Collaborative Research: MIMO Networking: From Principles to Protocols

近年来，用于通信网络的复杂的多输入多输出（MIMO）技术已经出现。实现这种技术所承诺的实质性性能增益不仅需要开发合适的网络协议，而且还需要在仔细考虑超越层次结构中各个层的问题的基础上重新思考相关网络架构的各个方面。到目前为止，大多数工作都集中在物理层和介质访问层，强调编码和多路复用问题。然而，人们逐渐认识到，更高网络层的问题至少同样重要且同样丰富，并且相关的更广泛的MIMO网络设计挑战对于实现该技术的性能潜力至关重要。所提出的研究正是针对这些挑战，并将从基本原则开始，以开发有效和强大的网络协议和资源管理算法的关键见解。随着人们对诸如802.16和802.11n等标准的开发的浓厚兴趣，这样的研究尤其及时，并且具有特别高的实际影响潜力。最重要的是调度问题：给定一个MIMO链路，用于将数据传输到一组最终用户，在这样的系统中，吞吐量、延迟、可靠性、公平性和复杂性之间的基本权衡是什么？在我们的初步研究中，我们已经建立了从根本上新的时空调度协议可以而且必须开发，这种协议的结构是显着不同的，从他们的非MIMO同行。在这项研究中，我们将探讨从在任何给定的传输间隔中如何选择用户，到如何利用反馈，再到如何完成训练等问题，并进一步探讨当这种类型的多个MIMO链路相互结合使用时的更丰富的问题，就像在任何实际网络中一样。在这种情况下，合作和中继成为有效的调度和资源管理的组成部分，研究将追求这样的算法的详细发展。在我们的研究中，我们将考虑强网络动态，包括多径衰落和阴影，用户连接和人口的快速波动，以及此类网络中流量的突发性。我们研究的突出特点包括：1）比迄今为止所考虑的更强的最优性概念; 2）超越简单渐近性的更细粒度分析;（3）在协议设计和分析中，将传统的信息论、通信论的处理方法与路由论、调度论和算法论的处理方法紧密结合起来。事实上，更普遍的是，要调查的问题的范围将需要协调应用广泛的设计和分析工具，使研究高度跨学科的性质。实际上，这些工具将从信息论、排队论、通信理论、图论、编码理论、算法和复杂性理论以及估计理论和机器学习中汲取。所提出的研究将显著推进利用MIMO技术的系统中的网络的基本理论和实践，并为开发新一代应用所需的更高性能网络基础设施奠定坚实的基础。更广泛的影响：拟议的研究是一项高度以教育为中心的活动，重点是学生的专业和智力发展。通过推广工作，PI将继续坚定地致力于让来自传统上在工程领域代表性不足的群体的学生参与他们的研究计划。PI还坚定地致力于将这项研究的结果纳入课程，特别是在各自机构正在开发的新网络主题中，这些主题专门集成了网络和物理层设计和分析。最后，该研究计划将有助于促进两个学术机构之间重要的新的实验室间研究合作，这两个学术机构有可能建立强有力的伙伴关系，以及与其他主要工业合作伙伴的合作。特别是，拟议的努力将促进俄勒冈州立大学和麻省理工学院之间在通信和网络领域更紧密的研究和教育联系，并通过与惠普，诺基亚和通用电气的网络实验室的密切互动，迅速将研究成果转化为实践，他们每个人都对有效的学术-工业研究合作有着坚定的承诺和良好的记录。PI将大力利用这些关系以及早期的基础设施投资，例如惠普在过去几年中在麻省理工学院建立的跨学科无线网络中心，以促进麻省理工学院内外的合作。

Recent years have seen the advent of sophisticated multiple-input, multiple-output (MIMO) technology for communication networks. Achieving the substantial performance gains promised by such technology requires not only the development of suitable network protocols, but also rethinking aspects of the associated network architecture based on careful consideration of issues that transcend individual layers in the hierarchy. To date, most effort has focused on the physical and medium-access layers, emphasizing coding and multiplexing issues. However, there is an emerging appreciation that the issues at the higher networking layers are at least as important and equally rich, and that the associated broader view of MIMO network design challenge is critical to realizing the performance potential of this technology. The proposed research is aimed at precisely these challenges, and will start from fundamental principles to the develop key insights from which efficient and robust networking protocols and resource management algorithms will follow. With the strong interest in the development of standards such as 802.16 and 802.11n, among many others, such research is especially timely and the potential for practical impact is particularly high.This broader research theme will be approached by investigating several specific questions and issues. Of central importance is the scheduling problem: given a MIMO link for delivering data to a collection of endusers, what are the fundamental tradeoffs between throughput, delay, reliability, fairness, and complexity achievable in such systems? In our preliminary research, we have established that fundamentally new spatiotemporal scheduling protocols can and must be developed, and that the structure of such protocols is strikingly different from those in their non-MIMO counterparts. In the proposed research, we will investigate issues ranging from how users should be selected in any given transmission interval, to how feedback should be exploited, to how training should be accomplished.The research will further pursue the still richer problems when multiple MIMO links of this type areused in conjunction with one another, as is the case in any practical network. In such settings, cooperation and relaying become integral aspects of efficient scheduling and resource management, and the research will pursue the development of such algorithms in detail. Throughout our investigation, we will take into account the strong network dynamics, including multipath fading and shadowing, rapid fluctuations in user connectivity and populations, and the bursty nature of traffic in such networks.Distinguishing features of our research include its emphasis on 1) stronger notions of optimality than have been considered to date; 2) its finer grain analysis beyond simple asymptotics; and 3) its tight integration of traditional information-theoretic and communication-theoretic treatments with queuing-theoretic, scheduling-theoretic and algorithm-theoretic treatments in the protocol design and analysis. Indeed, more generally the spectrum of issues to be investigated will require the coordinated application of a wide range of design and analysis tools, making the research highly interdisciplinary in nature. Indeed, these tools will be drawn from information theory, queuing theory, communication theory, graph theory, coding theory, algorithms and complexity theory, and estimation theory and machine learning.Intellectual Merit: The proposed research will significantly advance the fundamental theory and practice of networking in systems exploiting MIMO technology, and to provide a solid foundation for the development of the higher performance networking infrastructure needed to enable a new generation of applications.Broader Impacts: The proposed research is a highly education-centered activity with a focus around the professional and intellectual development of students. Through outreach efforts, the PI's will continue their strong commitment to involving in their research programs students from groups traditionally underrepresented in engineering. The PI's are also strongly committed to the integration of the results of this research into the curriculum, and in particular to new networking subjects under development at their respective institutions that specifically integrate network- and physical-layer design and analysis. Finally, the researchprogram will help foster important new interlaboratory research collaboration between two academic institutions for which there is the potential for strong partnership, as well as with additional key industrial partners. In particular, the proposed effort will foster closer research and educational ties between OSU and MIT in the communication and networking area, and rapid transition of the resulting research results to practice through close interaction with the networking laboratories of Hewlett-Packard, Nokia, and General Electric, each of whom has a strong commitment to and track record of effective academic-industrial research collaboration. The PI's will strongly leverage these relationships, as well as earlier infrastructure investments, such as the interdisciplinary Center for Wireless Networking established by HP at MIT over the last several years to foster such collaboration both within and beyond MIT.