Visual assisted formation control of distributed UAV swarm

Abstract

In recent years, unmanned aerial vehicles have been used effectively in many military and civilian areas. Because of it provides many advantages areas of surveillance, imaging, mapping, search and rescue etc. It has become an almost indispensable part of them. The use of more than one unmanned aerial vehicle in the performance of the tasks saves time and cost compared to a single vehicle, while allowing the completion of tasks that cannot be done with a single vehicle. For example, in missions where a certain area is required to be imaging or mapping, a single drone takes longer to scan the entire area than multiple drones. Since the airtime of small unmanned aerial vehicles is relatively short, a battery replacement will interrupt the task and increase this time even more. Using large-sized unmanned aerial vehicles that can stay in the air for a longer time can make the same task too costly. The use of unmanned aerial vehicles swarms allows different tasks to be performed by using methods such as triangulation or back azimuth that cannot be done with a single vehicle. Success of swarming often depends on preserving swarm formation. It is critical for all swarm members to maintain their position in the formation to successfully perform certain tasks, such as triangulation. In order to achieve this, each swarm member must precisely determine its relative position with respect to other agents. In this study a swarm of UAVs has been created by using method which relative position of UAVs determined by using only a monocular camera and IMU sensor, without GPS, VICON, LIDAR etc. which are external dependent or high-cost systems. UAVs are estimates the relative position between them by intersecting areas in the camera images and IMU acceleration measurements they send to each other. In formation acquisition, flocking and formation maneuver simulations, the swarm agents managed to create a swarm by staying within %5 error zone.