LiDAR-Inertial Integration for Rail Vehicle Localization and Mapping in Tunnels

Modern light detection and ranging (LiDAR)-based simultaneous localization and mapping (SLAM) systems have delivered excellent results in real-world scenarios. However, the potential of LiDAR SLAM still lacks well investigation for rail vehicle applications. This article proposes an SLAM method for rail vehicles in tunnel environments, which fully exploits the typical geometric feature structure in the tunnels. The system receives measurements from a LiDAR and an inertial measurement unit (IMU). As a front end, the estimated motion variation from IMU measurement deskews the denoised point clouds and produces an initial guess for frame-to-frame LiDAR odometry. A degeneracy-aware feature selection is employed to select the most informative features. As a backend, a factor graph is formulated to jointly optimize the multimodal information. Besides, we leverage the plane constraints from extracted rail tracks and the pole-like features to further constrain the 6-D state estimation. In addition, the real-time performance can be achieved with an onboard computer. Through extensive evaluation of datasets gathered over an extended time range in a railway scenario, it has been demonstrated that our proposed system delivers reliable localization accuracy even in long tunnel environments.