Review of LiDAR-based SLAM methods

As the core technology of robot navigation and environment perception, LiDAR-based simultaneous localization and mapping (SLAM) technology has been widely used in the fields of autonomous driving, drones, robots and so on. This paper summarizes the basic framework and key technologies of LiDAR-based SLAM, and focuses on the point cloud processing, front-end data registration, back-end optimization and closed-loop detection steps of LiDAR. It further reviews classical LiDAR-based SLAM methods and their corresponding improvements, discusses their technical innovations, and reveals the developmental trajectory and research trends of LiDAR-based SLAM, namely the transition from single-sensor-driven systems to multi-sensor fusion, from traditional geometric constraints to enhanced semantic understanding, and from local pose estimation to global consistency optimization. Meanwhile, an indoor 3D mapping experiment was conducted using a Livox HAP LiDAR to validate the effective support of high-precision maps constructed by SLAM for environmental perception tasks. Experimental results show that point cloud maps obtained through LiDAR-based SLAM can not only reconstruct the spatial structure of a scene with relatively high accuracy, but also provide a reliable data foundation for downstream tasks such as semantic segmentation, thereby demonstrating the significant application value of LiDAR-based SLAM in perception systems. Finally, the paper discusses the development potential of LiDAR-based SLAM in directions such as deep learning, neural radiance fields, and multi-sensor fusion, and points out that challenges remain in dynamic environment adaptation, the trade-off between real-time performance and accuracy, and large-scale engineering deployment.