Research On Hierarchy Policy And Reputation-oriented Multi-Agent Credibility System

Virtual organization is a kind of new business organization and management model. By creating an open and dynamic virtual organization, companies and individuals can establish appropriate alliances, and collaborative, negotiate, share resource in the virtual organization. Agent has some characteristics, such as autonomy, responsiveness, social and initiative. And it has strong behavior autonomy and collaboration ability. Therefore, it is a good technological method to select agent as a single autonomous element, and build virtual organization system based on multi-agent technology. However, owing to autonomy and black box of agent, when we build virtual organization system based on multi-agent technology, in order to pursue maximal individual interest, agent tend to make some behaviors, which will have an affect on community cooperation and system overall goal. So it will result in some problems, such as redundant autonomy and insufficient credibility.To solve the above problem, in this paper, in order to solve the problem of redundant autonomy and insufficient credibility in system, we propose a three-tier security system to stand the individual and group behavior of agent in virtual organization system, use trust and reputation mechanism to establish cooperation positively, use monitoring and arbitration mechanism to guarantee three safeguard systems reversely. Generally speaking, the main contents and contributions are as following two aspects:1) Study the credibility mechanism, which use the hierarchy policy as a core. In this paper, the policy has been divided into individual layer policy and organizational layer policy. We use individual layer policy to make sure that the user can guide and control their agents, use organizational layer policy to ensure that community cooperation behavior of agent carry on opportunly, and maintain the goal of community cooperation and the overall goal of system. In addition, introduce some common policy types which are always used in the system, give the formal representation and application examples of policy, study the hierarchy policy-oriented credibility behavior theory of agent and the credible collaboration and consultation process of multi-agent. We believe that it is possible to use the way of credible behavior combination and binding to transform policy into credible behavior, which can be implemented by agent directly. We use this method to standard the autonomy behavior of agent, and solve the problem of redundant autonomy and insufficient credibility in system.2) Study trust and reputation mechanism of forward promotion and monitoring and arbitration mechanism of reverse protection. Multi-agent credibility system faces to application background for virtual organization. And the effective collaboration among Agents is the core interest of the system. The trust and reputation mechanism can provide a credit reference for selecting collaborative object, avoiding blindly select collaborative object. Monitoring and arbitration mechanism provides a reverse safeguards, in order to monitor and punish those agent that violate policy norms and undermine group collaborative activities, forcing agent to comply with policy norms and system requirements. In addition, in order to adjust to multi-agent credibility system, we make some advances based on the existing reputation models, and present a reputation evaluation method which is more reasonable.