单线激光点云拼接SLAM算法研究

Research on Single Line Laser Point Cloud Splicing SLAM Algorithm

摘要: 针对传统单线激光SLAM算法，在地图构建过程中，由于单线激光单帧点云数据量过少，在某些局部相似的场景中可能出现得分很高的误匹配情况，本文对点云配准算法的迭代环节经行改进，并提出一种滑动窗口的点云拼接方法，利用新拼接的点云组，进行地图构建和回环检测。分别在仿真与实际环境中对本文所提算法经行验证，在仿真环境中本文改进的点云配准算法相对于原来的点云配准算法，迭代次数减少50.22%，运行时间减少39.67%；在真实环境中利用机器人平台进行定位与地图构建，结果表明本文所提算法在定位精度、建图精度和实时性上提升明显，可以有效解决单线激光运行过程中的误匹配情况。

Abstract: Aiming at the traditional single-line laser SLAM algorithm, in the process of map construction, due to the small amount of single-line laser single-frame point cloud data, there may be mismatches with high scores in some locally similar scenes. In this paper, the iterative link of the point cloud registration algorithm is improved, and a sliding window point cloud splicing method is proposed. Using the newly spliced point cloud group, we carry out map construction and loop detection. The proposed algorithm was verified in simulation and real environment respectively. In the simulation environment, compared with the original point cloud registration algorithm, the iteration times and running time of the improved point cloud registration algorithm were reduced by 50.22% and 39.67%. In the real environment, the robot platform is used for positioning and map construction. The results show that the proposed algorithm improves the positioning accuracy, map construction accuracy and real-time performance significantly, and can effectively solve the mismatching situa-tion during the operation of single line laser.