【作者】 刘栋；

【导师】 陈俊亮；

【作者基本信息】 北京邮电大学 ， 计算机应用技术， 2009， 博士

【摘要】 普适服务(Pervasive Services)是计算机网络和无线通信两大领域共同的研究愿景,其目的是构建一个开放的、无所不在的环境为用户提供智能化、个性化的信息和服务,而上下文感知计算技术与语义Web服务的结合是实现普适服务的有效途径之一。本文从上下文建模和推理技术出发,对上下文感知语义Web服务发现、提供的实现方法进行深入研究,概括而言,主要包括下列内容:1.由于上下文建模是实现上下文感知计算的基础和前提,因此,本文首先研究了基于本体论的上下文建模方法,并运用概念格和RDF具体化等工具对已有的上下文本体进行了改进,建立了一个通用的上下文模型。该模型不仅能够描述物理、计算、环境、社会和任务等各种上下文的语义,而且还将Allen区间代数和区域连接演算等时空数据表示方法引入了上下文感知计算中,提高了模型在时序、空间信息表示方法方面的能力,并为上下文推理、语义近似度计算以及上下文感知Web服务的研究奠定了基础。2.本文利用基于规则和基于本体的推理技术,设计实现了适用于上下文感知计算领域的时序和空间推理方法。在时序推理方面,提出了持续时间推理和时序关系推理方法;在空间推理方面,主要解决了空间拓扑关系一致性判定和隐性空间信息获取等问题。实验结果表明,本文提出的推理方法有效。3.本文利用标注机制解决了上下文模型与服务请求、服务描述结合的问题,并在此基础上提出了一种上下文感知Web服务发现方法。该方法主要由三个执行阶段组成:上下文匹配、接口匹配和基于QoS服务排序。在上下文匹配阶段除了一般的上下文匹配算法以外,本文还专门为时间、空间和设备这三种最重要的上下文类型专门设计了匹配算法。另外,由于上下文匹配是在接口匹配操作之前执行的,因此可以过滤掉一些与服务请求无关的备选服务,从而减少了接口匹配的次数,有效地提高了服务发现的效率和准确率。4.以何种形式为用户提供服务是Web服务发现过程结束之后需要解决的最重要的技术问题之一。为此本文提出了一种基于粗糙集理论的上下文感知规则生成方法,该方法将上下文感知系统视为一种决策信息系统,并利用基于可辨识矩阵的方法对上下文信息进行约简,进而生成规则。系统可以根据规则生成的结果自动选择适当的形式为用户提供服务。另外,由于在规则生成阶段所能利用的数据有限,生成的规则可能难以完全覆盖上下文取值范围,为了增加规则匹配的准确率,本文还设计了一种基于语义距离的规则匹配算法。更多还原

【Abstract】 Pervasive services, whose objective is to build an open and ubiquitous environment to provide users with intelligent and personalized information and services, are the shared vision of computer networking and wireless communications. Integration of context-aware computing and semantic Web services is one of the effective approaches to realize pervasive services. On the basis of context modeling and reasoning, context-aware semantic Web services discovery and delivery are in-depth researched in this dissertation. In sum, the main contributions are as follows:1. Context modeling is the basis and precondition of context-aware computing. Therefore, ontology-based context modeling is first lucubrated in this thesis, and the existing context ontologies are improved by utilizing context lattice and RDF reification to construct a universal context model which can describe the semantic of various types of contexts such as physical context, computing context, social context and task context, etc. Moreover, the description methods of spatiotemporal data, i.e. Allen’s Interval Algebra and Regional Connection Calculus, are introduced into context-aware computing so as to enhance capability of the expressions of the spatio-temporal data and lay the foundation for context reasoning, semantic similarity calculating as well as context-ware Web services.2. Rule-based and ontology-based reasoning technologies are utilized to implement spatial and temporal reasoning which apply within the context-aware computing. As temporal reasoning is concerned, duration reasoning and temporal relation reasoning are proposed. While, in respect of spatial reasoning, the problems such as consistency checking and implicit information obtaining are addressed. The performance evaluation results indicate that the presented methods are efficient.3. An annotation-based mechanism for combining context model with service request and description is presented in the thesis. On this basis, context-aware Web service discovery is also proposed, which can be divided into three phases: context matching, interface matching and QoS-based ranking. In addition to general-purpose context matchmaking algorithm which perform in the phase of context matching, special algorithms for matching of the temporal, spatial and device contexts are introduced. Furthermore, prior to interface matching, some candidates which are unrelated to the service request in terms of context, can be filtered through context matching, as a result the frequency of the execution of interface matching is reduced, and the efficiency and precision of service discovery are improved. 4. One of the most important technical problems arise after Web service discovery is how to determine the manner of service delivery. With regards to this, a rule generation method is proposed. Context-aware systems are regarded as decision systems, and context information are reduced with discernibility matrix so as to generate rules. System can automatically select an appropriate form of service delivery. In addition, because data can be utilized is limited, the generated rules can not entirely cover the domains of contexts. The rule matches current context probably does not exist. A rule matching algorithm based on semantic distance is presented as the solution to this problem.更多还原