【摘要】 在GNSS失锁情况下，如何基于三维激光雷达进行定位，是值得关注的问题。通过点云匹配或蒙特卡洛定位(MCL)的方法进行定位，存在初值敏感、计算时间长和搜索空间大的问题。本文提出了一种定位方法，首先利用描述子初始化，确定粗略的位姿，缩小搜索空间，然后使用MCL进行精定位。通过给定描述子初值，预先计算三维网格概率，达到了缩小搜索空间、减少计算时间的效果。初步的试验结果表明，该方法在GNSS失锁情况下定位精度保持较好。更多还原

【Abstract】 When GNSS loses lock, how to locate based on 3D LiDAR is a problem worthy of attention. Positioning by point cloud matching or Monte Carlo method has the problems of sensitivity to initial value, long calculation time and large search space. This paper proposes a two-step localization method, which first uses descriptor initialization, determines the rough pose, narrows the search space, and then uses Monte Carlo for fine localization. The use of descriptors in the method gives the initial value, reduces the search space, and pre-computes the three-dimensional grid probability, which reduces the computation time. Preliminary experimental results show that the positioning accuracy remains good in the case of GNSS loss of lock.更多还原