A Reasrarch On The One-To-Many Automated Negotiation System Based On Multi-Agent

Business negotiation is an important part in market-place. With the development of e-commence and globalization, Business negotiation would be more frequent and complex, which asks us to react quickly and flexibly to the change of market. The traditional negotiation which is occurred a lot of human resource, has not been adapted to the need of the development of global e-commence .It has been become an urgent problem that using computer program to replace the human resources. Automated negotiation based on MAS is a new negotiation fashion, which could increase automation of negotiation obviously, save time, reduce the cost, improve efficiency of every participator and agree on a preferable result.During the actual trade-off, negotiation usually more complex, it has kinds of fashion, the number of the negotiation issue often more than one. This paper focused on one-to-many multi-attribute negotiation fashion. We proposed a one-to-many negotiation model adopting elimination system, and put forward the concept of cooperation possibility, which was used to evaluate the sellers and eliminate the sellers having smaller cooperation possibility, so that reducing the negotiation cost, and using uniform proposal in order to avoid the unfair phenomenon for the seller agents. In additional, the formulae of compute cooperation possibilities were given. This paper designed the negotiation protocol adapted to the model in the paper, based on the CNP (contract Net Protocol), while adopted the negotiation strategy using uniform proposal based on unifying the different proposals for different types opponents. We used GA and Bayesian learning mechanism to estimate the ultimate agreement proposal of the opponents and compartmentalize the opponents.After making deep research on one of agent development frameworks——JADE(Java Agent DEvelopment Framework), we development the prototype system for one-to-many multi-issue negotiation, which can make full use of agent's intelligence, cooperation and self-adaptation, and promote the system's flexibility. Also, the prototype system proved the effectiveness of such a system.