Free Area Extraction and 2D Graph Map Construction based on a Geometric Property of Proximity Points
Rikuto NONOMURA, Yuichi Tazaki, Hikaru Nagano, Yasuyoshi Yokokohji · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2024
In a large-scale environment or during long-term use of SLAM, increase in data size and computational cost is crucial problems. Therefore, the need arises for maps with reduced data size, such as graph maps. Furthermore, it is essential to identify free space of the environment with low computational costs prior to constructing maps. Proximity point is a type of keypoints of point cloud that can be utilized for data reduction. In this paper, we propose a new method to extract free space with low computational cost based on a geometric property of proximity points. Subsequently, we construct 2D graph maps based on the modified Growing Neural Gas (GNG) algorithm which ensures that the edges do not interfere with the obstacles. Experiments are conducted using six sets of data acquired in an outdoor environment.