【作者】 杨文俊；

【导师】 汪秉文；

【作者基本信息】 华中科技大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 无线传感器网络是目前一个热门的研究领域,它涉及微电子制造、传感器技术、无线通信技术、嵌入式技术和分布式信息处理等学科。若要尽量发挥无线传感器网络的环境感知与数据采集能力,并充分利用各种监测数据,就必须跳出无线传感器网络的局限,与其他领域的理论与技术进行融合,以便对无线传感器网络进行更好的调度与控制。在这些领域中,分布式的信息广播系统无疑是与无线传感器网络结合最为紧密的一项。无线传感器网络所采集的监测数据可成为信息广播系统的数据基础,而信息广播系统则通过其内的各个组件对这些数据进行管理、决策、协调与控制。本论文即围绕以无线传感器网络为核心的信息广播系统展开工作,探讨并研究了其中的一系列核心问题。本文探讨了无线传感器网络中的数据获取和保存问题、大规模信息广播系统中层次服务器间的数据传播与更新问题,以及数据广播调度问题。通过这些研究,给出了从生产者到消费者的实时数据获取、存储与传递算法,以尽量降低节点能量消耗、增加数据的可用率,并减少消费者的等待时间。本文提出了一种使用内容发布／订阅的信息传播、处理架构EventX。在无线传感器网络层,从复杂事件订阅出发,提出了订阅分解与合成机制。将复杂的时间、类型等相关的复合主题事件分解为简单的监测、查询事件,并排除其中冗余的部分,以形成若干个彼此独立的子订阅,从而避免数据的重复采集与传输。同时,引入订阅以及子订阅权重机制及相关的调度算法,保证在资源有限的情况下优先处理重要的订阅,以提高系统应对多个订阅的服务质量。仿真结果表明,订阅分解与合成机制能够在保证数据采集效率的前提下,有效减轻网络负载,延长系统生存时间。通过分析无线传感器网络的特性及其应用特点,针对节点存储能力有限,无法将网内大量数据信息存储在若干个存储节点上的情况,本文提出了一种新型的传感器网络数据网内存储框架REALSTORE,以便在节约节点能量和内存开销的前提下提高数据的可靠性。给出了REALSTORE的设计目标和体系结构,提出了数据存储协作域、管理节点等基本概念,给出了协作域的建立机制和管理节点的选取算法,然后提出了两种在协作域内进行数据存储与提取的算法与流程。通过仿真测试,证明其能够在较低的资源开销下提高信息的可靠性。在接收监测数据的服务器层,本文提出了一种事件匹配算法SGEM,以订阅的统计分布为基础,根据订阅中的谓词对订阅进行分组,以便快捷、灵活并实时地响应消费者的查询请求。仿真实验证明,SGEM事件匹配算法能够有效地对事件进行处理,其性能优于其他现有算法,可降低匹配过程中的开销,能减小内容传递的延时,对系统性能的改善显著。本文深入探讨了层次化的信息广播系统中通过数据广播在一级服务器内进行高速数据更新,以及在二级服务器上进行大量数据查询请求处理的问题。主要的目标是降低消费者收到超时数据的百分比,并减少数据请求的平均等待时间。针对以上需求提出了具有广播判定功能的RU、RU+ESI和PF数据更新广播算法。这三种算法将数据更新与数据广播调度进行结合,改善了主服务上数据高速更新与二级服务器上高频率数据请求所带来的冲突问题,从而达到降低数据超时,减少消费者的等待时间的目的。此外,在二级服务器上还实现了两种广播判定准则——LWF与RxW,使得这三种算法能够适应不同的应用环境。仿真结果说明,RU、RU+ESI和PF数据更新广播算法都优于简单的基准调度算法,在不同的更新速率和消费者请求频率下,这三种算法都处于稳定状态,能够正常工作。针对多项查询的广播调度算法中的两种选择方法——查询选择方法和数据项选择方法,本文提出了一种采用数据项选择方法的新型的数据调度算法MLWF,以适应动态变化的查询列表。在MLWF中,引入了一种权重机制,根据数据项在未完成查询中出现的频率、所在查询的总体等待时间长度、广播该数据项能否使得尽可能多的查询进入完成状态等因素入手,综合考虑数据调度问题。MLWF算法能够选中使得大量查询状态变为完全应答的数据项,并进行广播。仿真实验将MLWF与LWF、Round Robin和FSO等传统调度算法进行了比较。结果显示,在消费者等待时间方面,MLWF的性能优于LWF、FSO和Round Robin算法。论文最后对所研究的内容进行总结,总结了本文内容与创新点,并提出了后续研究工作的前进方向。更多还原

