Flight control of multi-rotor helicopter by using Visual SLAM

In this study, we aim to realize a novel robust navigation system for Small Unmanned Aerial Vehicle (SUAV) in GPS and GPS-denied environment. Generally, the SUAV uses position and velocity information from Global Positioning System (GPS) for guidance and control. However, GPS could not be used in several environments, for example, GPS has huge error near buildings and trees, indoor, and so on. In such GPS-denied environment, several approaches for estimating the position and velocity of SUAV have been proposed. Optical motion capture system, Laser Detection and Ranging (LIDER) sensor based Simultaneous Localization and Mapping (SLAM), Visual SLAM are the kinds of approaches. However, these sensors also have the weakness. Therefore, it is desired to develop the integrated navigation system which is seamlessly applied to GPS and GPS-denied environments by using multiple sensors. To design the integrated navigation system, position estimation system by using each sensor should be constructed. In this paper, Visual SLAM system by using stereo vision camera is constructed and verified. ORB-SLAM2 which is one of the feature based Visual SLAM system is adopted for navigation of SUAV. We construct the system by using Robot Operating System (ROS) implemented in embedded computer mounted on SUAV, and reliability and accuracy of the position estimated by Visual SLAM are verified by flight experiment.