Outdoor Navigation of Autonomous Mobile Robot Using Localization and Road Detection Based on Visual Information

In recent years, autonomous mobile robots are expected to be useful in outdoor environments. However, it is difficult for the robots to respond to changes in the outdoor environment, which is an important problem. For an autonomous mobile robot, it is important which navigation system to use. In recent years, simultaneous localization and mapping (SLAM) is often used. This method has good performance if accurate maps is used, but it takes time and effort to create accurate maps. However, in case of human being, such accurate maps are not necessary. Our previous study of human environmental recognition ability demonstrated that human uses scenery and road information. In this study, we propose an autonomous mobile navigation method with image processing based on human environmental recognition ability. This method uses GIST feature for localization, and SegNet for road detection. We validated our method in outdoor using a mobile robot. Experimental results demonstrate that the proposed method enables autonomous mobile navigation without accurate map.