Research On Hierarchy Policy Approach-Based Multi-Agent System For Virtual Enterprise

Virtual enterprise is an important style of business pattern. Because of its uniqueadvantages for optimization and integration of resources and its rapid marketresponsiveness capability, virtual enterprise will become one of the most importantbusiness organizations in21st century. In view of the agent’s autonomy, initiative,responsiveness, Multi-Agent system is very suitable for the simulation of virtualenterprise. However, there are some problems to build virtual enterprise byMulti-Agent system.. For the insufficient credibility problem of virtual enterprise andinappropriate granularity problem of agent, the paper proposes solutions separately.The main contents of the paper are described as follows:(1) For insufficient credibility problem of virtual enterprise, this paper presents ahierarchical policy approach to regulate the behaviors of virtual enterprise. Accordingto application dimensions, policies are classified into individual layer policy andorganizational layer policy. Individual layer policies are designed for users to achieveguidance and control of agents; while organizational layer policies are used to ensurethe credibility of an autonomy agents system. This article gives full description ofpolicy types used in the system, and presents a formalization representation of thehierarchical policy approach. For the problems of policy conflict that may be broughtabout by policy mechanisms, detection and processing methods are proposed.(2) For inappropriate granularity problem of an agent to simulate an enterprise,the paper proposes dual-layer Multi-Agent System model to simulate the operation ofenterprises. The granularity of agents to simulate a virtual enterprise memberdetermines the complexity of agents’ functionality and management. The dual-layerMulti-Agent System proposed in this paper bases on functional decomposition ofenterprises, it takes an ‘enterprise Agent’ to manage the behaviors of a virtualenterprise member from a holistic angle, and sets up special functional agents tofulfill complex tasks. These functional agents release resources and turn to deactivatestatus automatically when its tasks were finished. In this case, it is sufficient to avoidcomplex functionality of large granularity agents or difficult management ofexcessive fine granularity agents, and keeps the advantages of large granularity agentand fine granularity agent methods. Further more, it can mitigate the resourcessharing problem resulted by idle occupy of resources in virtual enterprise efficiently. At the last, integrated the research work of these two aspects, a virtualenterprise-oriented and hierarchy policy approach-based dual-layer Multi-Agentsystem was designed and implemented to validate the above theory research.