Plane-based Loop Correction for LiDAR Inertial SLAM

LiDAR-based simultaneous localization and mapping (SLAM) exhibits excellent performance in indoor environments and is used to various aspects of robotic applications. Due to the inevitable odometry drift, map inconsistency always exists in mapping process. This issue is often addressed by loop correction, however, most loop correction method build pose to pose constraint based on appearance similarity, which is not able to detect map inconsistency with small overlap area.We proposed a plane-based loop detection method that is able to establish long-term association between plane feature and correct the drift without the need to go back to the same position. First, we identify map inconsistency between submap and global map. Then, we estimate relative transformation with revisited plane features. Then we utilize pose graph optimization (PGO) to correct the trajectory. Our proposed method has shown outstanding results compared with other state-of-the-art loop detection methods.