SpecEES: Toward Spectral and Energy Efficient Cross-Layer Designs for Millimeter-Wave-Based Massive MIMO Networks

SpecEES：面向基于毫米波的大规模 MIMO 网络的频谱和节能跨层设计

• 批准号: 2140277
• 负责人: Jia Liu
• 金额: $ 55万
• 依托单位: Ohio State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-06-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2140277&HistoricalAwards=false
• 关键词: SpecEES; Toward; Spectral; Energy; Efficient

Millimeter-wave (mmWave) and Massive MIMO (M-MIMO) technologies have strong potential to impact future 5G wireless networks and support data rates 50 times greater than the current 4G LTE wireless communications. As such, 5G multi-Gigabit wireless networks are poised to enable a myriad of applications for (e.g., Internet-of-Things, virtual/augmented reality, etc.). However, the highly directional propagation of mmWave signals and the special mmWave hardware requirements introduce fundamental technical challenges for mmWave-based M-MIMO network systems that may require a clean-slate of hardware and software beamforming architectures. In light of these challenges, the goal of this research program is to advance knowledge in both hardware design and theoretical foundations of mmWave and M-MIMO wireless networks. By exploring new hardware-software technologies for mmWave M-MIMO wireless networks, this research program is envisioned to serve a critical need in mmWave communications, signal processing, networking, and control research communities. In terms of broader impacts, the PIs plan to incorporate findings into graduate courses and develop new special topic courses on the fundamentals of mmWave and M-MIMO communication networks. Engagement of undergrad and high school students is also planned with the aim to provide hands-on experience in RF components, communications, networking, control, and signal processing techniques.The proposed research address foundational problems in mmWave large antenna arrays and communication networks, with potential breakthroughs in both theory and practice to enable the success of on future multi-Gigabit wireless communications and associated networking applications. This research program spans broad areas of communications and signal processing to establish a network-level understanding of mmWave M-MIMO networks through a unified research program, which includes the development and exploration of: i) tractable theoretical models, ii) theoretical performance bounds and capacity limits, and iii) low-complexity algorithms. The novelty of this project lies in the joint beam training and scheduling algorithms through the introduction of novel algorithms in the radio-frequency front-end, and in the exploitation of subarray clustering to increase throughput and reduce energy consumption. The PIs' efforts are organized around three interdependent research thrusts: i) Transceiver and beamforming architectures that offer large agility in frequency tuning and high performance at several metrics, ii) Spectral-efficiency optimization algorithms based on mmWave-based subarray clustering, and iii) Energy-efficiency scheduling algorithms based on mmWave-based subarray clustering. In addition to theoretical studies, the PIs plan to validate the analytical techniques and models via extensive simulations, trace-driven emulations, and field tests.

毫米波 (mmWave) 和大规模 MIMO (M-MIMO) 技术具有影响未来 5G 无线网络的强大潜力，并支持比当前 4G LTE 无线通信高 50 倍的数据速率。因此，5G 多千兆位无线网络有望实现多种应用（例如物联网、虚拟/增强现实等）。然而，毫米波信号的高度定向传播和特殊的毫米波硬件要求给基于毫米波的 M-MIMO 网络系统带来了根本性的技术挑战，可能需要全新的硬件和软件波束成形架构。鉴于这些挑战，该研究计划的目标是增进毫米波和 M-MIMO 无线网络的硬件设计和理论基础方面的知识。通过探索毫米波 M-MIMO 无线网络的新硬件软件技术，该研究计划旨在满足毫米波通信、信号处理、网络和控制研究社区的关键需求。 就更广泛的影响而言，PI 计划将研究结果纳入研究生课程，并开发有关毫米波和 M-MIMO 通信网络基础知识的新专题课程。还计划让本科生和高中生参与，目的是提供射频组件、通信、网络、控制和信号处理技术方面的实践经验。拟议的研究解决了毫米波大型天线阵列和通信网络的基本问题，理论和实践方面的潜在突破，使未来多千兆位无线通信和相关网络应用取得成功。该研究计划涵盖通信和信号处理的广泛领域，旨在通过统一的研究计划建立对毫米波 M-MIMO 网络的网络级理解，其中包括开发和探索：i) 易于处理的理论模型，ii) 理论性能范围和容量限制，以及 iii) 低复杂度算法。 该项目的新颖性在于通过在射频前端引入新颖算法来实现联合波束训练和调度算法，以及利用子阵列聚类来提高吞吐量和降低能耗。 PI 的工作围绕三个相互依赖的研究重点进行：i) 收发器和波束成形架构，可在频率调谐方面提供极大的敏捷性并在多个指标上提供高性能，ii) 基于毫米波子阵列聚类的频谱效率优化算法，以及 iii)基于毫米波子阵列聚类的能效调度算法。除了理论研究之外，PI 还计划通过广泛的模拟、跟踪驱动的仿真和现场测试来验证分析技术和模型。

项目成果

NET-FLEET: Achieving Linear Convergence Speedup for Fully Decentralized Federated Learning with Heterogeneous Data

NET-FLEET：利用异构数据实现完全去中心化联邦学习的线性收敛加速

• 发表时间: 2022-10
• 期刊: Proc. ACM MobiHoc
• 作者: Zhang, X.;Fang, M.;Liu, Z.;Yang, H.;Liu, J.;Zhu, Z.
• 通讯作者: Zhu, Z.

Finite-Time Convergence and Sample Complexity of Multi-Agent Actor-Critic Reinforcement Learning with Average Reward," in Proc. ICLR, Virtual Event, April 2022

具有平均奖励的多智能体 Actor-Critic 强化学习的有限时间收敛和样本复杂性，”Proc. ICLR，虚拟活动，2022 年 4 月

• 发表时间: 2022-04
• 期刊: Proc. ICLR
• 作者: Hairi, FNU;Liu, Jia;Lu, Songtao
• 通讯作者: Lu, Songtao

Prometheus: Taming Sample and Communication Complexities in Constrained Decentralized Stochastic Bilevel Learning

普罗米修斯：在受约束的去中心化随机双层学习中克服样本和通信的复杂性

• 发表时间: 2024-09-13
• 作者: Zhuqing Liu;Xin Zhang;Prashant Kh;uri;uri;Songtao Lu;Jia Liu
• 通讯作者: Jia Liu

CFedAvg: Achieving Efficient Communication and Fast Convergence in Non-IID Federated Learning

CFedAvg：在非独立同分布联邦学习中实现高效通信和快速收敛

• DOI: 10.23919/wiopt52861.2021.9589061
• 发表时间: 2021-10
• 期刊: and Wireless Networks (WiOpt
• 作者: Yang, Haibo;Liu, Jia;Bentley, Elizabeth S.
• 通讯作者: Bentley, Elizabeth S.

Interplay between Machine Learning and Networking Systems

机器学习和网络系统之间的相互作用

• 发表时间: 2023-08
• 期刊: IEEE network
• 影响因子: 9.3
• 作者: Chu, Xiaowen;Ibrahim, Shadi;Liu, Jia;Wang, Shiqiang;Wu, Chuan;Zeng, Rongfei
• 通讯作者: Zeng, Rongfei