Abstract: In underground sloped roadways, the stable high-precision localization of unmanned trucks is often hindered by challenges such as difficult signal transmission, road inclination, and a lack of effective feature information. These issues significantly impact the safe and efficient operation of unmanned mining vehicles. To address these challenges, a novel positioning and mapping algorithm for unmanned mining trucks in underground sloped roadways, termed GFRMINE-LIO, is proposed based on laser SLAM. Firstly, in response to the scarcity of feature points caused by the smooth concrete walls on both sides of the entrance to the sloped roadway, a novel positioning enhancement method based on artificial landmarks is designed. This method effectively increases the number of feature point clouds, thereby optimizing the pose estimation results and preventing drift during the mapping process.Secondly, the study introduces the SCSA (Slope and Curvature based Segmentation Algorithm), which integrates slope and curvature information. By analyzing the geometric features within the point cloud data collected by laser radar, this algorithm accurately calculates the slope angle and curvature of each point. This allows for the effective identification of inclined and uneven road surfaces in underground environments, ensuring more precise point cloud filtering in complex conditions, which significantly enhances the robustness and accuracy of the algorithm in challenging terrains. Finally, on the basis of the constructed map, the GICP (Generalized Iterative Closest Point) algorithm is employed to register the real-time acquired point cloud data, integrating the GFRMINE-LIO algorithm to correct point cloud distortions for efficient relocalization. Compared to the original algorithm, the proposed method demonstrates significant improvements in accuracy. Experimental results indicate that this algorithm is capable of achieving more stable and rapid high-precision localization in harsh environments. In practical applications, a case study conducted at an underground sloped roadway of China Steel Corporation in Shandong shows that the GFRMINE-LIO algorithm achieves a 2.90% improvement in accuracy and a 20.8% reduction in Z-axis error compared to the original algorithm. The quality of the generated map is markedly enhanced, and both localization accuracy and robustness are significantly improved, effectively addressing the challenges of mapping and localization for unmanned driving in underground settings.