網路服務是共同認同的標準以提供服務給使用者存取,當前有越來越多的服務被提供在網路上,然而使用者希望能夠依照個人的喜好來存取這些服務,因此,一個自主性的機制能動態以及智慧化存取這些個人化服務是必須的,而使用代理人與語意本體論的技術能夠提供這樣的需求,不幸地,當前缺乏足夠的知識表示去描述代理人的運作概念,並且,也沒有提供內部單元去使用這些知識,本論文提出一個新的以群組為基礎的代理人架構,其中作業本體論以及作業引擎也被提出來以解決上述的問題,同時結合使用者本體論,代理人能夠動態的評估使用者對於服務的喜好值以存取個人化的服務,另外,在使用者本體論中也記錄工作流的資訊以產生客製化的工作流,同時透過延伸FIPA的22個溝通基本動作,我們也提出了14個群組溝通基本動作及群組互動協定,來輔助群組之間的合作。最後以我們的架構為基礎實作了一個旅遊系統,其中顯示出使用者能夠透過行動裝置上的JADE Android 平台,來動態的建立使用者代理人到其他網路伺服器上的JADE 平台以取得語意服務,該語意網路是透過OWL-S來描述的,另外,也實作了書店以及醫院的系統,用來分別說明動態的客製化工作流與群組代理人的合作。Web service is a common standard for publishing services for users to access. Currently, more and more services are provided by various web sites. Notably, users access web services based on their preferences. An autonomous mechanism for dynamic and intelligent access to such personalized web services is thus a necessity. Using agent and ontology technologies seem to provide this much-needed capability. Unfortunately, there seems to lacks either enough knowledge representation to operational concepts of agent or units to uses them. This work presents a novel group-based agent architecture, in which the operational ontologies and engines are proposed to generate semantic behavior to solve the problem above. Combining it with user ontology, preferred value of services can be evaluated dynamically by agent to access personalized web service. Moreover, the user ontology also records the workflow information to generate customized workflow. We further extend the 22 FIPA communicative acts to propose 13 group communicative acts and 3 associated interactive protocols in order to facilitate the communication among group-based agents.Finally, the implementation of a travel system based on our architecture shows that a user dynamically creates a user agent moving from user’s mobile device that runs the Java agent development environment (JADE) Android platform to any web server that runs JADE on which semantic web services can be accessed. Notice that web service can be annotated by web ontology language for service (OWL-S) to form semantic web service. In addition, the two other systems, namely, a bookstore and a hospital system are illustrated to show the dynamics of customized workflow and describe the collaboration of group-based agents, respectively.