A Real-time Degeneracy Sensing and Compensation Method for Enhanced LiDAR SLAM

TL;DR

The paper addresses the problem of LiDAR SLAM degradation in degenerate geometries by introducing a real-time Degeneracy-Aware framework that combines a Hessian-derived degeneracy factor with a threshold-free DBSCAN perceptual method. It then applies a degeneracy-resistance fusion that projects and combines IMU motion with LiDAR estimates along the detected degeneracy directions, improving robustness in challenging environments. Key contributions include a clear degeneracy factor defined from eigenvalue ratios $S_{sR}=\lambda_{r1}/\lambda_{r3}$ and $S_{st}=\lambda_{t1}/\lambda_{t3}$, a DBSCAN-based degeneracy sensing method, and a reactive fusion scheme that mitigates degeneracy effects in real time. The approach is validated on eight sequences with diverse LiDAR modalities, demonstrating improved localization and mapping accuracy without extra sensors and across varying scene types.

Abstract

LiDAR is widely used in Simultaneous Localization and Mapping (SLAM) and autonomous driving. The LiDAR odometry is of great importance in multi-sensor fusion. However, in some unstructured environments, the point cloud registration cannot constrain the poses of the LiDAR due to its sparse geometric features, which leads to the degeneracy of multi-sensor fusion accuracy. To address this problem, we propose a novel real-time approach to sense and compensate for the degeneracy of LiDAR. Firstly, this paper introduces the degeneracy factor with clear meaning, which can measure the degeneracy of LiDAR. Then, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering method adaptively perceives the degeneracy with better environmental generalization. Finally, the degeneracy perception results are utilized to fuse LiDAR and IMU, thus effectively resisting degeneracy effects. Experiments on our dataset show the method's high accuracy and robustness and validate our algorithm's adaptability to different environments and LiDAR scanning modalities.

Figures (13)

• Figure 1: Research background. The lack of adaptability of LiDAR to environments with degenerate geometries leads to errors in state estimation and challenges LiDAR SLAM. The current algorithms use metrics with unclear meaning to determine whether the LiDAR is degenerate, which could be more effective in measuring the degeneracy degree of LiDAR and solving the degeneracy problem of LiDAR.
• Figure 2: Block diagram illustrating the full pipeline of the proposed Degeneracy-Aware and Resistance method. The system starts with the measurements of LiDAR and IMU. The features are extracted from the point cloud to provide information for later point cloud registration. During the point cloud registration, the LiDAR's degeneracy will be sensed. Simultaneously, the IMU integration will be performed. After LiDAR degeneracy, information from the LiDAR odometry is fused with information from the IMU to resist degeneracy in positioning and mapping accuracy caused by LiDAR degeneracy.
• Figure 3: Degenerate state Graph. LiDAR is now in a single-degree degeneracy environment.
• Figure 4: K-distance Graph. From a sample data set, we draw the k-distance graph and obtain Eps. The $\mathrm{X}$-axis represents the index of the data, and the $\mathrm{Y}$axis represents the value of the data. Eps is at the inflection point of the k-distance graph.
• Figure 5: Hardware system. (a) is a four-wheel robot car with an Ouster-64 LiDAR, GPS, and a Realsense-435i camera. (b) is a handheld LiDAR that includes a Livox-Avia LiDAR and a Realsense-435i camera.