Collaborative Research: CNS Core: Small: AirEdge: Robust Airborne Wireless Edge Computing Network using Swarming UAVs

合作研究：CNS 核心：小型：AirEdge：使用集群无人机的强大机载无线边缘计算网络

• 批准号: 2006604
• 负责人: Yu Wang
• 金额: $ 16.66万
• 依托单位: Temple University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2006604&HistoricalAwards=false
• 关键词: Collaborative; Research; CNS; Core; Small

Unmanned aerial vehicle-aided networks have been widely recognized by both cellular and internet industry and academia as an emerging technique to enhance current networking infrastructure. The research objective of this project is to design, analyze, and evaluate AirEdge, which is an innovative robust 3D airborne computing and networking system that exploits a swarm of aerial mobile radio access points and edge servers carried or deployed by unmanned aerial vehicles. The unique feature of AirEdge is to enable fast-deployable highly efficient on-demand edge computing and networking services. AirEdge will enable a series of applications in the areas of disaster rescue, public safety, anti-terrorism, battlefield assistance, and mobile entertainment. For example, AirEdge can be rapidly deployed to the area impacted by the disaster and allows first responders to locate and identify injured people using face recognition and provide their corresponding health information for first aids. This project also fosters interdisciplinary research and provides a unique training program for undergraduate and graduate students.This project aims to realize AirEdge through the communication-motion co-design principles for 3D airborne networking and communication-computation co-design principles to enable reliable and energy-efficient airborne edge computing. Toward this end, two fundamental research problems are investigated: 1) how to dynamically establish edge computing networks to enable flexible edge computing and 2) how to integrate dynamic computing resource deployments with the communication network provided by unmanned aerial vehicle (UAVs) to enable low-latency and high-performance edge computing on resource-constrained computing platforms. To address these problems, 1) a new edge-assisted optimal motion control scheme is designed to exploit the abundant computation power of the ground edge server and high-fidelity ray-tracing simulations to perform site-specific Air-to-Ground channel modeling; 2) a new energy-efficient motion planning strategy is developed for the UAV swarm with an objective to simultaneously enhance the area spectral efficiency of the entire serving site and satisfy the time-varying data rate requirements of the edge computing applications; 3) a novel multi-agent actor-critic (MA-AC) reinforcement learning method is developed to realize a more adaptive and robust model-free control scheme under the uncertainties of the deployment environment; 4) a new context-aware adaptive edge computing deployment solution is designed to optimally integrate the airborne edge computing with the airborne communication network; 5) a novel dynamic edge analytics framework is engineered for the airborne communication and computing. The framework leverages approximate computing to mitigate the tradeoff between computation quality, service latency, and energy efficiency in AirEdge.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.

无人机辅助网络作为一种增强当前网络基础设施的新兴技术，已经得到了蜂窝和互联网行业和学术界的广泛认可。该项目的研究目标是设计、分析和评估AirEdge，这是一种创新的强大的3D机载计算和网络系统，利用无人机携带或部署的大量空中移动无线接入点和边缘服务器。AirEdge的独特之处在于能够快速部署高效的按需边缘计算和网络服务。AirEdge将在灾难救援、公共安全、反恐、战场援助和移动娱乐等领域实现一系列应用。例如，AirEdge可以快速部署到受灾害影响的地区，并允许第一响应者使用面部识别来定位和识别受伤人员，并为急救提供相应的健康信息。该项目还促进跨学科研究，并为本科生和研究生提供独特的培训计划。本项目旨在通过三维机载网络的通信-运动协同设计原则和通信-计算协同设计原则实现AirEdge，以实现可靠和节能的机载边缘计算。为此，研究了两个基础问题：1)如何动态建立边缘计算网络，实现灵活的边缘计算；2)如何将动态计算资源部署与无人机提供的通信网络相结合，在资源受限的计算平台上实现低延迟、高性能的边缘计算。为了解决这些问题，1)设计了一种新的边缘辅助最优运动控制方案，利用地面边缘服务器丰富的计算能力和高保真的光线跟踪模拟来执行特定地点的空对地通道建模；2)提出了一种新的无人机群节能运动规划策略，以同时提高整个服务站点的区域频谱效率，满足边缘计算应用对时变数据速率的要求；3)提出了一种新的多智能体actor-critic （MA-AC）强化学习方法，在不确定的部署环境下实现了一种更具自适应性和鲁棒性的无模型控制方案；4)设计了一种新的上下文感知自适应边缘计算部署方案，将机载边缘计算与机载通信网络进行优化集成；5)为机载通信与计算设计了一种新的动态边缘分析框架。该框架利用近似计算来减轻AirEdge中计算质量、服务延迟和能源效率之间的权衡。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。

项目成果

Participant Selection for Hierarchical Federated Learning in Edge Clouds

• DOI: 10.1109/nas55553.2022.9925313
• 发表时间: 2022-10
• 期刊: 2022 IEEE International Conference on Networking, Architecture and Storage (NAS)
• 作者: Xinliang Wei;Jiyao Liu;Xinghua Shi;Yu Wang

EdgeML: Towards Network-Accelerated Federated Learning over Wireless Edge

• DOI: 10.1016/j.comnet.2022.109396
• 发表时间: 2021-10
• 期刊: Comput. Networks
• 作者: Pinyarash Pinyoanuntapong;Prabhu Janakaraj;Ravikumar Balakrishnan;Minwoo Lee;Chen Chen-Chen;Pu Wang

Data Placement Strategies for Data-Intensive Computing over Edge Clouds

边缘云上数据密集型计算的数据放置策略

• DOI: 10.1109/ipccc51483.2021.9679438
• 发表时间: 2021
• 期刊: 40th IEEE International Performance Computing and Communications Conference
• 作者: Wei, Xinliang;Mohaimenur Rahman, A B;Wang, Yu
• 通讯作者: Wang, Yu

Joint Optimization Across Timescales: Resource Placement and Task Dispatching in Edge Clouds

• DOI: 10.1109/tcc.2021.3113605
• 发表时间: 2023-01
• 期刊: IEEE Transactions on Cloud Computing
• 影响因子: 6.5
• 作者: Ieee A B M Xinliang Wei Student Member;Mohaimenur Rahman;Ieee Dazhao Cheng Member;Ieee Yu Wang Fellow;Xinliang Wei

Dystri: A Dynamic Inference based Distributed DNN Service Framework on Edge

• DOI: 10.1145/3605573.3605598
• 发表时间: 2023-08
• 期刊: Proceedings of the 52nd International Conference on Parallel Processing
• 作者: Xueyu Hou;Yongjie Guan;Tao Han