Automated Negotiation Research Based On Multi-agent

Along with more and more common online trades, a large number of electronic commerce services are being arisen on Inertnet. Nowadays, there are growing interests on some hotspots in agent-based e-commerce research. Agent-based soft technique has been considered as a useful tool to develop online business because of its autonomy, interactivity and intelligence, which can effectively meet the flexible needs of online trading, However, there are little technical supports of current e-commerce to automatization, especially to automatic decision-making process, such as "automated negotiantion". How to apply agent technique into automated e-negotiation effectively has been a mainsteam in both economics and computer science.Agent-based automated negotiation can be regarded as a decision-making process in dynamic enviroment lack of complete information among competitive and coopertative Agents. Agent proposal must consider personal preferences, negotiation history of offers, information of counterworker and environment and so on. It needs a negotiation system which can adapt to the change of dynamic environment to correspond the behavior among Agents timely. Hence, how to combine machine-learning theories into automated negotiation system has got more attention recently. My paper is just based on the above points mentioned for further expansion in many related fields. It consists of five parts:(1) Presenting the theoreitics tools used in Agent technology and outlining the background and application of Agent technology.(2) Introducing the definitions and related conceptions about Agent and MAS.(3) Introducing the related conceptions about electronic commerce and negotiation, and some advantages of Agent technology applied to e-commerce, as well as the prosess of Agent-based negotiation.(4) Describing a formalized multi-issue negotiation model and the process of Agent-based negotiation in details from three aspects of negotiation protocol, negotiation model and negotiation strategy. Suggesting using expected utility strategy and grouping proposal to prevent counterworkers' overbid and save jetton.(5) Discussing the necessary of learning, and analyzing the application of Bayesian learning and dynamic Q-learning algorithm to formalized multi-issue negotiation model.(6) Summarizing the content of this paper, and pointing out some shortages and the possible research in the future.