基于内容感知的无人机辅助通信网络主动缓存技术研究

The project intends to study the proactive caching scheme of unmanned aerial vehicle (UAV) enabled communication networks by using artificial intelligence methods to enable the prediction of users' content request distribution and mobility pattern. We first establish an assessment model on caching performance, and analyze the interconnections among UAV flight parameters, caching strategy and quality-of-experience (QoE), which makes sense to offer a theoretical support. Then, based on the considered assessment model, the deployment strategy of UAVs coupled with proactive caching is designed for UAV-enabled long-lasting communications. Due to the serious data sparseness problem in matrices of content popularity, machine learning method is adopted for precise prediction of users' content request distribution by extracting social information. Moreover, to improve the mobile users' QoE, the multi-UAV-based caching scheme will be presented. Finally, based on the concept of integrating multicast communication into caching-enabled UAVs, the proactive caching scheme, in case of synchronously content request, in multi-UAV-enabled multicast communications is studied, the goal of which is to achieve energy-efficient UAV-enabled communications by invoking computing resources from ground users and to reduce the complexity of data training. Additionally, by considering the indeterminacy of users' requested contents and the asynchronism of users' requests, UAV-enabled deterministic coding based caching scheme is proposed for improving predicative accuracy. Conclusively, these solutions and research results about the key problems in content-aware and user-behavior-based UAV-enabled caching schemes will provide important theory and key technical support to the practical deployment of UAV-assisted networking in the future.

本项目拟利用人工智能方法，通过感知用户需求内容和学习用户行为，研究无人机辅助通信网络的主动缓存问题。首先，建立缓存性能评价模型和分析无人机飞行参数-缓存策略-QoE（Quality of Experience）三者的理论关系，为缓存方案设计提供理论基础。其次，针对内容流行度数据稀疏的问题，利用机器学习方法提取社交信息用于设计无人机的部署和主动缓存方案，实现内容请求分布的精准预测和无人机持久通信；以及研究多个无人机的协作缓存方案，增强移动用户的通信体验。最后，引入多播通信技术，研究内容同步请求下无人机多播通信的主动缓存方案，通过充分调用用户的计算资源，降低数据训练复杂度并实现无人机节能通信；以及研究无人机多播通信的确定性编码缓存方案，解决用户内容请求的异步性问题，提高缓存命中率。由此获得一系列基于内容感知和用户行为的主动缓存方案，为未来实际应用无人机辅助通信系统提供重要的理论和关键技术支撑。

在国家自然科学基金“基于内容感知的无人机辅助通信网络主动缓存技术研究”(项目号61801243)的支持下，我们在过去的三年时间里实现一系列关于无人机辅助通信和缓存策略的技术突破。成功设计出保证用户QoE前提下无人机基站的智能部署方案，优化得到的传输功率、无人机部署数目、位置和缓存内容等策略，适用于多目标的无人机部署问题同时保持功率消耗和回程节省之间的最优折中；设计出面向地面终端QoE最大化的多无人机部署及发射功率分配的联合优化方案，适用于未获知地面信息情况下的多无人机辅助部署场景；设计出基于缓存的无人机中继辅助车联网传输优化方法，适用于在热门文件内容主动缓存下优化无人机轨迹与无人机-车辆关联策略的通信场景；设计出无人机-车辆的关联方案以及无人机的飞行策略，该方法实现了在能量消耗和信息因果关系的约束下车辆接收信息中最小可达率之和的最大化；设计出基于缓存的无人机增强无线通信系统策略以及轨迹和功率分配的联合优化方案，该方案实现了在无人机功率和移动性等约束下系统吞吐量以及能效等多目标的最大化；设计出无人机与卫星共同辅助地面通信下无人机飞行策略和功率分配方案，该方法在提升系统性能的同时达到绿色通信的目标；总共产生学术论文22篇，其中SCI论文15篇，国际会议论文7篇，发明专利申请5项，以项目负责人为第一作者发表的高质量国际知名SCI论文8篇。