Abstract: Grid computing has been a hotspot in recent years.Facing geographically distributed，hybrid，autonomous and dynamic changing grid resources，how to manage and schedule them is a key problem，and is still far to be solved satisfactorily.This paper presents a novel grid resource management model based on Agent.In order to take full advantage of intelligence and adaptability of agents，dynamic fuzzy knowledgebase and relevant fuzzy Q learning algorithm are designed for the agents.The model and algorithm have largely met the needs of intelligence，flexibility，scalability and optimization for grid resource management and scheduling.Simulation experiments show that the proposed model and algorithm works soundly and efficiently.

Key words: grid computing, multi-agent system, grid resource management, fuzzy knowledgebase, fuzzy Q learning

摘要： 由于网格资源的自治、分布、异构和动态变化等特性，如何有效地管理和调度资源是网格计算领域中的一个关键问题，至今仍未得到满意的解决。提出了一种基于Agent的网格资源管理模型，为各类Agent设计了动态模糊知识库，并研究了基于动态模糊知识的模糊Q学习算法。算法较好地满足了网格资源管理中的智能适应性、扩展性及调度优化的需要。通过模拟实验验证了所研究模型和算法的有效性和效率。

关键词: 网格计算, 多Agent系统, 网格资源管理, 模糊知识库, 模糊Q学习