CIF: Small: Fundamental Limits of Layered Wireless Networks

该研究的目标是确定何时以及如何以无缝和频谱高效的方式在共享和/或免许可频段上分层多个无线网络的基本信息理论限制。无线网络无处不在：WiFi、蓝牙、无绳电话充斥着越来越密集的未经许可的频段。蜂窝网络、卫星网络、军事网络和第一响应网络占用的是独家授权频带，这些频带正逐渐让位于动态的二次频谱共享授权形式。对于这两种类型的许可，多个无线网络以智能、高效和可扩展的方式在相同的有限频谱资源中共存是至关重要的。重新审视如何优化无线网络分层的时机已经成熟，远远超出了目前的“干扰限制”解决方案，该解决方案将正交访问和干扰作为噪声处理相结合。研究从经典的信息理论的单层，同质网络转向包括现实的遗忘约束，不同的层应该如何以分层的方式相互作用。了解如何优化分层网络是消除当前频谱效率低下的重要一步。所开发的分层网络性能的基本界限将揭示网络采用分层策略的激励何时存在，预计这将对未来网络的设计产生重大影响。对分布式和集中式分层网络进行了研究，目标是：1)开发分层网络的结构化框架；2)导出确定性分层网络和相关高斯分层网络在恒定间隙内的容量区域；3)通过考虑分层网络的广义自由度、每层节点数量的吞吐量缩放规律，通过渐近度量来理解分层网络；以及吞吐量随层数的变化规律。

The goal of the research is to determine the fundamental, information theoretic limits of when and how to layer multiple wireless networks on shared and/or unlicensed bands in a seamless and spectrally efficient manner.Wireless networks are everywhere: WiFi, Bluetooth, cordless phones populate increasingly dense unlicensed frequency bands. Cellular, satellite, military and first-responder networks occupy exclusively licensed bands that are gradually giving way to dynamic, secondary spectrum sharing forms of licensing. For both types of licensing, it is crucial that multiple wireless networks co-exist in the same finite spectral resources in an intelligent, efficient and scalable manner.The time is ripe for a fresh look at how to optimally layer wireless networks, reaching far-beyond today's ``interference-limited'' solutions which combine orthogonal access and the treatment of interference as noise. The research moves away from classical information theoretic single-layer, homogeneous networks to include realistic oblivion constraints on how different layers should interact in a hierarchical fashion. Understanding how to best layer networks constitutes a major step towards eliminating current spectral inefficiencies. The fundamental bounds on the performance of layered networks developed will reveal when incentives for networks to adopt a layered policy exist, which is expected to significantly impact the design of future networks.Both distributed and centralized layered networks are investigated, with goals to:1) develop a structured framework for layered networks, 2) derive capacity regions for deterministic layered networks and associated Gaussian layered networks to within a constant gap, 3) understand layered networks through asymptotic metrics by considering the generalized degrees of freedom of the layered network, the throughput scaling laws with number of nodes in each layer, and the throughput scaling law with number of layers.