With the significant development of unmanned aerial vehicles (UAVs) technologies, a rapid increase on the use of UAV swarms in a wide range of civilian and emergency applications has been witnessed. However, how to efficiently network the large-scale UAVs and implement the swarms applications without infrastructure support in remote areas is challenging. In this paper, we investigate a hierarchal large-scale infrastructure-less UAV swarm scenario, where numerous UAVs surveil and collect data from the ground and a ferry UAV (Ferry UAV) is designated to carry back all their collected data. We can divide UAV swarms into different areas based on their geographic locations due to the wide range of surveillance. To improve data collection efficiency of Ferry UAV, we introduce a single super cluster head (Super-CH) UAV in each area which can be selected by the proposed modified k-means clustering algorithm with low latency. Then, we design an iterative approach to optimize the 3-dimensional (3D) trajectory of Ferry UAV such that its data collection mission completion time is minimized. Numerical results show the efficiency and low-latency of the proposed clustering algorithm, and the proposed 3D optimal trajectory design for large-scale UAV swarms data collection admits better performance than that with fixed altitude.

随着无人机技术的飞速发展，目睹了无人机群在各种民用和紧急情况下的迅速使用。然而，如何在偏远地区没有基础设施支持的情况下有效地将大型无人机联网，并实现群应用。在本文中，我们研究了一种分层的，大型的，无基础设施的无人机群的情况，在这种情况下，众多无人机进行监视并从地面收集数据，并指定了轮渡无人机（轮渡无人机）来运回所有收集的数据。由于监视范围广泛，我们可以根据其地理位置将无人机群划分为不同的区域。为了提高渡轮无人机的数据收集效率，我们在每个区域引入了一个超级集群头（Super-CH）无人机，可以通过提出的改进的k均值聚类算法来选择低延迟的无人机。然后，我们设计了一种迭代方法来优化渡轮无人机的3维（3D）轨迹，以使其数据收集任务完成时间最小化。数值结果表明了该聚类算法的有效性和低时延性，并且针对大规模无人机群数据采集的3D最优轨迹设计比固定高度的性能更好。