Research And Implementation Of Multi-agent Workflow Model

Nowadays,Workflow Management System(WfMS) has been applied widely in theenterprise to support the business management and business control. The basicconcepts,theories and techniques in workflow management system are researchedthrough large amounts of literatures.This thesis analyzes the current workflowmanagement systems’ deficiencies, such as poor distribution, parallelism andintelligence which are not such sufficient as expected. A muti-agent workflow modelbased on role&task-based access control is presented to solve these problems.Thismodel is designed to improve the deficiency of the traditional workflow managementsystem through making a reference to WfMC model,combining with the characteristicsof Agent, taking into account the problem of access control in a multi-agent workflowmanagement system and making use of the agent features (e.g., autonomy, intelligence).In the model, the approach to access control module is proposed by introducing thetask set(T), the instance set of task(TI), the task agent set(TA) and the user agent set andthe concept of TaAC (specification of the Task Agent access control). The authorizationrelations are analysed and the formal specification and graphical representation areshowed to explain the model. This access control model enables workflow model withcentralized license management and high security characteristics, and follows theprinciples of Separation of Duty and least privilege.The multi-Agent workflow model is analysed and designed from several aspectsincluding workflow management system architecture, the workflow execution service,process operating mechanism and function modules. Firstly, the design andimplementation of the access control model are described.The agent types andfunctions,the way of communication and collaboration between the agents are discussedat the same time. The advantages of the multi agent Workflow model are also analyzed.Finally, the model is proved in the platform of WADE.Through analyzing the test resultsof the mutil-agent inlligence processing system, the validity of the model is verifiedand the requirements of access security and workflow management system are satisfied.