MAS动态协作任务求解模型与算法

Modeling and solving dynamic collaborative tasks in a multi-Agent system

针对网格环境的自治性、动态性、分布性和异构性等特征.提出基于多智能体系统(mutil agent system,MAS)博弈协作的资源动态分配和任务调度模型,建立了能够反映供求关系的网格资源调度动态任务求解算法,证明了资源分配博弈中Nash均衡点的存在性、惟一性和Nash均衡解.该方法能够利用消费者Agent的学习和协商能力,引入消费者的心理行为,使消费者的资源申请和任务调度具有较高的合理性和有效性.实验结果表明,该方法在响应时间的平滑性、吞吐率及任务求解效率方面比传统算法要好,从而使得整个资源供需合理、满足用户QoS要求.

A grid environment is characterized by its autonomy, its dynamic properties, its distributive properties, and its heterogeneity. We proposed a model for dynamic resource distribution and task scheduling based on a multiagent system （MAS） collaborative game. An algorithm for dynamically solving task scheduling of grid resources was developed. It reflected actual relationships between supply and demand. The existence and uniqueness of a Nash equilibrium point in the resource distribution game was proven, and then the Nash equilibrium solution presented. The proposed method can make full use of the learning and negotiating abilities of consumer agents and also intro- duces psychologically driven behavior. In this way the resource application and task scheduling of consumers became more reasonable and effective. Experimental results demonstrated that this approach improves smoothness, throughput capacity and task solving efficiency compared to traditional methods. Supply and demand became more manageable, meeting the requirements of quality of service （QoS）.