【摘要】 管道巡检机器人执行天然气管道内部任务的重要前提是具备定位建图能力。主流激光同步定位与地图构建（SLAM）技术在管道环境中面临两大问题：显著特征点数量不足和点云难以配准。这些问题导致SLAM算法定位精度降低，稳定性差。为了解决该问题，在FAST-LIO2的基础上提出了一种基于管件结构点云配准的激光SLAM算法。首先，针对特征点数量不足的问题，通过融合管件形状的几何先验，优化了特征提取算法。其次，针对管道中点云的几何特征配准退化问题，提出一种基于管件结构信息的配准优化方法，提取出管件的结构信息并根据管件类型进行配准。最后，通过实验结果证明所提算法较FAST-LIO2算法在管道环境下具有更好的定位精度和鲁棒性，定位精度（均方根误差）提升了30%，为管道机器人定位建图提供了一种可行的方案。更多还原

【Abstract】 Localization and mapping capabilities are critical for the operation of intelligent inspection robots within the interiors of natural gas pipelines. The mainstream LiDAR simultaneous localization and mapping(SLAM) technology encounters two significant challenges in pipeline environment: a scarcity of salient feature points and difficulties in point cloud registration. These challenges result in reduced positioning accuracy and poor stability of the SLAM algorithms. In order to solve the problem, a LiDAR SLAM algorithm based on point cloud alignment of pipeline fitting is proposed on the basis of FAST-LIO2. Firstly, the feature extraction algorithm is optimized by incorporating a priori knowledge of pipe fitting shapes to tackle the problem of insufficient feature points. Secondly, a registration optimization method based on the structural information of pipe fitting is introduced to address the issue of geometric feature registration degradation in pipelines. The structural information of the pipe fitting is extracted and registered according to the type of fitting. Finally,experimental results demonstrate that the proposed SLAM algorithm offers superior positioning accuracy and robustness in pipeline environment compared to the FAST-LIO2 algorithm, with a 30% enhancement in positioning accuracy(root mean square error), which provides a feasible solution for pipeline robot positioning and mapping.更多还原