Abstract: The paper puts forward a fuzzy-constraints-programming-based negotiation model in the application of trading agents.The model preferably describes the trade off preference of buyer agents by using prioritized fuzzy constraints.An intelligent algorithm for the programming model is put forward，which integrates neutral network，fuzzy simulation and gene algorithms.By running this programming algorithm，seller agents get a pareto optimized solution without global searching of the product database.The simulation indicates that the products given by the negotiation model is in consistent with real preference curve of buyers get by investigation.

Key words: fuzzy constraint programming, automatic negotiation, intelligent agents

摘要： 给出了一个基于模糊约束规划模型的自动协商系统。建立了模糊约束规划模型并利用模糊模拟、神经网络和遗传算法给出了求解Pareto最优解的混合智能算法；协商过程中卖方智能体根据神经网络拟合的效用函数并运行混合智能体算法得到当前协商步的Pareto最优解，避免了对大型商品数据库的反复搜索，为系统推向实际应用奠定了基础；协商模型仿真实验表明了协商系统返回的解与实际调查得到的用户偏好相一致。

关键词: 模糊约束规划, 自动协商, 智能体