A two stage-based approach for swarm UAVs landing in unknown 3D environments

The ability to autonomously land in unknown environments is essential for achieving high intelligence of swarm unmanned aerial vehicles (UAVs). Such a capability requires that UAVs necessitate four functions: autonomous selection of landing sites, mutual coordination among the individuals, real-time trajectory planning, and global optimization decision-making. However, these modules are not sufficiently integrated into existing methods due to the complexity and uncertainty of swarm system. To address these challenges, this paper proposes a completed landing process framework for swarm UAVs using a two-stage approach. Specifically, the candidate landing sites are generated automatically from the complex environment in a coarse-to-fine manner. At the high-altitude flying stage, the swarm UAVs are considered as a whole and directed towards dynamic target points, which are updated by the designed optimization cost model. Additionally, a virtual force model is introduced to maintain the balance between the interior formation of the UAVs and the external obstacles. At the approaching landing stage, a novel landing model based on rapidly exploring random tree (RRT) is proposed to plan the final landing paths, which can effectively avoid collisions with other UAVs and ground obstacles. The proposed method is performed in real-world scenarios and estimated via different fly stages. The results of comprehensive experiments demonstrate that the proposed approach achieves good landing performance in multiple scenarios, including landing site selection and progressive swarm path planning. This also supports industrial information integration by enabling coordinated sensing, communication, and decision-making for swarm UAVs in complex environments.