Research On The Base Of Compensation And Commitment Re-negotiation Between Agents

The advantage of E-commerce facilitates more and more completion of transactions on-line. Intelligent Agent technology has been introduced into the E-commerce which makes the possibility of automation and intellectualization in every trade stage. As a middle part of the transaction, negotiation is the process that buyer and seller coordinate with and consent to each other according to the facets of object bargained. It is the key point to decide whether the transaction is succeed, as well as whether the requirements of buyer and seller are met simultaneously. So, how to let Agent complete the trade automatically using certain strategy to replace human kind, namely automatic negotiation, always lays a hot and focal spot in the research of electronic commerce and MAS.At present, many scholars here and abroad have done some researches on the automatic negotiations and have registered some achievements. However, the current researches are generally aimed at once negotiation. In our real life, it's pervasive that a second trade is taken between both sides of the transaction due to their requirements, which needs a repeat negotiation.Our research is done under the background mentioned above, which posses a better theoretical value and application prospect. The research work includes the following aspects:In this thesis, we have analyzed the present situation of researches on the automatic negotiation systems, and built a bilateral repeated negotiation model based on compensation and commitment. In this model, we have portrayed Agent's structure, and defined the protocol of both the flow in repeated negotiation and interaction among Agents. Then we established the environment of repeated negotiation, and described the form of proposal which is used in the negotiation process.In this thesis, the first proposal is formed on base of case reasoning. After Agent receives the transaction request, it will match the request of source case which is most similar to the destination case through similarity computing. Then Agent takes that request as the first proposal of this negotiation. So, the fist proposal is formed considering not only the demand but also the transaction history, which conforms to the reality much better.Commitment and compensation have been introduced into the repeated negotiation process, which makes Agent have stable and high negotiation efficiency in the long-term negotiation processes by considering the negotiation history. At the same time, the cooperation has been introduced to measure the Agent's compensation degree. This has avoided Agent to be always at the concession position in long-term negotiation process. Finally, we have applied confidence to measure the degree of completing commitment which prevents Agent not completing commitment to some extent.The utility function can be used in many situations by considering the separate and continuous issue in the utility function and giving the corresponding overall utility function in view of different proposals. Many kinds of strategy functions also have been proposed, which enables Agent to be able to choose effective negotiation strategies according to his own scene. In view of the long-time repeated negotiation, Agent can give concessions according to the compensation he obtained, which makes the average effectiveness of negotiation Agent a balance in the entire negotiation process. In view of that the Agent receive commitment, Agent can avoids deceiving because it give the appraisal of commitment according to the degree of credibility.We have thoroughly analyzed the negotiation strategies being used in the repeated negotiation process. In this thesis, we have introduced the concessions strategy and the compulsion strategy considering negotiation history, introduced the balanced strategy considering the first negotiation, and introduced the partiality-priority strategy considering the separate issue. Then these strategies have been quantized to a calculated plan, which makes the choice of negotiations strategy and negotiation process into a calculated formalization process.We have constructed the experimental system using Agent Builder on foundation of the research mentioned above, and simulated the automatic negotiation process, including conforming transaction between buyer and seller, evaluating the received proposal, producing the counter-proposal, and so on. Through the experiment data, we certificate the feasibility and efficiency of our negotiation model and strategies.