Learning Model In Service Agent

Service-Oriented Architecture is new development architecture in systemdevelopment. Web Service has become more and more popular in manycompanies because of its low coupling feature. But traditional web servicecan not do good cooperation and because of many of services be published; itis also difficult to choose a right web service. Agent is an intelligent softwareentity and they can achieve their goals by their own action. Multi-agent canloosely coupled and through structured ACL messages to establishMulti-Agent System (MAS). In the MAS, Agent can complete missionsthrough their collaboration. Moreover, Agent is autonomy, initative,interactive, persistene and adaptability. So web service together with agenthas been proposed and extensively studied. Service agent can make up for theshortcomings of web services, which will manage a range of web service andthrough its own or combined with other services outside the agent to providecomprehensive and integrated service.The abaility of service agent is based on inner plan defined. Theresponsibility of plan is to manager web service in service agent and providesservice outside. At runtime, Agent will take the direct use of existing plan orcombine multi plan or re-organize plan in response to the event.Learning model in service agent contains reasoning model, storage modeland negotiation model. This paper summarizes the plan model and refines theontology-based semantic model and ECA workflow based syntax model.Reasoning model is based on planning theory. The semantic information canenter in and obtain the right sequence in the requirement of Qos. According tothe structured tree representation, this paper designs a storage model which can improve the work efficiency of Agent. In multi-agent environment, thispaper also give the negotiation model, it is based on multi-agent game andcooperation. It can effectively improve service need for a combination ofmulti-agent planning capabilities.Based on the above theoretical studies, the authors also implement Agentservice platform, the platform achieve the service model, plan model andlearning model. This platform has the good expansibility and robustness. Asthe result of software, it can be used for futher research in the laboratory.