Plan Model In Service Agent

Web Service technology can easily achieve low-coupled distributed system integration, and it has become an important way to publish and share resources inter-firm or inter-enterprise. However, Web Service technology holds stateless response function. There are many shortcomings exist in Web Service, such as single function, without initiative response to events, and lack of collaboration with other services. In addition, with the increasing number of Web Services, it becomes more difficult for users to select the correct services. Agent is a kind of intelligent software entity with autmatical actions in the environment to achieve goals. It has been applied widely to communicate and collaborate among multiple Agents using loosely Agent network and ACL to complete tasks. Also, Agent has the ability to make decisions independently, collaborate with each other and have social capabilities. Therefore, the combination of Web services and Agent is one of the popular research area in the development of Web technology. Service Agent can manage a group of closely related Web services, and react intelligently to the outside world. It provides comprehensive and integrated services.The ability of Service Agent is built on a group of plans. The plan concept is the manner of Service Agent to manage agroup of related Web Services. Service Agent lives in a distributed environment, so the requirements and circumstances change at any time. How to manage the plan to maintain the adaptive capacity, how to design plan model for tasks allocation, services scheduling and resources controlling, will be the research contents of this paper. The plan model is designed into a Process model to coordinate services. In this paper, the semantic model based on goal ontology and the syntactic model grounded upon ECA rules for plan are defined. And it implements ontology-based reasoning and develops fuzzy evaluation of plan quality. Based on above algorithms, it also designs how to choose the most similar with criterions of Requirement Goal expected quality. The the paper used depth-first search and heuristic search to build the abstract plan in the vertical level. Then the paper introduces three methods to resolve the problems, which the plan model was built throuth Modeling Environment but can not statisfy the dynamic and changed requirement goal. There are: (1)automatic planning to construct the plan model based on AI-technology; (2)planning update used event detector in plan runtime environment; (3)planning library optimization through transforming the plan model into tree model.Based on the research contents, we developed the Service Agent Platform, for integrated system to develop, operate, and distributed deploy services. It also support agents to provide integrated services can intelligently adapt to dynamic and uncertain demands. The Service Agent platform also provides good usability,. In a conclusion, the Service Agent will have a bright future.