Toward Autonomous In-flight Docking of Unmanned Multi-rotor Aerial Vehicles

Abstract

The aim of this research was to develop a cost effective, low complexity and off-the-shelf system for autonomous in-flight docking of unmanned, battery-powered MAVs using only onboard sensors. The robotic system presented herein consists of two MAVs, a computer vision system and a docking mechanism. The docking MAV is responsible for performing the docking maneuver, while the carrier MAV holds position and receives the docking MAV. The docking mechanism consists of a rigid mast attached to the top of the carrier MAV. The vision system's camera is at the top of the mast, pointed along the z-direction of the carrier MAV. The frame of the docking MAV is ring-shaped to fit over the mast. The ring-shaped frame of the docking MAV is surrounded with LED (light-emitting diode) fiducial markers used by the vision system. A simulation was constructed in order to model the system and develop the detection, tracking and navigation algorithms. The algorithms were shown to be successful in simulation. The docking MAV's ability to hold its position over the center of the camera was tested using only GPS and the performance of that test was compared to a similar test using the computer vision system. The use of the computer vision system reduced the root-mean-square error of the MAV's position hold by 89\% (0.73 m to 0.09 m) while also reducing its time spent outside of a desired horizontal radial limit of 0.1 m to 22\% from 95\%. These results suggest that this system is a viable approach to an in-flight docking system.