1A2-H05 Mobile Robot Localization by Scene Description using BoVW(Localization and Mapping (1))

Abstract

We have been investigating the construction of an environmental map using a laser range finder for autonomous mobile robot. However, mobile robot localization using the environmental map failed in monotonous environments. This study tries to solve the problem using image information which consists of a substantial amount of visual information. Specifically, we are using image information obtained from an omnidirectional camera, which is a camera with a 360-degree field of view. A robot learns omnidirectional images using self-organizing maps (SOM), and it generates a localization classifier. Thus, we verified effectiveness of the localization classifier for localization.