2P1-F06 屋外ナビゲーションのための見え系列生成

概要

論文の詳細を見る
Localization is indispensable for outdoor mobile robot navigation. In view-based localization, a robot localizes itself by comparing the current input image with an image sequence (called view sequence) which was taken in advance while the robot moved along a path. However, it is time-consuming to acquire the view sequence of all paths that the robot may take. Therefore, the purpose of this study is to develop a method of generating a view sequence including views from the positions where the robot never visits. For such a view generation, the environmental map is built by using the view sequence taken with an omnidirectional camera, and then it is used for generating views from new positions using the map. The method first estimates the robot's motion by visual odometry and calculates the robot's trajectory. It then makes a 3D feature map from tracked image features using triangulation. It finally generates a view sequence with prediction of views for new positions using the map.
一般社団法人日本機械学会の論文