Collaborative Research: SpecEES: Designing A Spectrally Efficient and Energy Efficient Data Aided Demand Driven Elastic Architecture for future Networks (SpiderNET)

可由移动的蜂窝网络使用的无线电频谱是有限的，但是蜂窝业务继续猖獗地增长。这就要求科学界设计出能够将频谱效率推向极限的网络。另一方面，使蜂窝网络节能是一个比以往任何时候都更加紧迫的目标，不仅是为了降低运营成本，而且是为了最大限度地减少膨胀的信息和通信技术行业的碳足迹。 最近的研究表明，除非在蜂窝架构中添加新的自由度和基于人工智能的动态适应性，否则频谱效率的任何显著提高都必须以能量效率为代价。该项目的总体目标是通过最先进的测试平台设计、表征、优化和验证一种新的架构，该架构使额外的自由度和智能能够在移动的网络的设计和操作中动态地利用这些新的自由度，从而同时在频谱效率和能源效率方面产生实质性的收益。 该架构被命名为SpiderNET：面向未来网络的频谱高效和节能数据辅助需求驱动弹性架构。SpiderNET背后的关键思想是引入额外的自由度，以放松严格的频谱效率-能源效率权衡，从而实现两者的同时增强。随着万物互联的到来，作为网络资源效率、增强电池寿命和服务水平的关键推动者，SpiderNET必将对依赖无线连接的不断发展的数字社会的几乎每个方面产生广泛的影响。此外，由于每比特收入的减少已经促使蜂窝运营商减少能源费用，SpiderNET实现的巨大节能可以大幅降低OPEX。减少蜂窝工业的碳足迹也是拟议研究的一个关键好处。该项目提供了一个非常抢手的多学科技能的劳动力培训，同时确保妇女和其他代表性不足的群体的参与，以及K-12的推广。与仅基于理论或模拟的研究相比，拟议研究的一个关键区别是对尖端细胞测试平台的实验研究，预计将产生更广泛的影响。该项目是与蜂窝生态系统的主要国家和国际利益相关者合作开展的，以确保各行业和政府机构及时适应项目成果。通过将运营重点从以基站为中心的刚性始终在线的小区转移到以用户为中心的按需小区，实现频谱效率和能源效率的同时提高。 为了实现这一点，SpiderNET由一层低密度的大面积控制基站组成，下面是高密度的可切换数据基站。开启/关闭数据基站、以用户为中心的小区（S区）的大小和其他参数由基于机器学习的自组织网络（SON）引擎主动地编排，该SON引擎利用数据和控制基站处的所选测量的数据库。 初步研究表明，S区的大小和活动数据基站密度两者的智能编排以及数据库的内容和时空分辨率的最佳设计沿着可以在不损害体验质量的情况下显著提高频谱效率和能量效率两者。该研究将通过以下三个研究重点将SpiderNET从一个想法转变为一个功能架构：1）开发分析和仿真模型，以充分表征SpiderNET的频谱效率和能源效率，以确定可以优化的关键设计参数，以最大限度地提高其频谱效率和能源效率增益。然后，这些模型将被用来设计算法，以最大限度地提高频谱效率和能源效率在动态交通条件。2)设计测量数据库并在控制和数据基站处利用该数据来开发用于主动小区发现和选择以及无线电资源分配的算法，以在不损害体验质量的情况下联合地最大化频谱效率和能量效率。3)在TurboRAN（NSF资助的端到端可编程测试平台）上验证SpiderNET的概念。该研究利用流体建模、随机几何、博弈论、机器学习以及随机和多目标优化等领域的工具来实现其目标。该奖项反映了NSF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。

Radio spectrum useable by mobile cellular networks is finite but cellular traffic continues to grow rampantly. This calls on scientific community to design networks that can push spectrum efficiency to its limits. On the other hand, making cellular networks energy efficient is a goal that is becoming more pressing than ever not only for operational cost reduction but also for minimizing the carbon foot print of the bulging information and communications technology industry. Recent studies show that unless new degrees of freedom and artificial intelligence based dynamic adaptability is added in the cellular architecture, any significant gain in spectral efficiency must come at cost of energy efficiency. The overarching goal of this project is to design, characterize, optimize and validate through a state-of-the-art testbed a new architecture that enables the additional degrees of freedom and intelligence to dynamically exploit these new degrees of freedom in the design and operation of the mobile network to yield substantial gains in both spectrum efficiency and energy efficiency, simultaneously. This proposed architecture is named as SpiderNET: Spectrally Efficient and Energy Efficient Data Aided Demand Driven Elastic Architecture for Future Networks. The key idea behind SpiderNET is to introduce additional degrees of freedom to relax the rigid spectrum efficiency-energy efficiency tradeoff and thus enable simultaneous enhancement of both. In wake of the internet of everything, as a key enabler of network resource efficiency, enhanced battery life and service level improvement, SpiderNET is bound to have a broad impact on nearly every aspect of evolving digital society that counts on wireless connectivity. In addition, as diminishing revenue per bit is already pushing cellular operators to reduce energy bills, huge energy savings enabled by SpiderNET can substantially reduce OPEX. Reducing the carbon foot print of cellular industry is also a key benefit of the proposed research. This project offers workforce training in a highly sought-after multi-disciplinary skill set while ensuring the participation of women and other underrepresented groups, and K-12 outreach. Compared to only theoretical or simulation-based research, a key distinction of the proposed research is the experimental research on a cutting-edge cellular testbed that is expected to cast a much broader impact. This project is collaborative undertaking with key national and international stakeholders in cellular eco-systems to ensure timely adaptation of the project outcomes by respective industry and government bodies.The simultaneous enhancement of both spectrum efficiency and energy efficiency is achieved by shifting the pivot of operation from the rigid always-on base-station-centric cells to user-centric on-demand cells. To enable this, SpiderNET consists of a layer of low-density large footprint control base station underlaid by high-density switchable data base stations. The switching on/off the data base station, the size of user-centric cells (S-Zones) and other parameters are orchestrated proactively by a machine learning based self-organizing network (SON) engine that leverages a database of selected measurements at data and control base stations. Preliminary studies show that intelligent orchestration of the both, the size of the S-Zone and active data base station density, along with optimal design of the contents and spatiotemporal resolution of the database can substantially enhance both spectrum efficiency and energy efficiency without compromising quality of experience. The research will transform SpiderNET form an idea into a functional architecture by pursuing the following three research thrusts: 1) Developing analytical and simulation models to fully characterize the spectrum efficiency and energy efficiency of SpiderNET to determine the key design parameters that can be optimized to maximize its spectrum efficiency and energy efficiency gains. These models will then be leveraged to design algorithms for maximizing spectrum efficiency and energy efficiency in dynamic traffic conditions. 2) Designing the database of measurements and leveraging this data at control and data base stations to develop algorithms for proactive cell discovery and selection and radio resource allocation for jointly maximizing both spectrum efficiency and energy efficiency without compromising quality of experience. 3) Proof of concept of SpiderNET on TurboRAN (an NSF-funded end-to-end programmable testbed). This research leverages tools from domains of fluid modelling, stochastic geometry, game theory, machine learning and stochastic and multi-objective optimization to achieve its goals.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.