【作者】 林跃东；

【导师】 刘发贵；

【作者基本信息】 华南理工大学 ， 计算机应用技术， 2013， 博士

【摘要】 语义Web服务将语义Web与Web服务相结合，利用语义Web丰富的语义描述能力和逻辑推理能力，从而使得计算机及系统平台能够“智能”的理解Web服务与用户请求，实现服务的自动化发现、组合、监控与调用。但由于单个语义Web服务的功能有限，当已存在的单个服务无法满足用户请求时，如何智能分析用户请求，自动化发现多个服务并组合成为满足用户请求的复合服务，已成为工业界与学术界共同关注的焦点问题与研究重点。虽然已经有研究者提出多种服务组合方法，但仍存在以下问题：①无法在数量庞大的服务中获取小规模可用的服务链，存在服务执行行为导致服务组合失效的情况；②难以处理子服务间复杂关联关系及执行顺序的动态组织；③缺少根据服务组合结果结构与关联关系度量其与用户请求的匹配程度的方法；④在系统框架上难以集成多种服务组合发现与组合方法并为其提供支撑。本文关注上述问题，以服务组合模型与方法、系统框架及组合结果的复合相似度度量方法为研究对象，以实现服务组合的执行顺序的动态组织为目标，探讨服务组合关键技术及应用，主要工作如下：（1）借鉴超图思想，提出基于超边图模型的面向组合的候选语义Web服务链集获取方法为了提供小规模可用服务集并避免服务的执行行为导致组合结果失效的情况，本文提出服务的单次执行过程获取方法以及基于超边图的候选语义Web服务链集的获取方法。主要成果包含：①以基于服务行为特征的服务单次执行过程作为本文方法的基础服务集，避免已有方法中存在的服务执行行为导致服务组合失效的情况；②借鉴超图思想，提出采用显示、全面而准确的形式化描述表示潜在候选服务链的超边图模型，其结构清晰又直观；该模型以服务单次执行过程为节点，控制超边图节点粒度，克服超图等模型中节点粒度过大或者过小的问题；③提出通用的超边图构图算法以及服务链获取算法，获取行为特征满足用户请求的候选服务链集。通过复杂性分析与实验，验证其有效性，结果表明其具有更低复杂性与良好性能。（2）对Petri网模型进行扩展并提出基于该模型的服务组合方法针对子服务间复杂的关联关系，为了实现子服务执行顺序的动态组织，本文对Petri网模型进行扩展并提出基于该模型的服务组合方法：①对Petri网的库所进行扩展，提出输入/输出库所的多种子类型并进行分类，表示并处理多来源的输入库所与多用途的输出库所的情况；②引入ε库所与ε变迁，通过其合并-选择与复制-分裂机制，表示并处理服务间一对多与多对一的复杂关联关系情况。通过①和②的扩展，克服原始Petri网无法处理复杂关联关系的问题。③提出基于扩展后的Petri网模型的服务组合方法，实现子服务执行顺序的动态组织。最后，通过实验验证其有效性并分析其复杂性。（3）提出基于子服务权重与关联相似度的复合相似度度量方法针对当前缺少在结构与关联关系上度量组合结果与用户请求的匹配程度的方法，对此，本文提出基于子服务权重与关联相似度的复合相似度度量方法：①分析并提出库所重要系数度量算法，计算不同用途输出库所的重要性系数；②根据子服务的不同用途的输出库所的分布情况，联合库所重要性系数，计算子服务权重，度量子服务在组合结果中的重要程度；③设计子服务关联相似度度量方法，充分考虑子服务权重以及I/O关联关系的模糊概念相似度；④提出基于目标输出的关联相似度与子服务个数的复合服务相似度度量方法，计算服务组合结果与用户请求的匹配程度，实验论证本文度量方法的有效性。（4）提出基于语义Web服务路由的系统框架本文提出基于语义Web服务路由的四角色系统框架，为了为服务组合方法提供系统框架的支撑，该框架引入语义Web服务路由，包含领域本体树融合、路由定位、服务发现与组合统一接口等核心算法与机制：①通过语义Web服务路由中的领域本体树融合算法、路由定位算法实现注册中心或服务路由的定位并进行消息的分发，适应多领域多注册中心的分布式应用环境；②通过服务发现与组合方法的统一接口，实现多种服务发现与组合算法的集成。通过对框架的实现及实验验证了其核心算法与机制的有效性，结果表明该框架适应多领域多注册中心的应用环境并实现多种算法的集成。（5）在物联网中的应用服务组合关键技术关键技术的应用，体现其实际价值。本文描述语义Web服务关键技术与物流供应链物联网结合的应用场景，选取物流供应链物联网环境下的车辆调度场景，采用支持用户自定义数据类型的模糊描述逻辑进行本体与服务的描述，并添加到所实现的系统框架中，应用服务组合关键技术，体现其应用价值。上述研究最终形成包含服务链获取、服务组合模型与方法、复合相似度度量、系统框架等语义Web服务组合关键技术以及应用。更多还原

