A Distributed Method to Form UAV Swarm based on Moncular Vision

The intermittent communication links or satellite navigation signals of UAV nodes will make it difficult for them to form a swarm through information exchange, this paper proposes a distributed method to form UAV swarm based on monocular vision information. Firstly, we construct a distributed swarm control model based on monocular vision information. Secondly, a modified YOLOV3-tiny method is proposed to reduce the resource consumption of target detection through model pruning. Thirdly, a geometric method is proposed to estimate the orientation of target UAVs and the distance to them. Finally, a prototype system is designed and implemented on the Gazebo simulation platform, and swarm flight tests are carried out in various scenarios to verify the feasibility of the proposed method.