LiDAR‐based robust localization for field autonomous vehicles in off‐road environments

Abstract

Robust localization is an essential capability for autonomous land vehicles. While a lot of work focused on structured environments, this article focuses on navigation in off-road environments. In the off-road environment, due to the lack of salient features, scan matching algorithms tend to degenerate. Therefore, the first contribution of this paper is to propose a reliable degeneracy indicator which can evaluate the scan matching performance. The evaluated degeneracy indicator is then integrated into the factor graph optimization framework which is used in both the offline mapping system and the online localization system. Moreover, a complete navigation system that can handle the incomplete and partly outdated LiDAR maps is developed. Extensive tests on real-world data sets show that the proposed system outperforms state-of-the-art approaches, especially in degenerate scenarios.

Kilometer-scale autonomous navigation in subarctic forests: challenges and lessons learned

“…Since the lidar scans are not dense enough to distinguish between single branches, the trees resemble large blobs of randomly distributed points in the scans and ultimately in the map. Therefore, point cloud degradation is not limited to the large open areas that were mentioned by Ren et al (2021). In particular, we have observed that point cloud degradation led to the ICP localization to jump forwards of up to 0.5 m. Such jumps in ICP localization lead to instability and eventually to system failure.…”

“…However, due to the location where this work was conducted, no analysis of the impact of snowfall is discussed. Recently, Ren et al (2021) have deployed a lidar localization system in a desert biome. They identified the lack of features and geometrical constraints as an issue for point cloud registration.…”

“…The purpose of the voxel manager is to limit the computational complexity of the map maintenance, instead of letting it grow with its size. Our implementation of the voxel manager module is similar to the one proposed by Ren et al (2021). It involves the following steps: the new map G Q is divided into voxels and only the voxels that are close to the robot are kept in the Random Access Memory (RAM).…”

Range-Aided LiDAR-Inertial Multi-Vehicle Mapping in Degenerate Environment

“…These directions were selected based on a dynamic threshold and real-time updated. Extending the above two methods, Ren et al [11] proposed a reliable degeneracy indicator that can evaluate the scan-matching performance in off-road environments. The evaluated degeneracy indicator was then integrated into a factor graph optimization framework.…”

“…The evaluated degeneracy indicator was then integrated into a factor graph optimization framework. However, these methods [9]- [11] only adopted a single sensor and were unable to optimize the degenerate dimension. Khattak et al [12] utilized a visual-inertial odometry and a thermal-inertial odometry to find robust priors for LiDAR pose estimation.…”