【作者】 程鹏；

【导师】 仇佩亮； 张朝阳；

【作者基本信息】 浙江大学 ， 通信与信息系统， 2008， 博士

【摘要】 随着人们对随时随地进行自由通信的渴求越来越强烈,有限的无线资源与多媒体业务不断提高的服务质量要求之间的矛盾日益尖锐,无线资源分配技术在一定程度上解决了这个问题。然而,随着下一代无线通信技术的飞速进步和业务需求的爆炸式增长,各国学者正在逐渐把多跳中继,异构网络等新技术引入到未来的无线网络中,这一方面形成了以多跳蜂窝网,无线Mesh网以及异构网络为代表的新一代无线网络,解决了传统无线网络一直存在的可扩展性低和健壮性差等诸多问题,但是另一方面也对传统的无线资源分配技术提出了更高的要求。事实上,在这些新型的无线网络中,跨层资源分配是不可或缺的,也是未来无线网络协议设计的必然趋势。本学位论文以下一代通信系统中最重要的三种无线网络形态为研究背景,在信息论、网络理论和凸优化理论的指导下,研究了跨层资源分配的问题。主要从以下几个方面展开了深入的研究,概括如下:研究了无线蜂窝网络中的跨层资源分配问题,并针对传统蜂窝网络和多跳蜂窝网络这两种不同的情况分别提出了两种不同的跨层资源分配策略。对于传统蜂窝网络,考虑了一个结合自适应调制和自动重传请求的蜂窝下行系统,研究了QoS保证业务的频谱效率最大化问题,并利用拉格朗日乘数法得到了自适应调制参数的最优选择算法。对于多跳蜂窝网络,考虑了一个采用自适应调制,并且数据缓冲区长度受限的两跳链路。首先提出一种跨层多跳排队分析模型,这种模型可以分析上述系统的吞吐量和数据包时延性能。然后研究了在保证QoS业务端到端时延需求的情况下,如何对上述系统进行联合最优的功率分配和带宽分配,从而最大化网络吞吐量。最后基于前面提出的多跳排队分析模型,分别提出了跨层功率分配算法、跨层带宽分配算法以及性能更好的功率、带宽联合分配算法,联合分配算法可以迭代地寻找最优的功率—带宽变量对,从而最大化网络吞吐量。研究了无线Mesh网络中的跨层资源分配问题,考虑了一个存在多个实时视频业务的无线Mesh网络,研究了如何通过联合的信源编码、功率控制、ARQ控制以及端到端时延分配而最大化视频接收质量的问题。首先对该问题涉及到的应用层、网络层、链路层和物理层进行了简单而合理的数学建模,然后将上述复杂的通信问题转化为一个最优化问题,并利用凸优化的性质证明了这个问题是一个log-convex的问题,从而可以用经典的内点法得到其全局最优解。随后又利用拉格朗日对偶法将原来复杂的优化问题分解为三个简单的子问题:信源编码问题、功率控制问题、端到端时延分配问题,这三个子问题都可以在无线Mesh网络中分布式地解决,并且可以用子梯度更新法来更新链路价格,从而得到全局最优解。研究了认知无线网络(Cognitive Radio Network)中的跨层资源分配问题,包括异构网络共存问题和动态频谱接入问题两个部分。对于异构网络共存问题,首先从多用户信息论的角度研究了当3G蜂窝网同2G蜂窝网共存时,认知用户在主用户QoS约束下的速率可达区域,以及获得和速率最大点的最优接入策略,然后又研究了当3G蜂窝网络同WLAN共存时,认知用户在主用户QoS约束下的功率可达区域,以及如何获得和功率最小点的最优功率控制算法。对于动态频谱接入问题,考虑了一个采用OFDMA接入方式,并且同时存多个主用户和认知用户的无线认知网络。研究了在每个主用户干扰温度(Interference Temperature)受限的情况下,认知用户如何通过联合地调整它们的速率、功率和频谱资源,使得速率加权和达到最大。首先对涉及到的物理层和链路层进行了数学建模,将研究的通信问题转化为最优化问题,然后用凸优化理论分析了该问题,并分别基于贪婪算法和拉格朗日对偶法提出了两个不同的中心式算法,可以证明,这两个算法都是近似最优的算法,同时具有比较低的复杂度。最后通过引入虚时钟,提出了分布式的实现协议。更多还原

