A LiDAR-IMU Tightly Coupled SLAM in Subway Tunnel Environments

Multi-sensor fusion for pose estimation has emerged as a critical methodology in simultaneous localization and mapping (SLAM). However, in subway tunnel environments where GNSS signals and other external references are unavailable and sensor observations suffer from degeneracy, achieving reliable mapping and localization using only LiDAR and IMU remains highly challenging, often leading to solution divergence. To address these limitations, this paper proposes a tightly coupled LiDAR-IMU mapping and localization algorithm that specifically leverages the operational characteristics of subway trains and the geometric features of tunnels. The proposed method incorporates a rail-constrained kinematic model, along with an effective point filtering module that enhances measurement model validity based on tunnel structural features. We validate the proposed algorithm using datasets collected in real-world subway tunnels. Experimental results demonstrate that our method achieves robust and accurate tunnel mapping, significantly advancing stable and precise localization for rail transport in tunnel environments.