关键词: Lidar SLAM, 无人车, 正态分布变换（NDT）, 迭代最近点（ICP）, 点云配准, 定位, 地图构建

Abstract:

Lidar SLAM technology is an important way for unmanned vehicles to carry out precise navigation, and it is also a prerequisite to realize the safe driving of unmanned vehicles in complex unstructured road environment in park. A method of fast and accurate localization and mapping is constructed. Through a large amount of point cloud data returned by vehicle lidar, noise point removal and pre-processing of voxel grid filter are carried out to realize point cloud registration while maintaining the original point cloud shape. Firstly, the NDT（Normal Distribution Transform） point cloud registration algorithm is used to estimate the position and posture of the unmanned vehicle. Then, in order to estimate the unmanned vehicle position and posture precisely, it corrects the registered point cloud using the ICP（Iterative Closest Point） point cloud registration algorithm. Finally, on the basis of the above, the map update process is completed. This method only needs the vehicle-borne lidar sensor to realize the fast and high precision lidar SLAM. The algorithm is applied to small cyclone unmanned vehicle and verified in campus environment. The results show that the algorithm is reliable and effective.

Key words: Lidar SLAM, unmanned vehicle, Normal Distribution Transform（NDT）, Iterative Closest Point（ICP）, point cloud registration, localization, mapping