Quatro++: Robust Global Registration Exploiting Ground Segmentation for Loop Closing in LiDAR SLAM

Hyungtae Lim,Beomsoo Kim,Daebeom Kim,Eungchang Mason Lee,Hyun Myung

2024-01-22

Abstract:Global registration is a fundamental task that estimates the relative pose between two viewpoints of 3D point clouds. However, there are two issues that degrade the performance of global registration in LiDAR SLAM: one is the sparsity issue and the other is degeneracy. The sparsity issue is caused by the sparse characteristics of the 3D point cloud measurements in a mechanically spinning LiDAR sensor. The degeneracy issue sometimes occurs because the outlier-rejection methods reject too many correspondences, leaving less than three inliers. These two issues have become more severe as the pose discrepancy between the two viewpoints of 3D point clouds becomes greater. To tackle these problems, we propose a robust global registration framework, called \textit{Quatro++}. Extending our previous work that solely focused on the global registration itself, we address the robust global registration in terms of the loop closing in LiDAR SLAM. To this end, ground segmentation is exploited to achieve robust global registration. Through the experiments, we demonstrate that our proposed method shows a higher success rate than the state-of-the-art global registration methods, overcoming the sparsity and degeneracy issues. In addition, we show that ground segmentation significantly helps to increase the success rate for the ground vehicles. Finally, we apply our proposed method to the loop closing module in LiDAR SLAM and confirm that the quality of the loop constraints is improved, showing more precise mapping results. Therefore, the experimental evidence corroborated the suitability of our method as an initial alignment in the loop closing. Our code is available at <a class="link-external link-https" href="https://quatro-plusplus.github.io" rel="external noopener nofollow">this https URL</a>.

Robotics

What problem does this paper attempt to address?

### The Problem Addressed by the Paper

The paper aims to address two main issues in LiDAR Simultaneous Localization and Mapping (LiDAR SLAM) global registration: sparsity and degeneracy.

1. **Sparsity Issue**:
- Due to the characteristics of mechanical rotating LiDAR sensors, 3D point cloud measurement data becomes very sparse at long distances. This sparsity reduces the expressiveness of feature descriptors, leading to a large number of incorrect matches and increasing the proportion of outliers in the correspondence estimation.

2. **Degeneracy Issue**:
- Outlier rejection algorithms sometimes excessively reject correspondences, resulting in fewer than 3 valid correspondences remaining, making effective pose estimation impossible. This is particularly severe in cases of large pose differences.

To address these issues, the authors propose a robust global registration framework named Quatro++. This framework enhances the robustness of global registration by utilizing ground segmentation techniques, especially during loop closing. Specifically, the main contributions of Quatro++ include:

- Proposing a new global registration framework, Quatro++, which can overcome gross outliers and degeneracy issues to achieve initial alignment.
- Utilizing ground segmentation techniques to make global registration more robust, reduce computational costs, and improve the accuracy and speed of loop closing in LiDAR SLAM.
- Analyzing the performance of the method from multiple perspectives, demonstrating its superiority over existing methods, particularly its training-free characteristic.
- Integrating Quatro++ into various LiDAR SLAM frameworks, verifying that it significantly enhances the overall performance of SLAM by providing accurate loop closure constraints as initial alignment.

In summary, the goal of the paper is to achieve robust global registration in the presence of inaccurate correspondences caused by degraded feature extraction and matching.