【Abstract】 Wireless sensor network has becoming a hot research topic recently. It combines research areas like the microelectronic manufacture, sensing technology, theories of wireless communication, and embed technologies. In order to make full use of the power from monitoring and collection of WSN, and take advantages of the monitoring data, the WSN should be integrated with other theories and technologies. Among the related areas, distributed information broadcast system is one of those techniques which are closely connected with WSN. The monitoring data collected by WSN will be the basis of information broadcast system that manages, makes policy and controls the whole WSN and its data. This dissertation carries out the research work in the design of WSN-based information broadcast system and focuses on the related core theories and algorithms.This dissertation discusses the following problems:data collection&restoration in WSN, data propagation, and data update&schedule in large-scale hierarchical information broadcast system.Also, this dissertation proposes the event-triggered based subscription model, and then proposes a novel message oriented publish/subscribe broadcast system named EventX to cater to the features of sensor networks and its applications. According to type and time interval, EventX implements an algorithm to analyze and decompose compound subscription. After that, another weight information based greed algorithm is introduced for the low-energy state of nodes. Simulation results show that the EventX is suitable to reduce communication overhead and prolong the lifetime of sensor networks.Considering the continuing advances in sensor networks and application design, it’s important to improve information availability in a class of delay-tolerant sensor network applications according to the scarce resources and sensor energy. This dissertation points out the design objective and introduces the system architecture of RealStore—a framework for providing in-network storage service in sensor environments, and defines the information storing cooperation zone as well as its zone manager in RealStore. RealStore implements an algorithm to store and extract information based on network coding theory in cooperation zone. Simulation test results show that the RealStore is capable to improve information availability efficiently.In the data server layer, existing event-matching algorithms are not very efficient, especially for interval range predicates and overlapping predicates in subscriptions. So this dissertation discusses the above challenge and proposes a dynamic and fast event matching algorithm called SGEM. SGEM algorithm can well support range predicates or overlapping predicates and provides single and composite event matching. It groups the subscriptions by the predicates and dynamically identifies appropriate number of groups considering different statistical distributions of subscriptions at run time. Also, we present an experimental evaluation and compare its performance with existing algorithms. The experiment results show that SGEM can significantly reduce the evaluation cost and guarantees the scalability with respect to the number of subscriptions as well as the number of predicates and events.This dissertation research updates propagation using broadcast delivery method in hierarchical data broadcast systems with a high rate of updates at the primary server and a high rate of data requests at the secondary servers. Updates are propagated from hierarchical data servers to single-item on-demand data requests from a population of data consumers. Three algorithms that combine update propagation with data broadcast scheduling in hierarchical data broadcast systems with a high rate of updates at the primary server and a high rate of data requests at the secondary servers are proposed named:RU, RU+ESI and PF algorithm. The proposed algorithms are functions of the rates of consumer requests and of data updates to increase data freshness and decrease consumers wait time. At the secondary servers, two implementations of the decision maker component of the proposed algorithms (i.e., LWF and RxW) are provided to demonstrate the versatility of the proposed algorithms. The experimental results show that the proposed algorithms are stable under various update arrival rates as well as consumers’ request rates.Considering the challenges of scheduling items for broadcast by data servers in response to data consumers’ multiple-item queries. This dissertation introduces the item selection approach, MLWF—a new data scheduling algorithm that adapts dynamic changes in the queue of pending queries. MLWF schedules a data item to meet the following conditions for broadcast:the item has a high frequency of occurrence among the pending queries, the total wait time of the queries in which the item occurs is large, and the broadcast of the item will cause a large number of queries to be completely satisfied. MLWF uses a normalized weight factor to give a higher priority to the broadcast of a data item that will cause the largest number of pending queries to be fully (or almost fully) answered. MLWF is compared with a generalization of LWF, Round Robin and FSO, All the experiment results show that MLWF performs better than LWF, FSO, and Round Robin in terms of consumers’ average wait time.In the last section, a summary has been given for all the proposed research contents. The innovative points are summarized. Also, the further work has been discussed.更多还原