Research On Process Integration Oriented Multi-Agents System

Multi-agents system is a system organization style which is combination of the traditional artificial intelligence problem solving pattern and distributed computing technology. Agent is the first-class citizen of a multi-agent system, which has characters such as autonomy, social ability, reactivity, pro-activity, and so on. Through massage based communication and coordination, multi-agents react with each other and accomplish their common tasks. Nowadays this style of system architecture play more important roles in complex system design, decision support, enterprise information system integration, especially in supply chain oriented information system fields.This paper first presents an agent language prototype aiming at supporting enterprise integration. This language, namely Agent Pattern Language, has the ability of describing complex data objects and business processes, and integrating with outer services such as database operations. The APL agent model is based on blackboard architecture, and driven by goals. The theory model, the language specification and the language compiler is discussed in detail in this paper. Thus we present a complete agent language system prototype.On this basis, a runtime platform for the APL agents has been built, and is discussed in detail in this paper. The platform provides living environment and many facilities for agents. Those facilities include message service facility, outer program interface and global service provider, and the J2EE application server is used as the basic implementation technology. The outer program interface provides inter-operation mechanism between the platform and the enterprise legacy systems.Finally, we have implemented and tested a multi-agents system according to the methods described in this paper. We think it an effective tool for building the virtual organization of enterprises.