This paper proposes a method for generating a map of a large scaled environment by observing scenes both along routes and at intersections. Panoramic representation is used for describing route scenes, and the number of routes connecting at an intersection is assumed to he known. A robot can obtain a closed loop by taking the same turn (leftmost for example) at each intersections when it moves along routes. By memorizing routes at intersections, the robot can select unmove routes for finding new closed loops. Further, by fusing new closed loops with found ones, the robot can build the map of environments. The effectiveness and robustness of our method are shown by our experiment in a real-world environment.
J-STAGEがリニューアルされました! https://www.jstage.jst.go.jp/browse/-char/ja/