A Research On Agent Automated Negotiation Based On Cooperation

With the development of Internet, E-commerce has been an important business mode. In order to resolve the conflicts and disputations about different attribute in commerce activities, negotiation is used in most cases to reach an agreement. Traditional negotiation is usually conducted face to face by the buyers and sellers, which will surely wastes lots of manpower, materials and time. Some times it may be influenced by the environment and human factors to bring some unfairness and randomicity, and even failure. Automated negotiation can resolve such problems.Currently, the research of agent-based E-commerce has been a hot topic. The software agent technology is considered as one of the most useful technologies in E-commerce, because its autonomy, interaction and intelligence can effectively meet the needs of flexibility in trade. How to apply the advanced Agent technology into automated e-negotiation has been a mainstream in both economics and computer science domain.Excepting price, the attributes involved in e-negotiation include quantity, quality, Warranty period, delivery mode etc. This thesis researchs in the multi-issue automated negotiation. Regarding the negotiations object, it has one-to-one, one-to-many and many-to-many forms. But in one moment, it has only one form. So the seller agent must decides how to respond buyer agents when many buyer agents send buying requests in one time. This is a key point in automated negotiation: decision-making. In order to win-win, the agent must consider social factors and the object's character except for interest.The major research work in this thesis can be viewed as follows.1. The thesis analyzes the state of the arts of the automated negotiation in AI and E-commerce, and the sociality of agent. To construct a multi-issue automated negotiation system based on agent, and to formalize the seller agent, buyer agent and information agent.2. The concept of spread and cooperation are proposed. And we use tree mode to compute spread degree, and use BP Nerve Network to compute cooperation degree.3. To discuss the negotiation strategy based on time restraint. Then a new biding strategy and negotiation system based on cooperation is proposed. When many buyer agents send buying requests in one time, the seller agent decides which one to response first and using what strategy by the cooperation.4. Finally, based on the research mentioned above, we design and implement an experiment system using Java language. Through the experiments, we verify the validity and feasibility of the multi-issue negotiation, and compare varieties of negotiation strategies.