【Abstract】 Wireless resource allocation is a vital way to handel the conflict between the limited wireless resource and the increasing Quality of Service (QoS) requirement of multimedia traffics. However, with the development of multi-hop and heterogeneity techniques, the future wireless networks will consist of multi-hop cellular network, wireless mesh network, and cognitive radio network. These new types of networks have solved many problems that exist in the traditional wireless networks, e.g., lack of expansibility and robustness. However, just as every coin has two sides, these novel and powerful techniques will also change the way of utilizing wireless resource, and further become a challenge for the traditional layered resource allocation algorithm. In fact, significant performance gains can be achieved by various cross-layer approaches in these networks, and cross-layer resource allocation (CLRA) is necessary for the future wireless network protocol stack design. In this dissertation, under the guidance of information theory, network theory and convex optimization, the cross-layer resource allocation techniques in the typical three emerging types of wireless networks are intensively studied. The contents of this work are listed as follows:Two different CLRA methods are proposed for traditional cellular networks and multi-hop cellular networks, respectively. For the former case, a downlink system combining AMC and ARQ is considered. We study the problem of spectral efficiency maximization for QoS-guaranteed services, and propose a cross-layer link adaption algorithm to get the global optimal solution. For the multi-hop cellular network, a two-hop wireless link employing AMC and finite-length buffers is considered. We propose a multi-hop queuing model to analyze the network throughput and delay performance. Furthermore, we consider the problem of optimal power and bandwidth allocation for QoS-guaranteed services. We first discuss the optimal bandwidth allocation and the optimal power allocation. Then, we propose a joint allocation algorithm, which can iteratively find the optimal power-bandwidth pair and thereby improve the network performance.For CLRA in wireless mesh networks, we consider the problem of joint optimization of source coding, power control, ARQ control, and delay partitioning functionalities, our studied problem is to maximize the video quality under strict end-to-end delay con- straints through adjusting the source coding rate, end-to-end delay distribution, and each node’s transmit power. First, the performances of application, physical, MAC, and network layers are modelled by some classical models under reasonable assumptions. Then, we formulate the studied problem as a mathematical optimization problem. This optimization problem is proved to be a nonlinear but log-convex one. Finally, we propose a centralized solution based on the geometric programming theory, as well as a partially distributed solution based on the Lagrangian dual decomposition technique. And both solutions are proved to converge to the global optimum of the above problem.For CLRA in cognitive radio networks, we study two types of problems, which are heterogeneous network coexisting and distributed dynamic spectrum access. With respect to the former case, we study the problem of 3G cellular network coexisting with 2G cellular network and the problem of 3G cellular network coexisting with WLAN, respectively. From the view of multiuser information theory, cognitive users’ optimal access method and optimal power control policy are proposed for these two problems, respective. Then, for the case of dynamic spectrum access, the problem of wireless resource management in broadband cognitive OFDMA networks is addressed. The objective is to maximize cognitive users’ weighted rate sum by jointly adjusting their rate, frequency, and power resource, under the constraints of multiple primary users’ interference temperatures. We formulate the studied problem as two nonlinear and non-convex optimization problems, and propose a centralized greedy algorithm to solve one problem, as well as a centralized algorithm based on Lagrangian duality theory for the other. The two centralized algorithms are shown to be optimal, and both have polynomial time complexities. Finally, we present that the two centralized algorithms can be distributively implemented by introducing the idea of virtual clock.更多还原