ERI: Empowering Data-Driven Resource Management in Indoor 5G+ Wireless Networks

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。数据流量的未来趋势需要每秒数千兆比特的高质量无线连接速率和小于10毫秒的延迟。然而，传统的无线电频谱非常拥挤，因此无法满足如此高的要求。因此，第五代及以上（5G+）无线网络将采用未使用的高频频段。然而，高频段的无线连接受到用户移动性引起的频繁中断的挑战。最近的研究表明，基于人工智能的先进网络管理技术可以与用户移动性保持可靠的高质量链接。然而，要发展如此先进的技术，需要全面、高精度的无线信道质量数据集。不幸的是，这些数据集对研究界来说是不可访问的。该项目的第一个目标是开发一个逼真且高精度的模拟器，该模拟器可生成400 - 800太赫兹频率范围内5G+无线信道的丰富数据集。该模拟器将向公众开放，以支持数据驱动的5G+网络管理解决方案的研究工作。生成的数据集将通过最先进的测试平台进行验证。此外，生成的数据集将被表征，以了解用户移动模式对无线信道质量的影响。此外，将开发新的方法来预测由于用户移动性而导致的信道质量，这将进一步有助于开发有效的5G+网络管理工具。通过增强未来数据驱动型网络管理解决方案的研究能力，该项目将实现5G+无线网络中高频频段的广泛集成。因此，该项目支持智能互联社区和万物互联时代急需的高速率低延迟5G+技术。因此，该项目广泛地影响了不断发展的数字社会的各个方面，特别是室内移动应用。此外，该项目提供非常理想的多学科技能的劳动力培训，同时确保妇女和代表性不足群体的参与。5G+无线网络将在未使用的高频频段运行，例如可见光（VL）频段（400 - 800太赫兹）。虽然它们可以支持超高吞吐量和超低延迟流量需求，但这些频段的无线信道的衍射能力有限。这将导致由于静态和/或移动对象的阻塞而导致用户移动通信链路频繁中断。初步研究表明，使用一般概率分布模型描述这些链路中断是不实际的，因为此类中断与环境限制的用户移动性细节有关。因此，经典的优化工具对于5G+网络管理将无效。另一方面，数据驱动策略可以设计智能网络管理策略，从环境中学习，自适应地为移动用户分配资源。然而，采用数据驱动的网络管理策略面临以下挑战：1)缺乏高质量的室内移动VL信道增益数据集；2)VL信道增益数据的稀疏性，阻碍了传统数据驱动工具的采用。为了解决这些限制，在考虑办公空间布局的同时，拟议的项目追求以下研究重点：T1)开发高效的5G+移动信道模拟器，该模拟器反映了VL频段的真实时空特征，并捕获了由于物体和用户身体的动态阻塞而导致的链路不可用的影响。该模拟器将向公众开放，为5G+数据驱动网络管理的进一步研究提供支持；T2)开发高效的5G+信道预测器，尽管信道数据集具有高稀疏性，但该预测器可为未来时间框架提供有用的VL信道状态信息。该预测器将为各种主动数据驱动的5G+网络管理策略提供支持。本项目中开发的方法和工具将通过模拟基于虚拟现实的室内网络设置的测试平台进行验证。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。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).Future trends in data traffic require high-quality wireless connections of multi-gigabits per second rates and less than ten milliseconds delay. However, the conventional radio spectrum is quite congested, and hence, it cannot satisfy such high demands. Consequently, unused high-frequency bands will be adopted in the fifth generation and beyond (5G+) wireless networks. Yet, wireless connectivity in high bands is challenged by frequent outages induced by user mobility. Recent studies show that advanced network management techniques based on artificial intelligence can maintain a reliable high-quality link with user mobility. However, to develop such advanced techniques, comprehensive highly-accurate datasets of wireless channel quality are required. Unfortunately, these datasets are not accessible to the research community. The first goal of this project is to develop a realistic and highly-accurate simulator that generates rich datasets of 5G+ wireless channels in the frequency range 400 – 800 Terahertz. This simulator will be made publicly available to empower research efforts in data-driven 5G+ network management solutions. The generated datasets will be validated through a state-of-the-art testbed. Moreover, the generated datasets will be characterized to learn the impact of user mobility patterns on the wireless channel quality. In addition, novel methods will be developed to predict the channel quality due to user mobility, which will further help in developing effective 5G+ network management tools. By empowering future research in data-driven network management solutions, this project enables a broad integration of high-frequency bands in 5G+ wireless networks. As a result, this project supports high-rate low-delay 5G+ technologies, much needed in the era of smart and connected communities and the internet of everything. Thereby, this project broadly impacts myriad aspects of the evolving digital society, particularly, for indoor mobile applications. Furthermore, this project provides workforce training in a highly desirable multi-disciplinary skillset while ensuring the participation of women and underrepresented groups.The 5G+ wireless networks will operate in the unused high-frequency bands, e.g., the visible light (VL) frequency band (400 – 800 Terahertz). While they can support ultra-high throughput and ultra-low latency traffic demands, the wireless channels at such bands suffer from limited diffraction capabilities. This results in frequent outages in communication links with user mobility due to blockage from static and/or mobile objects. Preliminary studies demonstrated that it is not practical to describe these link outages using a general probability distribution model, as such outages are tied to the environment-confined user mobility details. As a result, classical optimization tools will be ineffective for 5G+ network management. On the other hand, data-driven strategies can be used to design intelligent network management strategies that learn from the environment and adaptively allocate resources to the mobile users. However, adopting data-driven network management strategies is challenged by: 1) the absence of high-quality datasets of indoor mobile VL channel gains and 2) the sparsity of the VL channel gain data, which impedes the adoption of conventional data-driven tools. To address these limitations, the proposed project pursues the following research thrusts while considering office room layouts: T1) Development of efficient 5G+ mobile channel simulator that reflects realistic spatio-temporal features in the VL band and captures the impact of link unavailability due to dynamic blockages with the objects and users' bodies. The simulator will be publicly available to empower further research in 5G+ data-driven network management; T2) Development of an efficient 5G+ channel predictor that provides useful VL channel state information for future time frames despite the high sparsity in the channel dataset. The predictor will empower various proactive data-driven 5G+ network management strategies. The developed methods and tools in this project will be validated through a testbed that mimics a VL-based indoor networking setup.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.