DMLO: Deep Matching LiDAR Odometry

LiDAR odometry is a fundamental task for various areas such as robotics, autonomous driving. This problem is difficult since it requires the systems to be highly robust running in noisy real-world data. Existing methods are mostly local iterative methods. Feature-based global registration methods ar...

Full description

Saved in:
Bibliographic Details
Main Authors: Li, Zhichao, Wang, Naiyan
Format: Conference Proceeding
Language:English
Subjects:
Feature extraction
Laser radar
Learning systems
Noise measurement
Pose estimation
Singular value decomposition
Task analysis
Citations: Items that cite this one
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:LiDAR odometry is a fundamental task for various areas such as robotics, autonomous driving. This problem is difficult since it requires the systems to be highly robust running in noisy real-world data. Existing methods are mostly local iterative methods. Feature-based global registration methods are not preferred since extracting accurate matching pairs in the nonuniform and sparse LiDAR data remains challenging. In this paper, we present Deep Matching LiDAR Odometry (DMLO), a novel learning-based framework which makes the feature matching method applicable to LiDAR odometry task. Unlike many recent learning-based methods, DMLO explicitly enforces geometry constraints in the framework. Specifically, DMLO decomposes the 6-DoF pose estimation into two parts, a learning-based matching network which provides accurate correspondences between two scans and rigid transformation estimation with a close-formed solution by Singular Value Decomposition (SVD). Comprehensive experimental results on real-world datasets KITTI and Argoverse demonstrate that our DMLO dramatically outperforms existing learning-based methods and is comparable with the state-of-the-art geometry- based approaches.
ISSN:2153-0866
DOI:10.1109/IROS45743.2020.9341206