EARS: Collaborative Research: Maximizing Spatio-Temporal Spectrum Efficiency in the Cloud

在过去的几年里，无线数据流量急剧增加。为了应对这一趋势，必须大幅提高时空频谱利用率。然而，实现这一目标需要解决几个基本挑战，包括在地理位置数据库通常失效的城市地区发现更多的电视白色空间（TVWS），通过具有过度小区间干扰的网络致密化来提高频谱效率，以及实现从一维频谱共享到多维基础设施共享的转变。该项目的目标是通过系统地利用演进的云无线电接入网络（Cloud-RAN）架构的潜力来应对这些重要挑战。本项目的研究解决方案旨在从根本上解决当前封闭分布式无线接入网频谱效率低下的问题，满足沿着网络应用快速发展和多样化的快速增长的移动的业务需求，为网络用户提供统一、泛在的网络服务。其通过创新的云增强频谱映射、基于云的频谱资源编排和支持虚拟化的动态基础设施共享来最大化时空频谱效率。该项目由四个高度相关的重点组成：（1）迭代贝叶斯决策框架，该框架将贝叶斯空间预测和贝叶斯实验设计相结合，最佳地选择少量的移动的用户及其位置，以使大城市规模的地理定位数据库具有高空间分辨率和高TVWS检测精度;（2）利用该频谱图，提出了一种吞吐量最优的联合分簇调度框架，该框架联合选择远程无线头端（RRH）的分簇模式和网络用户的传输调度，使得每个用户具有有界的平均接入时延，并且通过RRH分簇形成的虚拟基站（VBS）的和速率最大化;（3）开发了新的无线虚拟化工具，可以在公共无线物理基础设施上抽象、切片和实例化多个虚拟网络;（4）通过基于商用软件无线电前端和高性能服务器的实验平台，对所提出的解决方案进行了验证。

Over the last few years wireless data traffic has drastically increased. To encounter this trend, spatio-temporal spectrum utilization has to be dramatically improved. Achieving this goal, however, need to address several fundamental challenges including discovering more TV white spaces (TVWS) in urban areas where geo-location databases generally fail, increasing spectrum efficiency through network densification with excessive intercell interference, and enabling the shift from one-dimensional spectrum sharing to multidimensional infrastructure sharing. The objective of this project is to address these important challenges by systematically exploiting the potential of the evolutionary cloud radio access network (Cloud-RAN) architecture. The research solutions of this project are expected to fundamentally address the spectrum inefficiency of current closed and distributed radio access networks and to meet the demands of fast-growing mobile traffic along with the rapidly-evolving and diverse network applications, thus providing uniform, ubiquitous network services for network users.This project develops a new and holistic spectrum management framework, which maximizes spatio-temporal spectrum efficiency through innovative cloud-augmented spectrum mapping, cloud-based spectral resource orchestrating, and virtualization-enabled dynamic infrastructure sharing. The project consists of four highly interrelated thrusts: (1) an iterative Bayesian decision framework, which coherently combines Bayesian spatial prediction and Bayesian experimental design, which optimally selects a small number of mobile users as well as their locations to enable metropolitan-scale geo-location databases with high spatial-resolution and high TVWS detection accuracy; (2) utilizing such spectrum map, a throughput-optimal joint clustering and scheduling framework is developed, which jointly selects the clustering patterns of remote radio heads (RRHs) and the transmission schedules of network users, such that each user has bounded average queueing delay and the sum-rate of the formed virtual base stations (VBSs) through RRH clustering is maximized; (3) novel wireless virtualization tools are developed, which can abstract, slice, and instantiate multiple virtual networks on a common wireless physical infrastructure; (4) the proposed solutions are demonstrated with an experiment testbed based on commercial software-defined radio frontends and high-performance servers.