融合反光柱的2D激光SLAM和高精度定位系统

2D Laser SLAM and High-Precision Positioning System

提出一个新的融合反光柱的2D激光SLAM和纯定位系统,当前,2D激光SLAM在几何特征丰富的场景中可以稳定地工作,但是,当场景中的几何特征较少时,或者当系统工作在玻璃等特殊环境内,系统的建图和定位模块都容易失败。提出一个融合反光柱的激光SLAM实时建图和定位系统,该系统建立在现有的激光SLAM系统Cartographer基础上。在建图时,提出将反光柱作为位姿图优化的一个节点,放到后端和激光帧以及地图中的子图节点一起优化;在融合反光柱的纯系统中,提出对于当前帧中的反光柱,采用最近邻匹配的方式寻找地图中匹配的反光柱节点,在找到匹配的反光柱后,采用联合优化激光点误差和反光柱匹配误差的方式,定位激光帧的位姿。实验表明,当系统工作在几何特征少,或者玻璃等环境下时,融合反光柱的激光SLAM系统的定位精度和鲁棒性都有大幅提升。

Proposes a new 2D laser SLAM and pure positioning system that integrates reflective columns.Currently,2D laser SLAM can work stably in scenes with rich geometric features,but when there are fewer geometric features in the scene,or when the system working in special environments such as glass,the system's mapping and positioning modules are prone to failure.We have proposed a real-time laser SLAM mapping and positioning system that incorporates reflective columns.Our system is based on the existing laser SLAM system Cartographer.When constructing the map,we proposed to use the reflective column as a node to optimize the pose graph,and put it on the backend to optimize it with the laser frame and the submap nodes in the map.In a pure system that integrates the reflective column,we propose to use the reflector column in the map,the nearest neighbor matching method is used to find the matching reflector node in the map.After the matching reflector column is found,the position of the laser frame is determined by jointly optimizing the laser point error and the reflector column matching error.Experiments show that when the system works in environments with few geometric features or glass,the positioning accuracy and robustness of the laser SLAM system incorporating reflective columns are greatly improved.