网络编码在无线通信中的应用研究

【作者】 王宁；

【导师】 林家儒；

【作者基本信息】 北京邮电大学 ， 信号与信息处理， 2009， 博士

【摘要】 1956年,Feinstein和Shannon指出：“通信网络端对端的最大信息流是由网络的最小割决定的”,但是传统路由器的存储转发模式根本不可能达到香农最大流最小割定理的上界。直到本世纪初,香港中文大学的R. Ahlswede等人在IEEE Transactions on Information Theory上发表了一篇题为Network information flow的文章中,基于网络信息流的概念提出了网络编码的思想。他们指出网络节点对不同的信息流进行编码组合,可以获得网络组播速率的最大流限,而通过传统的存储和转发不能获得该最大流限。为了研究网络编码在无线通信系统中的性能,.从网络编码的构造方法入手,分析了线性网络编码、随机网络编码和代数型网络编码的具体实现,尤其对随机线性网络编码做了深入研究,并通过实际例子进行详细论述。在以上理论工作的基础上,接下来研究了网络编码应用在几种无线通信环境下的性能增益,构造通信系统模型,提出不同的算法,设置不同的参数环境,论证网络编码的应用性能,得到理论结果,并仿真验证。主要结论及创新点如下：研究网络编码在中继网络及协作通信网络中的性能,设计了三种Turbo式网络编码系统结构,并分别与不使用网络编码的相应系统性能作比较,突出了网络编码带来的误码性能增益。在此基础上,提出信道编码与网络编码联合设计的模式应用在协作分集通信中的观点,网络编码可以与LDPC码、卷积码以及空时码等联合编码,带来误码性能及吞吐量等方面的进一步改善,并给出了仿真验证。重点论述了网络编码在分布式存储中的性能,针对分布式存储中的一个重要参数-信源节点的度,从理论的角度推导出信源节点的度与恢复数据所需查询的存储节点数量之间的折衷关系,并给出了相应的定理和公式。这一理论公式也直接体现了网络编码代价及其带来的增益之间的折衷关系,具有现实指导意义。最常用在分布式存储中的码型是纠删码(erasure code),如RS码、数字喷泉码(LT码,Raptor码)。针对这些码型的局限性以及网络编码的优势,研究了网络编码在分布式存储系统中的应用。以传感器网络模型为例,对分布式网络中实施网络编码所带来的系统能耗、吞吐量等方面的增益做了深入研究,设置不同的网络环境,分析了几种仿真参数下的性能,论证了网络编码在分布式存储中富有潜力的应用前景。为了分析随机线性网络编码在无线Mesh网络中进行文件共享时的性能增益,分三种情况仿真：上传数据受限且MAC层理想无冲突情况下,阐明了随机线性网络编码对文件共享成功率的增益；MAC层存在冲突情况下论证了随机线性网络编码对文件共享下载时间的增益；节点移动且MAC层存在冲突情况下,分析了随机线性网络编码对文件共享下载时间的增益,都分别与传统路由的情况进行比较,突出网络编码所带来的性能改善。更多还原

【Abstract】 In 1956, Elias, Feinstein and Shannon draw a conclusion that the communication network end-to-end maximum flow is decided by the network minimum cut.But storing-forwarding (SF) of traditional router cannot achieve the upper bound of information flow in maximum flow minimum cut theorem. Network coding is introduced by R.Ahlswede in the "Network information flow" published by IEEE Transactions on Information Theory in 2000, which is based on the conception of network in formation flow. The max-flow bound of network multicast can be achieved if the nodes encode ifferent information flow, which can not always be achieved by traditional SF mode.To research the performance of network coding in wireless communications networks, starting from the basic principles and the structure of network coding, analyses linear network coding, random network coding and algebraic network coding.Especially studies the design of random linear network coding.Basic on above network coding theory, clarifies gain of network coding applying in wireless communication, provides system models, demonstrates several algorithm, sets up different simulation parameters, proves the application performance, concludes the theory results and simulation analysis.The main work and innovation of the dissertation includes the following:In order to study the performance of network coding in relay and cooperative communication networks,three stuctures of Turbo-network coding are provided, and compares the BER gain of system with network coding and that of without network coding.On the basis of Turbo-network coding,puts forward the application of joint channel and network coding in cooperative diversity communication. System performance of BER and throughput can be further improved via joint LDPC,convolutional code, or space-time code and network coding, and confirms the results via simulation diagrams.The performance of network coding in distributed storage is studied. An important parameter-source node degree is emphased, develops the tradeoff relationshio between source node degree and the number of queried nodes brought out as theorem and formula. This theory formula reflects the tradeoff betweem network coding cost and gain, which has practical guiding value.Erasure code, such as RS code, LT code and Raptor code, is the most commonly used in distributed storage. According to the limitations of erasure code and the advantages of network coding, research the application of network coding in distributed storage. Devoloping a sensor network model as an instance, analyzes network coding gain on energy consumption and system throughput. Furtherly, combines with node degree, sets up several network conditions, demonstrates the performance and confirms the potential future of network coding applying in distributed storage network.To analyze the gain of down-load time and success rate brought by random linear network coding in wireless mesh nework, set up three simulation conditions:upload-limited but MAC layer be ideal conflict-free, MAC layer be conflict and mobile nodes and MAC be conflict. Simulation results compare the performance between the system with random network coding and the system with traditional route.更多还原