving the accuracy of LiDAR-Odometry is point cloud distortion removal. In existing research, laser point cloud undistortion is mostly performed on the entire frame of the point cloud, and the lower the frequency of the point cloud, the more difficulty it is to remove distortion. The main innovation of this study is the proposal of a robust point cloud partition boundary division method, Section-LIO, that takes into account the matching and undistortion effects of each point cloud. The accuracy of the overall laser odometer is improved using inertial measurement unit to remove distortion from the segmented point cloud. Section-LIO supports both 360° mechanical rotation LiDAR and light and small solid-state LiDAR. In experiments with public and private datasets, Section-LIO outperforms three existing state-of-the-art algorithms in terms of accuracy. All implementations of our Section-LIO are open-sourced on GitHub.

Keywords

• Simultaneous localization and mapping (SLAM)
• nonlinear distortion
• sensor fusion