Improvement of LiDAR SLAM Front-end Algorithm Based on Local Mapin Similar Scenes

In the corridor, tunnel and other similar scenes, performances of traditional LiDAR SLAM (simultaneous localization and mapping) algorithms will seriously degrade, and the algorithms might even be completely invalid due to the similarity of observation data.To solve this problem, the motion prediction model is improved firstly with the hdl_graph_slam algorithm based on the assumption of uniform motion to obtain a more accurate initial pose estimation.Then, the concept of local map is introduced to densify the point cloud, and the performances of the front-end odometer are improved in the similar scenes.In the indoor experiment, the average restoration rate of the scene reaches 99.54%, which is 57.25% higher than that before improvement.In the outdoor experiment, the odometer drift is reduced from 111.62 m to 7.65 m by the improved algorithm.The experimental results show that the proposed algorithm can bring a significant performance improvement in both indoor and outdoor similar scenes.