面向实时业务延迟需求的多跳无线网络数据传输跨层控制机制研究

Cross-layer control is s a new research trend of data transmission control in multi-hop wireless networks. Compared with traditional transmission control mechanisms, cross-layer control mechanism is more suitable for multi-hop wireless networks, and has been proved to be able to achieve both the maximum network capacity region and the maximum network utility. However, one serious problem of cross-layer control is that it leads to large end-to-end delay. On the other hand, the real-time applications in the Internet are becoming the dominating applications and are still growing rapidly. As we know, real-time applications usually have tight requirements on network delay. Cleary, there is a sharp contradiction between the delay requirements of real-time applications and the large delay problem caused by cross-layer control mechanism. How to quantitatively understand the reason of the large delay problem, and further to solve the large delay problem, is the aim of this proposal. In this proposal, we plan to first study the coupling relationship of the dual variables of the transport optimization problem and the queue length on the network nodes, based on optimization theory and the queuing networks theory. With the thorough understanding of the relationship, we can design flow-grained delay analysis model. On this basis, we will design innovative cross-layer control mechanism for delay sensitive real-time applications. In particular, we will study the delay differentiation technology from the perspective of scheduling, and the transmission path optimization technology from the perspective of routing. Combined with the flow-grained delay analysis mode, the innovative cross-layer control mechanism will be able to perform controllable adjustment for delays of real-time applications, so as to guarantee the delay performance of real-time applications, while simultaneously maintains the network utility ?maximization..

跨层控制作为适用于多跳无线网络的数据传输控制新机制，能达到最大的网络容量空间和网络效用。然而，跨层控制的一个严重问题是它会导致很大的端到端延迟。随着网络中实时业务的快速增长，跨层控制导致大延迟的问题与实时业务的延迟需求形成了尖锐的矛盾。如何定量理解跨层控制导致大延迟的原因，进而设计面向实时业务延迟需求的数据传输跨层控制机制，是本课题拟研究的核心问题。本课题拟首先基于优化理论和网络排队论，深入研究数据传输优化问题的对偶变量与节点队列长度间的耦合关系，提出业务流粒度的延迟分析模型。在此基础上，本课题将研究面向实时业务延迟需求的创新的数据传输跨层控制机制，分别从调度和路由的层次实现业务流间的延迟区分和实时业务流的传输路径优化，在保持网络效用最大化的同时，结合所提出的延迟分析模型对业务流粒度的延迟进行可控性调整，实现实时业务的延迟保障。

随着网络中实时业务的快速增长，跨层控制导致大延迟的问题与实时业务的延迟需求形成了尖锐的矛盾。如何定量理解跨层控制导致大延迟的原因，进而设计面向实时业务延迟需求的数据传输跨层控制机制，是本课题拟研究的核心问题。围绕这一问题，本课题对对多跳无线网络数据传输控制机制进行了深入的研究，重点对兼顾延迟与网络效用的跨层控制机制进行了系统性的研究。课题研究内容包括以下三个方面：.1) 兼顾延迟与网络效用的跨层控制：本课题在多网关多跳无线网络中跨层控制的问题模型、优化求解算法、算法性能理论证明、仿真实验评估多个层次开展了系统深入的研究。我们提出了兼顾延迟与网络效用的算法CLC_DGS_DD，证明了CLC_DGS_DD能够达到最大网络效用；同时该算法提供了一个灵活的控制框架来动态地调整各个业务流的延迟优先级，可实现不同业务流的线性延迟区分，从而可保障实时业务的延迟需求。.2) 物理干扰模型下多跳无线网络分布式链路调度算法：物理干扰模型是一种更准确反映无线通信特性的干扰模型。本课题对物理干扰模型下分布式链路调度算法进行了深入研究，提出的局部调度算法在oblivious和uniform发送功率形式下分别保证所达到的网络的整体吞吐为网络所能支持的最大吞吐的常数近似和对数近似。本工作为实际多跳无线网络链路调度问题提供理论保证，将有助于部署实际的多跳无线网络并提高网络整体性能。.3) 多终端协同实时流媒体协同传输调度：在保障实时业务的服务质量方面，除了上述传输层次（包括调度、路由、速率控制）的优化工作外，本课题也对多终端应用层协同的方法进行了初步的探索：提出了一种流媒体下载任务协同算法，该算法能够实现多个移动终端间的流媒体协同下载与分享，既能够确保流媒体应用流畅的用户体验，又能够保证整个协同过程所消耗蜂窝网流量的公平性。.在本项目的开展过程中，我们共发表论文14 篇（SCI论文4篇，EI论文10篇），其中顶级国际学术期刊和会议论文包括IEEE transactions on Parallel and Distributed Systems 1篇 (CCF A类期刊)，IEEE transactions on Vehicular Technologies 1篇，ACM MobiCom 1篇（CCF A类会议），ACM MobiHoc 1篇；并取得授权专利2项，受理专利4项。