Study Of Multi-Issues Negotiation And Decision-Making In Virtual Enterprise Based On MAS Technology

Virtual enterprise (VE), as a new form of enterprise organization, is increasingly being focused in academic and industrial fields. Along with the incessant studies and the developments of information technology, VE has successfully exceeded the conception formation stage and gradually moved into the practical operation stage. But, for the leader enterprise, how to conduct effective negotiation with the member enterprises and how to make fast and effective decisions become the bottlenecks for VE in the practical operation. Actually, the study of negotiation and decision-making mechanism has become a new subject in studying VE application.On the basis of analyzing traditional modeling methods, we conclude that the Agent technology more fits to the VE modeling and propose a MAS-based VE negotiation and decision-making model. And in this model, for the negotiation problem, we improved the traditional Bayesian negotiation mechanism for the negotiation problem in VE. Then the improved negotiation mechanism is successfully applied to the multi-issues negotiation in VE. And for the decision-making problem, we introduced the multi-objects genetic algorithm to the leader enterprise's decision-making mechanism. At last, we successfully applied our model and its corresponding algorithm to an air separation device company in HangZhou.The main research results include:1) Applying Agent technology to VE modeling, define the structures of each Agent in the model, extend the KQML communication words and add thinker parameters into the communication protocol;2) Improving the Bayesian self-learning negotiation algorithm and successfully applying it to the course of multi-issues negotiation in VE;3) Proposing a fast and effective multi-object genetic-algorithm decision-making mechanism for the leader enterprise and adopting the dynamic punishment function in the multi-object genetic algorithm.