【Abstract】 As the combination of Web service with semantic Web, semantic Web service is makingWeb service more readable to machine with its rich capacity of description and great strengthof reasoning, and thereby offering great help to service discovery, composition, monitoringand invocation. But due to the limited function, there may not exist a single service to meet acustomer’s requirement. At this situation, how to composite the single ones to a powerfulcomposite-service is becoming a hot spot in both industry and academic. Researchers haveproposed a variety of service composition methods, but there still exist the followingproblems:①An available and small-scale service chain can’t be obtained from a huge numberof services and some service’s behavior may led to the failure of service composition;②Complex connecting relationship among sub-services can’t be represented and dealt, andfailed to organize the dynamical processes of sub-services;③matching degree betweencomposition result and user request can’t be measured;④the existing system framework can’tsupport and integrate variety methods of service discovery and composition.Focusing on the above problems, key technologies of semantic Web service compostionwill be launched with this dissertation, with the objection of dynamic composition. The maincontent and contributions are listed as follows:(1) A method of getting candidate service chains for composition in chain is proposed,based on the hyper-edge graph modelIn order to get an available and small-scale service chain and avoid the condition thatsome service’s behavior leds to the failure of service compsotion, methods of getting SingleExecution Process and candidate semantic Web service chains based on hyper-edge graph areproposed in this dissertation, providing candidate service chain sets for compostion methodsto dymantically organize the execution process.①choose the execution-behavior-basedSingle Execution Process to be the basic service set, avoiding the condition that leds to thefailure of compositon;②a hyper-edge-graph model for candidate service chain is proposed,describeb by formal, accurate and comprehensive description, which has very strongcommonability with clear and intuitive structure; in the model, making Single ExecutionProcess be the node can control the node granularity and the scale of hyper-edge model.③common algorithms of building hyper-edge graph and analyzing hyper-edge graph areproposed to achieve the candidate service chain based on node-edge path. Finally, theeffectiveness and affectiveness are verified by the prototype system and experiments. (2) Extend the basic Petri net model and An algorithm for service composition based onextended Petri-net is proposedTo address the complex conntecting relationship among subservices and to realizedynamic organization of subservices, the basic Petri net model is extended, and based onwhich, a semantic Web service compostion method is proposed:①Places of basic Petri netfor Input and Output are extended to some subtypes for multi-source input place and multiuseoutput place;②ε Place and ε transition with combine-select and copy-split mechanisms areintroduced for complex connecting relationship such as1:n and n:1;③a service compositonmethod based on the exended Petri net is proposed to dynamically organize the executionprocesses of subservices. The experiments show our methods’ affectiveness and effeciveness.(3) A method for measuring the composite matching degeree with user request isproposed, based on subservice weigth and connecting similarityFor measuring the matching degree between the results of composition and user request,a measure method for composite similarity degree is proposed, which is based on subserviceweight and connecting similarity.①Place Importance Factor Measure algorithm formeasuring the importance of different subtypes of outplaces is proposed based on analyzingtheir different usages.②the subservice weight is measured by the distribution of outputplaces and their importance factor;③a method for measuring the connecting similarityamong subservices is proposed, based on subservice weight and fuzzy ontology similarity ofI/O to transfer the information of similarity from one subservice to another.④the compositesimilarity is based on both the connecting similarity and the amount of subservices, propsosedto measure its matching degree of user request and convenient for service selection. Theexperiment shows their affectiveness.(4) A system framework based on semantic Web service router is poposedA semantic-Web-service-router-based system framework is proposed to provide systemsupport for service compositon, with domain ontology tree fusion, routing location algorithmand unified interfaces of service discovery and composition methosds:①In the framework,domain ontology tree fusion algorithm and routing location algorithm to achieve the locationof the suitable role, making the framework better suited to the distributing environment withmulti-domain registry;②through unified interfaces of service discovery and compostion, theframework is able to support and integrate a variety of service discovery and compositionmethods. Finally, an integration platform of the proposed framework is realized and the keyalgorithms and mechanisms is proved effective. (5) Applications of composition technique in Internet of ThingsApplication of key technologies reflects their practical value. The application oflogisitics suplly chain IoT combined with semantic Web service is proposed and described inthis dissertation. In vehicle scheduling scenario selected for application of key technologise inthis dissertation, the ontology and services are described using the key technology of fuzzydescription logic with custom Datatype, and then apply the key technologies proposed in thisdissertation to verify their practical values.Above all, the research achievement includes: method for getting candidate servicechains, models and methods for service composition, composite similarity measure, thesystem framework and platform, which are key technologies of semantic Web servicecomposition and their applications.更多还原