NA-LOAM: Normal-Based Adaptive LiDAR Odometry and Mapping

Light detection and ranging (LiDAR)-based simultaneous localization and mapping (SLAM) exhibits excellent performance in large-scale real-world scenarios and is widely applied in robot navigation systems. However, the adaptability of LiDAR-based SLAM algorithms in different environments remains a challenge. The fixed parameter settings and local information-based weighting strategies can influence the performance and reliability of LiDAR-based SLAM algorithms across various environments and application scenarios. To address the above issues, this article introduces a method based on point cloud normals to evaluate the degree of environmental degradation. This approach adaptively weights point clouds and dynamically adjusts optimization hyperparameters. Specifically, we first utilize distinct lookup tables for ground and nonground points based on the scanning structure of the LiDAR, allowing for the rapid computation of the point cloud normals. Subsequently, we used the weighted covariance matrix (WCM) of normal vectors to assess the degree of environmental degradation. Finally, based on the degradation level, we dynamically adjust optimization hyperparameters and compute the weight of each point. The proposed method demonstrates higher accuracy and robustness in diverse environments through validation on the KITTI benchmark and real-world scenarios.