Abstract: To address the harsh underground coal mine conditions with complete GNSS outage and sparse roadway geometric features, as well as the inherent drawbacks of conventional SLAM algorithms including heavy computational overhead and severe long-distance cumulative drift, this paper proposes a lightweight LiDAR-inertial SLAM localization method constrained by UWB measurements for underground mining applications. In the mapping process, an IMU-driven piecewise dynamic distortion compensation scheme is employed to rectify point cloud distortion from solid-state LiDAR. A tightly coupled LiDAR-IMU odometer is formulated, where roadway geometric constraints and ground plane constraints are fused to strengthen registration robustness in feature-sparse scenarios. Global UWB position measurements are exploited to suppress low-frequency pose drift, and incremental local map management is adopted to achieve lightweight mapping. During the relocalization phase, real-time matching localization is implemented based on the pre-established prior point cloud map, and UWB readings serve as the judgment standard for drift monitoring and relocalization activation. An early warning is triggered alongside matching range adjustment once the positional deviation between LiDAR matching outputs and UWB data exceeds the preset warning threshold; if the deviation goes beyond the relocalization threshold, UWB coordinates are utilized to assist in recovering the system localization. Experimental results collected in real underground roadways demonstrate that the presented method achieves satisfactory mapping quality. It outperforms state-of-the-art algorithms such as DLO and LILI-OM in localization accuracy and real-time performance, which fulfills the high-precision and lightweight positioning demands of driverless trackless rubber-tyred vehicles in underground coal mines.