Vision-Based Monocular SLAM in Micro Aerial Vehicle

Abstract

Micro Aerial Vehicles (MAVs) are popular for their efficiency, agility, and lightweights. They can navigate in dynamic environments that cannot be accessed by humans or traditional aircraft. These MAVs rely on GPS and it will be difficult for GPS-denied areas where it is obstructed by buildings and other obstacles.  Simultaneous Localization and Mapping (SLAM) in an unknown environment can solve the aforementioned problems faced by flying robots.  A rotation and scale invariant visual-based solution, oriented fast and rotated brief (ORB-SLAM) is one of the best solutions for localization and mapping using monocular vision.

 In this paper, an ORB-SLAM3 has been used to carry out the research on localizing micro-aerial vehicle Tello and mapping an unknown environment.  The effectiveness of ORB-SLAM3 was tested in a variety of indoor environments.   An integrated adaptive controller was used for an autonomous flight that used the 3D map, produced by ORB-SLAM3 and our proposed novel technique for robust initialization of the SLAM system during flight.  The results show that ORB-SLAM3 can provide accurate localization and mapping for flying robots, even in challenging scenarios with fast motion, large camera movements, and dynamic environments.  Furthermore, our results show that the proposed system is capable of navigating and mapping challenging indoor situations.