【作者】 薛小平；

【导师】 张思东；

【作者基本信息】 北京交通大学 ， 通信与信息系统， 2009， 博士

【摘要】 RFID网络已成为RFID应用的研究热点,研究人员提出了以中间件为基础的应用框架,并在诸多领域得到了广泛应用,如物流、供应链、零售业、食品以及铁路车号自动识别系统等。早期的RFID应用,由于应用的简单性,研究人员对应用架构的研究较少,侧重关注RFID数据获取以及制造成本等问题,包括:廉价的标签和阅读器的研究与设计、标签及阅读器冲突、特定应用环境下的标签、以及数据读取的可靠性等问题。传统的RFID应用主要是针对特定领域的应用,大都采用基于数据中心的应用方式,即由阅读设备获取标签对象数据,并集中进行存储和处理。重点关注如何将RFID数据与现有的应用系统相关联,因而,中间件技术被认为是解决RFID网络的关键手段和技术方法。近年来,随着RFID的普及,RFID应用正在经历由简单应用向复杂应用的转变;从简单的数据处理转向面向事件的处理;从局部应用走向全球化应用;从RFID与后台单一应用系统相结合,转变为在异构环境下实现数据共享等。这些转变给RFID数据管理和系统架构带来了新的挑战,要求新一代的RFID应用架构能处理具有以下特征的数据事件:RFID数据的时态性、动态性和关联性,数据的流特性和海量数据,冗余和差错,以及在异构网络实现数据共享等特征。为满足新一代RFID网络应用的要求,本文提出了面向复杂事件处理的普适应用环境下的RFID网络体系结构,并对其中的关键理论和方法进行了深入的研究。本文的主要工作包括:(1)将Pub/Sub系统与RFID复杂事件处理相结合,提出了RFID分层网络结构,给出了RFID网络关键描述及定义,包括RFID网络的形式化描述、匹配、路由等,研究并讨论了RFID网络中的关键组成部分一代理的结构和主要功能。(2)考虑到海量数据条件下RFID网络的负载和可扩展性,基于复杂网络和结构化P2P,提出了分层RFID网络结构和模型。在此基础上,针对著名的基于内容的CBCB路由,通过考虑代理邻居间关系,提出了簇首代理网络中的代理遍历路由算法,实验结果表明:考虑代理邻居间关系的代理遍历算法,可有效地减少簇首代理网络由于遍历而带来的通信开销。(3)为限制簇首代理网络内部开销,考虑到客户订阅间的相关性,提出了带有源地址的基于边的覆盖合并精确路由算法,并进一步给出了Chord环簇首与簇内代理的基本算法,算法可以减少传统的Chord环覆盖路由中的订阅重复传播的问题。同时,为适应Chord环的动态变化,采用数据代理的方法,对Chord环进行维护。研究表明:利用上述方法一方面可有效地减少网络内订阅的传播数量,另一方面也使得Chord维护开销更低。(4)研究并讨论了基于Chord环的簇内和簇间移动性问题,提出了基于Chord环的簇内移动性支持算法以及簇间移动性支持算法,包括订阅路由、订阅取消路由算法等,并根据弱有效路由准则,证明了算法的正确性。实验表明:在移动情况下与标准算法和延时优化算法相比,在路由重构开销、网络负载等方面具有较好的表现。更多还原

【Abstract】 With RFID network becoming hot in RFID applications,researchers proposed the middleware-based application framework which has been widely used in many fields, such as logistics,supply chain,retail,food,automatic train identification system and so on.Due to the simplicity of the early RFID applications,few researches paid attention to the structure,while most of the work focused on the RFID data acquisition as well as the manufacturing costs,including:low-cost tags,research and design of the readers, the tag collision,the reader collision,tag in a specific application environments,and the reliability of the data reading.Traditional RFID applications are mainly used in specific fields,and the application modes are mostly based on the data center,that is,labeled object data acquisition through reading equipment,then centralized storage and processing.How to associate RFID data with the existing application systems will be the focus point,thus, the middleware is considered to be a key method and technology in the RFID network.Recently,with its popularizing,RFID is going through the application changes from simplicity to complexity;from simple data-processing to event-oriented processing;from local applications to global applications;from the RFID combined with back-end single application systems to data sharing in a heterogeneous environment,and so on.These changes bring new challenges to the RFID data management and system" architecture,and demand a new generation of RFID application framework which can process data events with the following characteristics: RFID data tense,dynamic and relevant,the data flow and the mass data,redundancy and error,as well as data sharing in heterogeneous network.In order to meet the needs of the new generation of RFID network applications, this paper presents a complex event-oriented processing RFID network architecture for the universal application environment,and makes an intensive investigation about the key theories and methods.The main work includes as follows:(1) Pub/Sub system is combined with the RFID complex event processing,and a hierarchical RFID network structure based on Pub/Sub System is proposed,as well as some key descriptions and definitions about the RFID network are given,including the formal description of the RFID network,matching,routing,and so on.The key components about RFID network - the structure and main function of the broker are also discussed.(2) Considering the load and scalability of the RFID network in conditions of massive data,a hierarchical structure and model of the RFID network based on the complex network and the structuring P2P is put forward.On this basis,for the well-known content-based routing CBCB,considering the relationship between the broker neighbors,a broker traversal routing algorithm in a network of cluster-head broker is proposed,and results show that:the broker traversal algorithm considering the relationship between broker neighbors can effectively reduce communication overhead brought by the traversal in the network of cluster-head broker.(3) For the restrictions of network overhead on cluster-head broker network, considering the dependency of the client subscription,this paper proposes a edge-based precisely cover routing algorithm combined with the source address,and further gives the basic algorithm about cluster-head broker and broker in the cluster,which can reduce the repeated spread of subscribers in the traditional Chord ring coverage route. At the same time,in order to adapt to the dynamic changes of the Chord ring,a data proxy approach is used in Chord ring maintenance.Research shows that:on one hand, these methods can be effective in reducing the number of subscribers spreading in the network;on the other hand,they can also reduce the Chord ring maintenance costs.(4) This paper studies and discusses problems about mobility of intra-cluster and inter-cluster based on the Chord ring,and proposes algorithms which can support the mobility for botfi the intra-cluster and inter-cluster based on the Chord ring,including the subscription route algorithm,subscription to the abolition of the route algorithm and so on.This paper also proves the correctness of the algorithms based on weak efficient routing criteria.Experiments show that in comparing with the standard and delay optimization algorithms in route reconstructing expenses,and network load,the algorithms have better performance when in the situation of mobile clients.更多还原