Collaborative Research: SWIFT: Decentralized Intelligent Spectrum Sharing in UAV Networks (DISH-uNET) via Hardware-software Co-design

智能无人机（uav，或“无人机”）作为提供新功能、扩展无线网络基础设施并使其更具灵活性和弹性的“工具”，正吸引着网络社区的兴趣。无人机辅助无线网络将使现在和未来的物联网（IoT）和5G应用成为可能，并成为战场检查、边境控制和空中监视、精准农业、环境监测、运输和货物交付等新的军事和民用应用的驱动力。目前，尽管无人机网络具有巨大的潜力，但由于无人机的高机动性、有限的功率约束以及非静止环境，无人机网络的设计面临着很大的挑战。在这个项目中，我们将开发毫米波无人机网络（DISH-uNET）中分散智能频谱共享的新方法。该项目将产生重大的工程和社会影响，并大大推进具有强弹性属性的智能无人机网络设计的最新技术。该方案通过(1)特定领域的节能收缩加速器，(2)针对高机动性无人机的新型基于学习的收发器设计，(3)新的分散式频谱共享多访问控制（MAC），实现高效的频谱管理和利用率优化，(4)针对无人机网络的机动弹性毫米波波束学习的快速适应，实现了基于软硬件协同设计的高效率和弹性。这个项目有四个重点。(1)节能收缩加速器，用于同时进行实时信号处理和机器学习。(2)高机动性无人机通信收发器设计。(3)连接Lyapunov优化框架、博弈论和强化学习在分散频谱共享中的应用。(4)移动弹性毫米波波束学习。该研究团队拥有独特的大型无人机测试平台，配备了最先进的毫米波收发器。该试验台将用于对拟议的DISH-uNET系统进行广泛评价。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Intelligent unmanned aerial vehicles (UAVs, or “drones”) are attracting the interest of the networking community as a “tool” to provide new capabilities, to extend the infrastructure of wireless networks and to make it more flexible and resilient. UAV-aided wireless networks will enable present and future Internet of Things (IoT) and 5G applications and be a driver for new military and civilian applications spanning battlefield inspection, border control and aerial surveillance, precision agriculture, environmental monitoring, transportation and delivery of goods. Currently, despite the great potential of UAV networks, the design of such networks faces great challenges due to the high mobility of UAVs, the limited power constraint of UAVs, and the non-stationary environment. In this project, we will develop novel approaches for decentralized intelligent spectrum sharing in mmWave UAV networks (DISH-uNET). This project will have a significant engineering and societal impact and substantially advance the state-of-the-art in the design of intelligent UAV networks with strong resiliency attributes. The proposed DISH-uNET achieves high efficiency and resilience based on hardware-software co-design through (1) domain-specific energy-efficient systolic accelerators, (2) novel learning-based transceiver design for high mobility UAVs, (3) new decentralized spectrum sharing multiple access control (MAC) for efficient spectrum management and utility optimization, (4) fast adaptation of mobility resilient mmWave beam learning for UAV networks. This project has four main thrusts. (1) Energy-efficient systolic accelerator for simultaneous real-time signal processing and machine learning. (2) Transceiver design for high mobility UAV Communication. (3) Bridging Lyapunov optimization framework, game theory, and reinforcement learning in decentralized spectrum sharing. (4) Mobility-resilient mmWave beam learning. The research team has the unique access to a large-scale UAV testbed, equipped with the state-of-the-art mmWave transceivers. This testbed will be used for extensive evaluations of the proposed DISH-uNET system.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.