| Summary: | 博士 === 元智大學 === 資訊工程學系 === 92 === Due to the rapid development of the Internet, and everyone's wishing to be interconnected and to be able to access data at any time from anywhere, current information environment is open, large, and heterogeneous. In such a setting, the intelligent agent that is capable of autonomous, proactive, and social action has become one of the most active areas of research in recent years. Meanwhile, the world is a multi-agent environment, so goals cannot be met without consideration of others (and their goals). Consequently, as a coordination mechanism and its ubiquity in everyday event, agent negotiation has always been one of principal subjects in the agent research.
This dissertation presents a general problem-solving framework for modeling multi-issue multilateral agent negotiation using fuzzy constraints. Such a model addresses the problem of how participants negotiate for multilateral contracts to be mutually beneficial without a central coordinating agent. Agent negotiation is formulated as a distributed fuzzy constraint satisfaction problem (DFCSP). Fuzzy constraints are thus used not only to define each agent's demands involving human concepts, but also to represent the relationships among agents.
Negotiation strategies determine how agents evaluate and generate offers to reach an agreement that is most in their self-interest. Agents exchange offers throughout the negotiation according to their own negotiation strategies. In this dissertation, a concession strategy, based on fuzzy constraint-based problem-solving, is proposed to relax demands and a trade-off strategy is presented to evaluate existing alternatives. With the inherent capability of self-relaxation and gradual contraction, this approach provides a systematic method for reaching an agreement that benefits all agents at the highest possible satisfaction degree of constraints. Meanwhile, by applying the method, agents can move towards an agreement more quickly, since their search focuses only on the feasible solution space. A set of meta negotiation strategies, different combinations of concession and/or trade-off strategies, then are further developed for various scenario applications.
Additionally, agents should be able to communicate and understand each other. To construct such framework, communication language must be standardized so that different parties have common knowledge to interact. Thus, we propose the state transition function to specify the agent interaction protocol, give a model of agent negotiation behavior to illustrate the whole negotiation process, and then provide the proofs on the correctness of our negotiation model as well.
To illustrate the negotiation process of our model in various negotiation strategies, an e-Marketplace application is presented in this study followed by an example application of international trade to demonstrate the effectiveness of the proposed model in a multi-issue multilateral negotiation. Then, the usefulness of the proposed negotiation model is further proved by its successful application to a distributed planning/scheduling at a UPS manufacturer and a PC chassis collaborative design.
|
|---|
