Research On Double-Layer Policy-Based Multi-Agent Cooperation System

To overcome the shortcoming of traditional information system, which is closed, lack of function and hard to control and manage, an idea of combining virtual organization technology into information system is proposed. The enterprise can build enterprise unions and cooperate with other enterprises to achieve its goal through virtual organization. To implement the information system which is intelligent and easy to control and manage, some problems need to be resolved:how to implement dynamic, flexible, on-demand and autonomy cooperation on resources and services in a virtual organization, which with a certain degree of autonomy and controllability, including the controllability of enterprise business strategy and cooperation process?Aiming at the above problems, we begin our research work with focus on the autonomy and controllability of virtual organization, take the software agent as an autonomic element, with multi-Agent coordination as the target of our research, Multi-layer policy is introduced into the management of multi-Agent cooperation system, a double-layer policy-based multi-Agent cooperation system is proposed. In general, our works and contributions are summarized as following:1) A mechanism about double-layer policy-based multi-Agent cooperation is presented, which divides the policies into individual policies and organizational policies, and uses four steps of policy load, policy refinement, policy-behavior mapping and policy execution to change the policies to the agents' behaviors. In addition, in a multi-Agent cooperation process, we use the concerted control of double-layer policies and the principle of priorty and first-update-first-prior to resolve the conflict between policies, which achieves the deterministic control on system objectives. Compared with the previous study, the flexibility and controllability in cooperation process are enhanced, and provide the conflict resolution mechanism to enhance the stability and consistency in cooperation process.2) An implement mechanism and method for double-layer policy-based multi-Agent system are presented. On the one hand, we establish the architecture of double-layer policy-based multi-Agent system, present the management framework and structure features which take the policies as a core. On the other hand, aim at the real-time requirement of system, use of the asynchronous transfer characteristics of Ajax to implement real-time display mechanism, real-time access to agent status, ensure system users can load policy to guide agent behavior according to the real-time status of agents. Compared with the previous related systems, our system can flexible and real-time respond to and deal with external events, and put the policy-control idea into specific modules, which enhances the efficient, flexibility and controllability of the system.