【作者】 陈怡；

【导师】 胡瑞敏；

【作者基本信息】 武汉大学 ， 通信与信息系统， 2011， 博士

【摘要】 Ad Hoc网络在众多通信技术中,以其快速布网能力、结构灵活、价格低廉、鲁棒性强、不需要基础设施的特点,得到越来越广泛的关注。由于带宽限制和网络拓扑结构的动态变化,在Ad Hoc网络中保证服务质量(QoS)是一个具有挑战性的工作。目前,对于Internet和其他网络结构中支持QoS已做了大量的研究工作,但并不适用于Ad Hoc网络。无线Ad Hoc网络技术的发展尚短,特别是如何满足用户对网络资源的需求,提高网络传输性能存在许多不足,其中包括：1)带宽分配产生负载流量波动导致Ad Hoc网络工作环境恶劣的问题；2)带宽需求异质性的多播用户业务流对Ad Hoc网络带宽分配满意度低的问题；3)多数据流构成的复杂任务在Ad Hoc网络中服务质量低的问题。这些问题导致Ad Hoc网络环境承载多媒体数据传输未能达到期望的服务质量,网络资源的利用低,极大阻碍了无线Ad Hoc网络大量商用的步伐。本文以武汉市科技攻关项目“车载多媒体通信系统”和国家自然科学基金重点项目“移动音频技术理论与关键技术”为依托,对其中无线多媒体网络的传输质量展开研究。针对无线Ad Hoc网络多媒体传输特性,研究网络传输质量的难点问题,提出了新的解决方案,并通过理论分析、证明和实验验证了本文理论成果的有效性,本文的理论贡献与创新如下：1、提出了一种分布式PD控制动态资源分配算法ORAA。本文通过把控制理论中的PD控制机制嵌入原有非线性优化分布式算法,改进了传统算法GPA比例控制方式。当无线Ad Hoc网络端到端延时较大时,比例控制调节作用会使网络系统出现振荡甚至失稳的现象。本算法在资源分配价格体系下,采用分布式PD控制,达到改善系统动态性能、提前预见网络可用资源变化的目的。使网络快速收敛,降低网络带宽波动。本文算法与传统GPA算法相比,在随机动态网络环境下,收敛时间缩短接近50s,平均提高17%左右的收敛时间；比GPA-Z缩短40s,平均提高13%左右的收敛时间；同时仅提升2%左右的网络开销。这对于提高Ad Hoc网络系统传输的服务质量,提升上层用户运行环境具有重要意义。2、提出了一种带宽异质多速率多播资源分配算法AMC。本文在Peng模型基础之上,引入用户效用感知精确曲线和竞价机制,克服原有Ad Hoc网络多播带宽资源分配算法中未考虑多媒体用户带宽需求的差异,仅适用单一业务类型多播流,网络资源利用率不高的问题。本算法采用用户分段效用竞价方式,减少零散带宽分配次数,特别适用动态变化条件下的Ad Hoc网络,从而提高整体带宽资源分配效率。本文算法与传统MCT算法相比,在同等30个无线节点移动速度为8m/s的随机移动拓扑结构中,平均效用提高17%。提高了模型的精度,获得比MCT更高的网络效用,比MCT算法更适应带宽异质的多播用户。这对于提升系统在动态异构环境下资源利用率,保障上层应用体验具有重要意义。3、提出了一种基于多数据流任务协作的自适应分配算法MRD。本文从各数据流相关性出发,在非线性最优化理论的框架下,通过调整任务内各数据流单位效用,动态满足业务对数据流带宽的要求,从而有针对性的分配资源,克服原有技术没有考虑不同数据流之间依赖关系,造成任务内部分数据流缺失或质量下降,导致当前任务失败的问题。本文算法与现在最新MST算法相比,网络整体效用提升4%。这对于提升任务完成质量,满足新型应用对带宽分配模式改变的要求,提高模型自适应性,优化网络带宽资源利用率具有重要的意义。综上所述,本文研究成果从Ad Hoc网络带宽分配的角度探索高效可靠的网络质量(QoS)控制策略,为Ad Hoc网络系统科学设计提供开创性的指导思想,具有重要的理论意义,对于保障上层多媒体应用服务质量和提高网络有限资源的利用率具有重要的应用价值。本论文首先明确给出研究成果涉及的三个创新点；绪论部分对研究的背景和研究方向的国内外研究现状做了综述,阐述了研究内容及关键问题,并给出整篇论文的组织结构；正文部分首先引入相关最新技术的模型问题分析,再详细论述了本文的技术路线,接着进行了理论分析和实验验证。本论文最后给出了对未来工作的展望。本文在展望部分指出未来研究工作需要解决的主要问题是：1)如何在变化的网络环境下选择匹配的收敛参数(如收敛步长),以提高算法的自适应性；2)考虑如何避免模型受到测量误差等网络参数影响,提高算法的鲁棒性；3)考虑多数据源内容重叠对数据分组的产生的冗余,如何结合网络编码和视频内容建立合理的协作机制降低分组数据流量,以提升算法的效率。更多还原

