Abstract: LightLaser Detection and Ranging (LiDAR) and Ultra Wide Band (UWB) are currently widely used underground positioning technologies. To address the problem that single LiDAR positioning in mine roadways with sparse and repetitive structural features produces deviations due to insufficient geometric constraints, and UWB positioning is prone to accuracy degradation caused by Non-Line-of-Sight (NLOS) errors, this paper proposed an underground UWB/LiDAR integrated positioning method based on DS evidence theory. Signal energy in channel statistical parameters was used as the feature parameter, and kernel density estimation was employed for feature extraction. Based on the extracted energy features, Simulated Annealing-Support Vector Machine (SA-SVM) was applied to identify NLOS environments, achieving effective handling of NLOS errors in UWB positioning. DS evidence theory was then adopted to fuse UWB and LiDAR positioning data to improve positioning accuracy. Experiments were carried out in an underground roadway of a coal mine. The results showed that SA-SVM achieved 94% accuracy in NLOS environment identification. The maximum error of the UWB/LiDAR integrated positioning method based on DS evidence theory was 1.073 50 m, the minimum error was 0.002 05 m, the mean error was 0.259 34 m, the standard deviation was 0.110 05 m, and the root mean square error was 0.331 08 m. This method outperforms UWB positioning, LiDAR positioning, and the integrated positioning method based on Extended Kalman Filter.