The Research On Multi-agent Generating Environment (MAGE) And MAS Cooperation And Planning

Multi-agent system attracts many Artificial Intelligence researchers' interest in recent years, for it can better imitate the social intelligence of human being and it fit better the open and dynamic environment. At the same time, with the rapid application of Internet, Software developers need the design guide and developing tools for the open and complex distributed computing system urgently. These two forces make the research on agent oriented programming a new hot research topic. The social behavior of agent, especially its cooperative behavior, is a often studied research topic of MAS and AOP, and a research theme of this dissertation. The cooperation and multi-agent planning are investigated. The concrete topics of this dissertation include the formal representation of cooperation, multi-agent distributed planning based on constraint propagation, the description of cooperation and its organization mode, multi-agent generation environment and its application.In more detail, the contribution of this dissertation are as follows,1. This dissertation presents a formal system of cooperation. The formalization is realized as two levels: the ground level is a basic logic system, which describes some basic concepts such as action, knowledge, capability, and goal. Upon the ground level, some cooperation related notions are formally defined.2. The constraint propagation method of distributed multi-agent planning is studied. An algorithm of judging the consistency of partial order constraint among different agent's action is presented. Based on this algorithm, a planning model and a planning algorithm are designed.3. Cooperative agent team, an organization mode of multi-agent cooperation, is presented and applied in multi-agent generating environment MAGE. Cooperative agent team is built up dynamically according the the cooperation network via the clone operation of agents who intent to joint the cooperative team. This organization mode provides standard cooperative team formation and maintenance protocol and default behavior, and it also support the concurrent cooperation processes.4. Multi-Agent Generation Environment(MAGE) is designed and implemented for the requirement of developing tool of multi-agent system. It provides system design supports, the description methods and automatic configuration of knowledge and ability, standard API of cooperation and negotiation for the agent-based computing on the Internet. The standard services provided by MAGE, including naming service, security mechanism, query and advertisement service, facilitate the management and maintenance of multi-agent system.5. The research results, especially Multi-Agent Generation Environment MAGE, are used in some representative MAS application domains, including parallel multi-media information retrieval, group intelligence decision support system, and E-commerce. These prototype systems are built by MAGE.