Abstract: In the field of robotics, loop closure detection is recognized as a core component of simultaneous localization and mapping (SLAM) technology. Compared with visual loop closure detection methods, LiDAR (light detection and ranging) loop closure detection is able to provide more stable and robust geometric features by virtue of its hardware properties, and the interference of illumination and texture changes, which often affects visual methods, is effectively avoided. A novel global triangular descriptor with distance information is proposed in this paper. Key points are extracted through adaptive Euclidean clustering in this method; distance vectors are constructed based on the distance relationships between the key points; and the global descriptor is generated by fusing triangular geometric features. To verify the effectiveness of the proposed method, experiments are conducted on the public KITTI dataset and in real-world environments respectively. Experimental results show that the algorithm in this paper achieves significant improvements in both localization precision and accuracy, and the pose drift problem in SLAM systems, which is caused by the accumulation of sensor errors during long-term operation, can be effectively suppressed—this further enhances the long-term localization and mapping stability of the system.