【作者】 孙岩；

【导师】 马华东；

【作者基本信息】 北京邮电大学 ， 计算机应用技术， 2007， 博士

【摘要】 随着监测环境的日趋复杂多变，迫切需要将信息量丰富的图像、音频、视频等媒体引入到以传感器网络为基础的环境监测活动中来，实现细粒度、精准信息的环境监测。近年来，对多媒体传感器网络(MSNs，Multimedia Sensor Networks)技术的研究已引起了学术界和工业界的密切关注。MSNs可提供的服务多种多样，因而衡量每类服务的服务质量(QoS)的标准也不尽相同。如何保障不同服务的QoS且有效利用全网资源，成为MSNs研究领域的一个突出问题。本论文围绕MSNs服务质量体系在各个层面的相关保障机制和算法进行了研究，侧重于服务模型的建立，服务感知的业务调度策略，服务感知的路由选择算法以及自适应QoS保障机制等几个方面，主要考察了MAC层、网络层、传输层以及应用层的QoS保障策略以及相互影响。论文工作主要包括以下几个方面：(1)根据MSNs的特点，在基于簇的MSNs体系结构上，借鉴DAVIC服务模型的思想，并对其加以改进，从而建立一种适用于MSNs的服务模型。在此基础上设计了该服务模型的协议流程。该服务模型使MSNs以服务为核心，为用户以及上层应用提供保障其QoS需求的多媒体服务。(2)面向MSNs的两种基本服务模式(基于事件驱动和查询驱动的服务模式)，提出了一种提供区分服务的MSNs业务调度模型，进而设计了一种服务感知的业务调度策略，使用非抢占的优先级排队模型保障两类服务的时延要求，以提供满足多媒体应用需求的QoS。(3)针对MSNs中三类基本服务：异常事件告警服务、信息查询服务以及流媒体查询服务，分析了不同服务的QoS需求，并抽象出多媒体传感器网络QoS路由模型。进而，设计了一种基于蚁群优化的服务感知路由算法ASAR(Ant-colonyoptimization based Service Aware Routing)。该路由算法旨在依据各类媒体数据包的不同QoS需求，选择相应的路由以合理利用全网资源、提高网络性能。(4)考虑到MSNs中同时存在的基于端到端(end-to-end)和基于事件到汇聚节点(event-to-sink)的两种传输模型，结合传输控制协议和媒体采集方案生成方法形成了一个基于二级反馈的QoS自适应保障机制。根据网络的QoS和传输负载状况将网络划分为10个典型状态。在传输层，通过传输控制协议调整媒体流和事件流的发送率，完成网络向最优状态的迁移；在应用层，通过传输层计算得出的媒体流数据包发送率，计算节点多媒体信息采集的最佳方案。上述方法适合于MSNs的混合传输模型，有效地保证了两类基本服务的QoS。(5)在基于簇的MSNs体系结构上，我们设计并开发了一个多媒体传感器网络原型系统。在此基础上，为了更好地保障MSNs的服务质量，设计了灵活的、分布于各类多媒体传感器节点的中间件—AuViM，并详细描述了中间件中各模块功能以及两种主要服务模式流程。进而提出一种描述MSNs中QoS需求的方法，以便灵活地为每个用户定制相应的QoS需求。综上，论文从MAC层、网络层、传输层以及应用层研究了MSNs中QoS保障的关键技术，并设计了适用于MSNs的中间件。目前在MSNs中，QoS保障技术还很不成熟，多种服务的并存，使得QoS问题比传统传感器网络更难解决。因此QoS保障问题研究对MSNs的发展具有重要的意义更多还原

【Abstract】 With the complexity and variation of monitoring environment, it is urgent to introduce rich media (i.e. audio, image and video) into environment monitoring activities on the basis of sensor networks, thus perform a complete and accurate environment monitoring. Recently, multimedia sensor networks (MSNs) has been paid wide attention. Due to the variety of services provided by MSNs, their QoS criterions are different. How to guarantee different QoS while maximizing network utilization is becoming a hot topic in this research field. We study the corresponding guarantee mechanisms and algorithms of QoS in different layers of MSNs. This thesis focuses on service model, service-aware traffic scheduling scheme, service-aware routing algorithm and self-adaptive QoS guarantee mechanism. The main contributions of this thesis are as follows:(1) Considering the characteristic of MSNs, based on the cluster-based architecture of MSNs, we improve the DAVIC model to establish a service model suitable for MSNs. Furthermore, we design the protocol flows of this service model. This architecture and service model make MSNs focus on service and provide multimedia services to meet the QoS requirements of users and upper applications.(2) For two kinds of basic service modes (event-driven and query-driven) in MSNs, we propose a traffic scheduling model which can provide differentiated service. Based on the queuing model with non-preemptive priority, we design a service-aware traffic scheduling scheme, which can ensure delay-constrained requirements of two kinds of services to satisfy different QoS of multimedia applications.(3) For three types of MSNs services: event-driven service, data query service and stream query service, we analyze different QoS requirements, thus abstract a QoS routing model for MSNs. Moreover, we propose a QoS routing model for MSNs. By differentiating three types of typical services, we satisfy with diverse QoS requirements of all these services. Moreover, we improve the traditional ant-based algorithm and quantify pheromones value to speed up the convergence of our algorithm. In particular, we propose a novel ant-based service-aware routing algorithm (ASAR). According to different QoS requirements of all types of media data, this ASAR algorithm can select the optimal paths to satisfy with their individual QoS requirements, thus optimize network utilization.(4) There exist two kinds of streams in MSNs: end-to-end stream and event-to-sink stream. Combined with transport control protocol and generating approach of capturing task, we present a self-adaptive QoS guarantee scheme based on two-layer feedback. We partition the network into 10 typical states according to the network QoS level and network congestion status. MSNs perform the state transfer to the optimal one in the transport layer by means of adjusting delivery ratios of two streams in the transport control protocol. Moreover, application layer evaluates the optimal scheme of capturing multimedia according to the multimedia packet delivery ratio calculated by the transport layer. Our method is fit to the hybrid transport model for MSNs, and it can ensure the QoS of two kinds of basic services.(5) According to the cluster-based architecture of sensor networks, we design and develop a prototype system for MSNs. To guarantee QoS in multimedia sensor networks, we also present a flexible, distributed middleware named as AuViM. Moreover, we detail the functions of different components in AuViM and the flows of two service modes. In particular, the QoS description structure is introduced to help users to customize individual QoS requirements.In all, this thesis studies the key problems of QoS guarantee from MAC layer, network layer, transport layer and application layer in MSNs, and designs a middleware suitable for MSNs. At present, the technologies of QoS guarantee in MSNs are immature. The coexistence of all kinds of services makes QoS in MSNs difficult to handle. Therefore, the research on QoS has a significant meaning for the development of MSNs.更多还原