Collaborative Visual SLAM Framework for a Multi-UAVs System Based on Mutually Loop Closing

Abstract

In a typical visual Simultaneous Localization and Mapping (SLAM) algorithm integrated within a single unmanned aerial vehicle (UAV), the positioning drift will increase cumulatively due to the motion and dynamic of UAV platform, which could be efficiently alleviated by the introduction of loop detection. However, a large number of loop movements by a single UAV will result in too many turns of the drone with significant reduction of coverage area per unit time. Therefore, in this paper, we propose a collaborative framework of visual SLAM algorithm with a multiple UAVs system. In the proposed framework, a series of mutually closed loop will be detected and executed by multiple UAVs within the position map. By this coordination method, the position accuracy of the system could be obviously improved, through the experimental results compared with a single UAV SLAM system.