Study on Map Construction of Spherical Robot based on Statistical Filtering

Traditional robots lacked the detection and preservation of environmental information. This paper proposes to apply SLAM to robots. We put a camera on the robot to capture the environment and map it. Traditional SLAM algorithms typically generate point cloud maps, which are bulky and unreadable and cannot be used for robot navigation. Therefore, point cloud map is generally converted into octree map, but due to environmental interference, sensor error and other reasons, point cloud map often contains a lot of noise, which leads to a large error in the generated octree map. In this paper, based on OBR SLAM2 algorithm, proposed a filtering method, through the judgment of point cloud map points automatically select filtering mode, thus generating high quality octree map. Compared with unfiltered maps through experiments, the map generated by this method is smaller volume and more accurate.