In communication-deprived disaster scenarios, this paper introduces a
Micro-Unmanned Aerial Vehicle (UAV)- enhanced content management system. In the
absence of cellular infrastructure, this system deploys a hybrid network of
stationary and mobile UAVs to offer vital content access to isolated
communities. Static anchor UAVs equipped with both vertical and lateral links
cater to local users, while agile micro-ferrying UAVs, equipped with lateral
links and greater mobility, reach users in various communities. The primary
goal is to devise an adaptive content dissemination system that dynamically
learns caching policies to maximize content accessibility. The paper proposes a
decentralized Top-k Multi-Armed Bandit (Top-k MAB) learning approach for UAV
caching decisions, accommodating geotemporal disparities in content popularity
and diverse content demands. The proposed mechanism involves a Selective
Caching Algorithm that algorithmically reduces redundant copies of the contents
by leveraging the shared information between the UAVs. It is demonstrated that
Top-k MAB learning, along with selective caching algorithm, can improve system
performance while making the learning process adaptive. The paper does
functional verification and performance evaluation of the proposed caching
framework under a wide range of network size, swarm of micro-ferrying UAVs, and
heterogeneous popularity distributions.

在通信受限的灾难情景中，本文提出了一种微型无人机增强的内容管理系统。该系统在缺乏蜂窝基础设施的情况下，部署了一个由静止和移动无人机组成的混合网络，为孤立的社区提供重要的内容访问。通过装备有垂直和水平链接的静态锚定无人机，该系统满足本地用户的需求，而通过装备有水平链接和更高灵活性的机动微型运输无人机，系统可以覆盖多个社区中的用户。该系统的主要目标是设计一种自适应内容传播系统，动态学习缓存策略以最大限度地提高内容可访问性。本文提出了一种分散式的 Top-k 多臂赌博机学习方法，用于无人机缓存决策，以适应地理和时间上的内容流行度差异和多样化的内容需求。所提出的机制涉及一种选择性缓存算法，通过利用无人机之间的共享信息，自动减少内容的冗余副本。实验证明，Top-k 多臂赌博机学习和选择性缓存算法可以改善系统性能，同时使学习过程适应性更强。本文在各种网络规模、微型运输无人机群体和异构的流行度分布下对所提出的缓存框架进行了功能验证和性能评估。