一种基于多目标优化的Agent多议题协商模型

Agent多议题协商研究是多Agent合作求解的核心内容之一,一般基于对策论的方法实现Pareto最优的协商结果。由于很多学者将其转化为单目标约束满足问题,因而只能满足一方的效用最大化要求。Nash指出在理想情况下Agent应追求自身效用最大和对手效用最大的多目标优化,以达到快速达成一致并能最优化自身效用的目的。针对该问题,本文给出一种用指数型功效系数法求解的一揽子交易多议题协商模型NMMOP,该模型能够实现双方Agent的效用最优,提高协商双方的总效用。实验结果验证了该模型的优化效率优于Fatima和Faratin等人的工作。

An Agent Multi-Issue Negotiation Model Based on Multi-Objective Planning

Agent multi issue negotiation is one of the key problems in multi agent cooperative solving. Most researches realize the Pareto optimal negotiation results based on the game theory. Due to single objective optimization, most negotiation models can only satisfy the optimal requirements of one of the participates. However, as pointed out by Nash, agents should search for the optimal utility of both themselves and the opponent. Therefore, agents can reach an agreement quickly and obtain the  optimal utility for themselves. Thus, this paper proposes a negotiation model NMMOP with the package deal procedure based on the method of exponential efficacy coefficient. It can realize the optimal utility for both participants in negotiation, and increase the sum of their utilities. The experimental results indicate that the efficiency of the NMMOP model takes an advantage over the models of Fatima and Faratin et al.