面向自主网格的多Agent动态协同图构造算法

Autonomic grid-oriented algorithm for multi-agent dynamic cooperative graph construction

网格环境的复杂性和动态性迫切需要自主计算技术的支持。在前期工作中给出地自主网格体系结构基础上,为解决任务执行过程中资源或服务失效情况下多Agent间协同的问题,提出了多Agent动态协同图的概念和任务偏序集驱动地多Agent动态协同图构造算法。图中的顶点是由Agent和自主网格服务组成的序偶,构造算法由任务偏序集到服务集的映射,逐层构造图中的顶点。该图通过Agent对本地服务的感知和Agent间的通信,达到任务执行过程中服务间自主协同的目标。模拟实验的结果验证了算法的正确性,表明算法的时间复杂度主要由任务哈斯图的层数决定,并且Agent的感知时间具有鲁棒性。

Complicated and dynamic environment of grid needs urgently supporting of autonomic technology.In order to solve difficult problem of out-of-order of resources or services during tasks implemented.This paper proposes the notion of multi-agent dynamic cooperative graph and the algorithm of multi-agent dynamic cooperative graph construction driven by task posers based on autonomic grid architecture given by previous work.The vertex is ordered pair composed of agent and autonomic grid service. The graph is constructed by layers of the vertex on mapping from task posets to service set.The graph achieves the goal of auto- nomic cooperation between services during tasks implemented by means of the perception of agent to local services and the communication between agents.Running of simulation experiment verifies the algorithm is effective.It also shows that time complexity of the algorithm is decided by layer numbers of task Hasse graph and that perception time of agent takes on robustness.