基于无人机多层空中网络智能高效的通信覆盖机制研究

To meet the communication link requirements of unmanned aerial vehicles (UAV) UE, multilayered air network architecture based on different types of UAV base stations is proposed to realize the high reliability and low delay. However, due to the nonhomogeneous distribution of UAV UE, the unbalanced load on the intra-layer and on the inter-layer are detrimental to the network performance as it results in a significant level of throughput degradation, and increase in the end-to-end communication delay, which greatly leads to low utilization of wireless resources. Meanwhile, the limited battery energy of drone also brings the efficient utilization problems of the energy resources. Aiming to improve the resource utilization in the multilayered air network, this proposal would focus on the coverage mechanism to implement low delay, high data service and persistent communication of multilayered air network. The major studies include the following research: Firstly, we will deploy the UAV-BS based on inter-layer cooperation in the two-layer air network to intelligently distribute the traffic by jointly controlling the height between the layers and traffic scheduling through deep Q-learning network, which can offload the traffic with low delay and high network throughput; Secondly, we will solve the 3D cell association area partition optimization problem using the powerful mathematical framework of optimal transport theory to realize efficient and fair coverage in the LAP layer air networks; Thirdly, by means of the actor critic learning theory, we will set up an efficient transfer actor-critic learning model to effectively enhance the learning efficiency, then by using of the distributed learning method in the air network with energy harvesting to intelligently allocate power and subchannels to achieve the optimal energy efficiency goals and ensure the persistent air communication.

为了满足无人机终端的通信链路需求，不同类型的无人机基站协作构成多层空中网络架构被提出以实现高可靠低时延通信。然而由于无人机终端分布不均匀，多层空中网络层内、层间负载均衡问题造成无线资源利用率低下，同时有限的电池能量也带来能源高效利用的问题。本课题以提高空中网络资源利用率为目标，研究无人机多层空中网络智能高效的通信覆盖机制。研究1）针对HAP-LAP双层空中网络建立基于层间协作的无人机基站部署算法，通过深度Q-learning算法决策层间高度和流量调度策略，控制层间流量分布实现低时延高吞吐量的层间流量分载；2）针对LAP层空中网络建立基于最优传输理论的3D小区关联区域划分，实现高效公平的小区关联覆盖；3）建立基于迁移Actor-Critic学习的无线资源分配机制，考虑无人机基站的能量收集功能，智能地分配功率和子信道，提升学习效率且最大化空中覆盖能效，实现持续通信的空中网络。

基于无人机多层空中网络智能高效的通信覆盖机制项目针对空中网络资源利用率低、有限能量下无人机持续通信以及无人机安全等方面的问题，提出了多无人机协作的高效、公平部署策略，通过设计深度强化学习算法，智能决策无人机能量补给和通信服务，优化无人机飞行轨迹，实现公平覆盖下可持续通信；优化无人机的悬停位置和无人机飞行轨迹以及用户关联问题，实现有限能量下多无人机协同数据采集任务，减少数据采集时间，提高数据采集效率；提出基于最优传输理论的3D小区关联区域划分，将连续非凸问题转换为线性凸问题求解，实现服务公平且最大化数据传输服务；提出基于迁移Actor-Critic学习的无线资源分配机制，考虑无人机基站的能量收集功能，智能地分配功率和子信道，提升学习效率且最大化空中覆盖能效，实现持续通信的空中网络；提出了安全数据传输策略，保证了无人机网络中用户信息传输的安全性。项目取得了一系列的研究成果：已发表论文11篇，其中SCI论文5篇，EI论文6篇；已申请专利4项；参与国际交流2次; 提次ITU提案1篇。项目成果为无人机可持续通信覆盖提供了解决方案，为无人机通信应用推广提供了技术基础，推动空天地一体化网络的发展；培养了一大批创新型人才，显著提升了我国自主创新水平、核心技术竞争力。