【作者】 许华杰；

【导师】 李国徽；

【作者基本信息】 华中科技大学 ， 计算机软件与理论， 2008， 博士

【摘要】 集成了传感器技术、微机电系统技术、无线通信技术和分布式信息处理技术的无线传感器网络是一种全新的计算模式,是继因特网之后将对21世纪人类生活方式产生重大影响的IT热点技术。因特网改变了人与人之间交流、沟通的方式,而无线传感器网络将逻辑上的信息世界与物理世界融合在一起,将改变人与自然交互的方式。无线传感器网络是高度应用相关的,无线传感器网络的应用大多具有监测性质,因此对无线传感器监测网络的研究具有重要意义。在无线传感器监测网络应用中,测量误差、网络传输错误等因素的客观存在,数据采集和传输受到资源的限制而只能以离散的方式进行与外界物理量(如温度和压力值)的连续变化之间的矛盾不可避免,因此通过无线传感器监测网络获得的数据本质上是不确定性数据。不确定性数据对数据库领域传统的数据处理方法提出了新的挑战。无线传感器监测网络节点的主要特点是电能、带宽、计算和存储能力等高度受限,尤其是其电源的不可替换性导致在保证对监测目标完全监测的同时延长系统工作寿命成为无线传感器监测网络应用的一个中心问题。针对该问题,提出无线传感器监测网络的扩展工作寿命的定义,并在此基础上提出一种延长无线传感器监测网络工作寿命的分布式节点调度策略,在各节点簇内对节点进行调度以实现有差别监测服务并延长系统的工作寿命。不确定性数据查询与更新是不确定性数据处理技术的基础。概率性查询基于不确定性数据的概率不确定性数据模型,根据数据的不确定性区间及其概率分布(不确定性概率分布函数)为查询结果提供置信度信息。在深入分析各类不确定性数据的概率性查询及其计算方法的基础上,对概率性最近邻居查询ENNQ的计算方法进行改进。基于信息熵的概念,提出概率性范围查询ERQ查询质量度量方法,并在此基础上提出一种基于信息熵的不确定性数据更新策略,目的是以较小的能耗代价实现查询质量的提高。在无线传感器监测网络环境下,可能存在大量用户需要访问传感数据,因此数据的有效分发是不确定性数据处理的另一个核心问题。在移动计算环境中,数据广播是一种有效的数据访问方式。在深入分析不确定性数据特点的基础上提出数据平均不确定率的概念,并创造性地将Push-based在线广播方法应用于不确定性数据的分发,提出一种不确定性数据在线广播调度策略,在进行数据广播调度时综合考虑数据的访问率和不确定性,并考虑网络传输错误和多信道对数据广播的影响。无线传感器监测网络应用中往往伴随着海量的数据,研究从这些海量数据中挖掘出有用的知识意义重大。无线传感器监测网络环境中数据的不确定性会对数据挖掘结果的正确性产生显著影响,对传统数据挖掘方法提出了严峻的挑战。在对当前不确定性数据聚类的主要研究成果的深入分析并结合不确定性数据的特点的基础上提出基于密度的不确定性数据概率聚类算法,根据数据不确定性区间的概率分布信息提高算法的准确性并通过R树索引和概率阈值索引PTI提高算法的效率。更多还原

【Abstract】 Wireless Sensor Network (WSN), which integrates the technologies of sensor, micro-electro-mechanism system (MEMS), wireless communication and distributed computing, is a novel mode of computing and a hotspot of information technology after Internet. It will have a profound influence on many areas in 21st century. Internet changes the way people communicate and exchange, while WSN connects the physical world to the logical information world, and will bring on the revolution of the interacting way between human and nature. WSN is application specific. Most of the applications of WSN have the character of surveillance, so the research of Wireless Sensor Surveillance Network (WSSN) is very significant. Measure errors and network transmission errors cannot be avoided entirety in the applications of WSSN. Furthermore, extreme limited system resources like network bandwidth and battery power in WSSN can only afford sampling data in a discrete manner, while the values of the entities being monitored (e.g. temperature, pressure) is changing constantly. The intrinsic inconsistency or uncertainty of data related in WSSN makes such data uncertain data in nature. Uncertain data offer a new challenge for traditional data processing methods.In WSSN composed of a large number of low-power, short-lived, unreliable sensors, one of the most important design challenges is to obtain long system lifetime, as well as maintain sufficient surveillance to targets. The definition of the general lifetime of system is proposed and a round-based decentralized nodes scheduling scheme is presented, which schedule nodes in each cluster independently, therefore extend the general lifetime of system in differentiated surveillance. Fault-tolerance can be achieved with our scheme by taking nodes status and residual energy into account.The query and update of uncertain data are the foundations of the processing technology of uncertain data. Probabilistic query on uncertain data, which is based on the probabilistic uncertainty model, places confidence to query answers based on the uncertainty intervals and their probability distributions (uncertainty pdfs). By analysing the classification of probabilistic queries and related evaluation methods, an improved evaluation method for Entity-based Nearest Neighbor Query (ENNQ) is proposed. Metrics used to measure the qualities of the results returned by Entity-based Range Queries (ERQ) are proposed based on the notion of information entropy. An entropy-based updating scheme for uncertain data is presented, so as to improve the qualities of queries by the minimum energy overhead.There may be a large amount of clients who need to access sensing data on WSSN environment. Dissemination of uncertain data is another important issue in the processing of uncertain data. Data broadcasting is an effective means for data dissemination method on mobile computing environment. Definition of the mean uncertainty ratio of data is presented and a broadcasting scheme is proposed for uncertain data dissemination, based on the push-based online data broadcast. The demand probability and the uncertainty of data are considered in the process of broadcast scheduling. The effect of transmission errors and multiple broadcast channels are also taken into account in the scheme.The applications of WSSN usually involve a large amount of data. It is significant for the research on data mining on such large volume data. The uncertainty of data on WSSN environment affects the correctness of data mining remarkably, which offers new challenges for traditional data mining methods. The issue of clustering of uncertain data is focused on and a probabilistic density-based clustering algorithm for uncertain data is proposed based on the probability distribution of uncertainty. Effectiveness is improved by taking the probability distribution information in the uncertainty intervals of data into consideration, efficiency is achieved with R-tree and Probability Threshold Index (PTI).更多还原