Accurate and augmented navigation for quadcopter based on multi-sensor fusion

In this paper, we propose a navigation system consisting of a novel multi-sensor fusion method for calculating precise and accurate aerial coordinates and orientation, of a quadcopter in indoor and GPS-silent environments. A prototype system is developed that is composed of 2 modules: Simultanious Localization and Mapping (SLAM) system that uses Oriented FAST and rotated BRIEF (ORB) features, also known as ORB-SLAM and an Extended Kalman Filter (EKF). A Proportional Integral and Differential (PID) controller is employed for achieving accurate aerial maneuvers. The novel fusion method used in the system enables the quadcopter to calculate the geographical pose of the quadcopter and simultaneously calculate the map of the surroundings with minimal error. Most importantly, we develop an accurate scale-estimation system which calculates the scale of the map generated by the monocular SLAM system. The results of the experiments conducted validate the effectiveness of the system in achieving an accurate estimation of aerial movements in an indoor environment with minimal error. The proposed system can be used even in large areas as it uses a DBoW (Bag-of-Words) based Loop Closure and an optimal Bundle Adjustment which is a non-linear technique for optimizing the pose and the 3D map-points. This navigation system can be helpful for quadcopters used in diverse applications such as monitoring, surveillance, transportation, construction, etc.