A Tightly Coupled LiDAR-Inertial Localization Method for High-Speed Trains

Abstract

Accurate and robust localization is fundamental to the autonomous operation of high-speed trains (HSTs). However, conventional satellite-based localization becomes unreliable in degraded environments, such as long tunnels, which pose serious challenges to continuous and precise train localization. To address this issue, a tightly coupled light detection and ranging (LiDAR)-inertial method is proposed for HSTs autonomous localization, which does not rely on global navigation satellite system (GNSS). Specifically, the method adopts an on-manifold error-state Kalman filter (ESKF), which effectively constrains error propagation under high-speed motion and ensures consistent state estimation. In addition, a point-by-point update mechanism is employed, allowing the system to perform state updates at each LiDAR point cloud, where all points are deskewed and sequential residual updates are performed. Moreover, a multiframe merged update aggregates several consecutive LiDAR scans into an information-consistent refinement step, improving robustness in degraded scenes. Extensive experiments are conducted on real-world HSTs, including multiple long tunnel environments. The proposed method outperforms existing LiDAR-based methods, especially during high-dynamic motion. Significantly, this study demonstrates the potential of the method to enable reliable autonomous localization for HSTs.