RID: LiDAR Range Image Descriptor for Fast and Efficient Loop Closure Detection in Indistinguishable Environments

摘要

Loop closure detection is a crucial aspect of simultaneous localization and mapping (SLAM) systems in robotics, as it helps to enhance the accuracy and consistency of robot localization by recognizing previously visited locations. LiDAR SLAM is a widely adopted approach for localization tasks due to its robustness to changes in illumination. However, the use of existing global descriptors for loop closure detection in indoor environments with narrow corridors and confined spaces is challenging due to limitations in representing the environment’s geometry. To address this issue, we propose a novel approach to loop closure detection using LiDAR Range Image Descriptors (RID) that can effectively capture the geometry of indistinguishable environments with low computation cost. Our proposed descriptor compresses 3D point clouds onto a 2D image plane while preserving distance information. In this paper, we demonstrate the efficacy of our approach through experiments in large-scale indoor and outdoor environments, showing that our descriptor outperforms existing methods in terms of computational efficiency in challenging, indistinguishable environments. Based on the experimental results, our proposed approach is suitable for low-power robots and can be implemented in real-time, making it a promising technique for loop closure detection in challenging environments.