Vision-based autonomous flocking of UAVs in unknown forest environment

This paper studies tandem flocking of a team of unmanned aerial vehicles (UAVs) equipped with limited range sensors exploring an obstacle-rich GPS-denied environment such as a forest or urban canyon. There is no communication between UAVs and the follower can only estimate its leader's relative position by its on-board camera. A vision-based leader-follower flocking strategy is proposed to achieve this kind of flocking in unknown cluttered environments. The leader UAV explores its way to reach a given goal position and the follower UAV follows its corresponding leader UAV via on-board camera and avoid obstacles at the same time. The simulation of a team of six UAVs demonstrates the effectiveness of our proposed flocking strategy in a forest environment.