Research On MultiAgent Concurrent Negotiation Strategies In E-Business Environment

E-business has taken advantage of Internet, which ignores the distance between companies and customers. It has brought companies infinite business opportunities and customers more choices. The success or failure of E-Business seriously depends on business negotiations. However, negotiation is a matter that needs many resources such as persons, money and time. It has become more and more important that how to find an automate negotiation mechanism to promote effective and efficient intercommunion among the negotiators. In a word, E-business needs automatization or half-automatization of business negotiations.Because software agents have the features of autonomic, mobile and flexible, agent-based negotiation (namely automated negotiation) technology is useful to give an effective and efficient solution in business negotiations. Today automated negotiation has given increased importance to E-commerce.To secure better deals and acquire more utility, a negotiator may engage in negotiating concurrently with many opponents for a particular good or service, namely concurrent negotiation. Then he could choose the best trade solution. Concurrent negotiation also could improve negotiation efficiency and save negotiation cost. It exists in our lives widely.In this paper, we present a novel model and some strategies for bilateral multi-issue automated concurrent negotiation in E-bussiness environment. Our major research work in this paper can be viewed as follow.1. We present a novel model, called MACNM (MultiAgent Concurrent Negotiation Model), in which buyer agent can adopt different strategies to negotiate concurrently with many seller agents who have different preferences, in semi-competitive situations in which there exists information uncertainty and deadlines.2. We present the strategy based on Q learning algorithm to propose own offer, the strategy based on similarity criteria to evaluate the opponents' offers and the strategy based on time-ultility to manage commitment. These strategies compose an automated concurrent negotiation mechanism.3. Our model introduces a learning mechanism by which the buyer agent can learn automated useful knowledge about the preperence and the belief of the seller agents from their offers during negotiation. These knowledge help the buyer agent generate its offers which are more possiblely accepted by the seller agents and help the buyer agent choose one seller which will bring it the most utilities. The mechanism is easily incorporated into the model.4. Our model provides a computable solution for the concurrent negotiation which lets the process of chosing and using the negotiation strategies from the unformalized and uncomputable form into the formalized and computable process.5. We construct the negotiation platform, implement simulation experiment.Through the experiment data, we certificate the feasibility and efficiency of our negotiation model and strategies.