【Abstract】 Ad Hoc network attracts more and more attention compared with other communications by its abilities, including rapid implementation, low cost, robust structure and without the infrastructure. Due to bandwidth limitation and dynamic topology, it is a challenge to support Quality of Service (QoS) in Ad Hoc network. A lot of research work done for Internet and other network are not suitable for Ad Hoc network.With a short history of wireless Ad Hoc network technology has many deficiencies especially about how to meet the needs of resource for network users. These deficiencies include 1) bandwidth allocation fluctuations leading to poor working conditions in Ad hoc network; 2) heterogeneous users of bandwidth demand being with low satisfaction level; 3) complex mission consisted of multiple streams being with low satisfaction level. These problems fail the service quality of multimedia carried by Ad Hoc network and hamper the commercial use of wireless Ad Hoc network.Based on Wuhan Science Research Project "On-board multimedia communication system" and the National natual Science Foundation of Key Project "Mobile audio technology theory and key technology", this paper research the difficult problem of these issues and propose new solutions. Theoretical analysis, theorems and simulation results verify the positive effect of our work. The theoretical contributions and innovations are as follows:1. A PD control of distributed dynamic resource allocation algorithm ORAA.In this paper, the PD control is embedding into optimal distributed nonlinear algorithm to anticipate the change trends of network resources. The method overcomes these disadvantages of GPA, which lead to oscillations and instability because of proportional control algorithm. Our algorithm can get quick convergence and low bandwidth oscillations. Compared with original GPA algorithm, in random dynamic network, the convergent time is shortened by 50s, namely cut down 17% average convergence and only increase 2% network overhead. And our algorithm cut down 13% convergence time (40s) compared with GPA-Z. It is vital important to improve the transmit quality of service and environment of applications.2. A bandwidth heterogeneous multirate multicast resource allocation algorithm AMC.This algorithm, based on Peng model, introduces accurate perception utility curves and auction mechanisms to overcome the weakness of original algorithm which leads to under-utilization and waste of network resources without consideration different bandwidth demands of multimedia users and is suitable for homogeneous users. This algorithm uses discrete utility curve and bidding mechanism to reduce iteration number and increase the aggregated utilization, especially for dynamic Ad Hoc network. Compared with original MCT algorithm, in the same topology of 30 random mobile wireless nodes, AMC algorithm outperform over MCT algorithm by 17% utilization. The AMC algorithm is more accurate than MCT algorithm, more suitable for bandwidth heterogeneous users, which upgrades the resource utilization in dynamic heterogeneous network and guarantees the upper application experiences.3. An adaptive adjustment algorithm MRD for cooperative mission based on multiple data flows.Based on the correlation of multiple flows and nonlinear optimization framework, the algorithm adjusts the marginal utility to meet the dynamic bandwidth demands of missions. This method, considering the correlation of multiple flows, can over the defects of the loss or low quality of some flows in the mission. Our method outperforms the latest MCS algorithm by 13% the overall network utility and outperforms the original MST algorithm by 4% the overall network utility. It is vital important to enhance the quality of new applications for bandwidth allocation, to meet the changing requirements of model, to improve model adaptability and optimize the network bandwidth utilization.In summary, the paper explored reliable network quality (QoS) control strategies for Ad Hoc network system and to provide innovative science guiding ideology. This paper has theoretical significance and application value to guarantee service quality of top multimedia applications. In conclusion, future work is proposed.This paper firstly gives three innovation points. Then the introduction is about the research background, the current research state, the research content, the key problems and the whole organization. In the main part, a problem analysis is proposed, and technical theory and experimental verification are followed. In the end, the future work is discussed.In the future, the main work needs to solve these issues:1) how to choose parameters (e.g., step length) under the dynamic networks; 2) avoid the measure errors and improve the robustness of algorithm; 3) how to improve the efficiency of algorithm by combining network coding and video content information.更多还原