D-LIO: 6DoF Direct LiDAR-Inertial Odometry based on Simultaneous Truncated Distance Field Mapping

TL;DR

This work tackles the challenge of real-time, 6DoF LiDAR–inertial odometry with online distance-field mapping on CPU. It introduces Direct LiDAR-Inertial Odometry (D-LIO) that performs nonlinear SE(3) registration of raw LiDAR data against a Fast-TDF map, while updating the map in constant time. A binary-kernel, bitwise-AND-based Fast-TDF representation enables map updates that scale with the number of points rather than the map size, providing an efficient, continuous distance-field for downstream tasks. Experimental results on public datasets show competitive odometry accuracy with the added benefit of an online, usable distance-field map for planning and collision avoidance.

Abstract

This paper presents a new approach for 6DoF Direct LiDAR-Inertial Odometry (D-LIO) based on the simultaneous mapping of truncated distance fields on CPU. Such continuous representation (in the vicinity of the points) enables working with raw 3D LiDAR data online, avoiding the need of LiDAR feature selection and tracking, simplifying the odometry pipeline and easily generalizing to many scenarios. The method is based on the proposed Fast Truncated Distance Field (Fast-TDF) method as a convenient tool to represent the environment. Such representation enables i) solving the LiDAR point-cloud registration as a nonlinear optimization process without the need of selecting/tracking LiDAR features in the input data, ii) simultaneously producing an accurate truncated distance field map of the environment, and iii) updating such map at constant time independently of its size. The approach is tested using open datasets, aerial and ground. It is also benchmarked against other state-of-the-art odometry approaches, demonstrating the same or better level of accuracy with the added value of an online-generated TDF representation of the environment, that can be used for other robotics tasks as planning or collision avoidance. The source code is publicly available at https://anonymous.4open.science/r/D-LIO

Figures (3)

• Figure 1: 2D example of $5 \times 5$ binary kernel using 4 bits encoding. (a) Correspondence between mask and $L_1$ distance. (b) 2D occupancy representation of the input point cloud over the grid. (c) Merged distance map resulting from the bitwise-AND operation over overlapping binary kernels centered at each occupied cell.
• Figure 2: D-LIO Workflow. Red indicates the Kalman filter, green the TDF grid map, and blue the LiDAR preprocessing and optimization.
• Figure 3: 3D map reconstructions (from left to right) of "eee", "nya", "sbs", and Newer College (Quad-Easy) datasets.