A Visual Localization Technique for Unmanned Ground and Aerial Robots

Abstract

In this paper, we propose a technique for UGV robot's localization using map data of an UAV(Unmanned Aerial Vehicle) with RGB-D camera. UGV has a large loading capacity and a long battery life, but it has limited movement. On the other hand, UAV is free to move but has a small loading capacity and a short battery life. In case of UAV SLAM(Simultaneous Localization And Mapping), location recognition error occurs due to the influence of terrain and obstacles. Since odometry information of UAV SLAM is more complicated than UGV SLAM, location recognition error is accumulated when continuous position recognition is performed after generating map data. Therefore, we proposed a technique to make the UGV SLAM possible by sharing map data of two robots after making map data by using UAV which is relatively mobile. It is useful to complement each other's disadvantages between UAV and UGV. And we presented the advantages of RGB-D camera based SLAM by comparing the accuracy of a SLAM using 2D-camera and 2D LRF (Laser Range Finder) SLAM. Through a series of experiments, the effectiveness of the proposed UAV SLAM using RGB-D cameras was introduced by comparing UAV SLAM with UGV SLAM with RGB-D camera.