Abstract:

Mobile robots in a variety of tasks need to be built in map, location and path planning, but the current vision SLAM can only output camera trajectory map, but cannot generate a route planning and navigation maps. Therefore, this paper proposes a real-time 3D raster map construction algorithm, based on the ORB_SLAM2, combined with RGB-D camera. An Inverse Sensor Model（ISM） is first established. Secondly, the algorithm of 3D grid map is reconstructed for the ISM model. Finally, it combines with ORB_SLAM2 for data set experiments, simulation environment experiments and real-time construction experiments. The experimental results show that the algorithm can construct a 3D grid map with scale in real time by using ORB_SLAM2, and can clearly display the obstacle position, and verify the effectiveness of the algorithm.

Key words: ORB_SLAM2, RGB-D camera, Inverse Sensor Model（ISM）, 3D grid map

摘要：

移动机器人在各种任务中需要进行建图、定位和路径规划，但是目前的视觉SLAM只能输出相机的运动轨迹图，而不能生成用于路径规划和导航的地图。因此，在ORB_SLAM2的基础上，与RGB-D相机相结合，提出了一种实时3D栅格地图构建算法。建立了一个逆传感器模型（Inverse Sensor Model，ISM）；针对ISM模型，重新构建了3D栅格地图的算法；联合ORB_SLAM2进行数据集实验、仿真环境实验和实时构建实验。经实验验证，该算法能够利用ORB_SLAM2实时构建出具有尺度的3D栅格地图，且能够清晰地显示障碍物位置，验证了该算法的有效性。

关键词: ORB_SLAM2系统, RGB-D相机, 逆传感器模型（ISM）, 3D栅格地图