摘要

Modeling is always one of best ways for us to recognize and describe the real world. As the most important sub-system of decision support system, MB (Model Base) and MBMS (Model Base Management System) undertake much more expectations and responsibilities. In fact, we had so many models to help us to describe and understand problems of decision, but we hadn't found the efficient and perfect approaches to do with it. After all, the manipulation of model is much more complex than data, Although there are some attentions on architecture of model base, few researches are related with distributed artificial intelligence. In this article, we attempted to provide a novel architecture of model base management system in which we took MBMS as an independent system, just like the DBMS. Then, we drew a model of MBMS based on cooperative multi-agent system, and argued the key problems of MBMS. To improve the efficiency of cooperation among individual agents, we extended the traditional agent structure of BDI. About the implementation of engineering, we discussed the agent-oriented methodologies compared the traditional object-oriented methodologies.*