基于多Agent系统的资源服务聚合模型研究

发布时间：2018-03-08 14:12

本文选题：多Agent系统　切入点：资源服务聚合模型　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着计算机技术和网络技术的发展与广泛应用,分散在网络中资源的种类和数量也越来越多。用户可以通过资源的聚合,将这些分布式异构的资源重新组织起来,呈现出特定的服务,满足自己的需求。然而,数据资源庞大、分布异构等因素给资源服务的聚合带来了困难。本文研究多Agent系统中的资源服务聚合问题,在资源服务聚合过程中,分析和讨论利用多Agent技术来实现资源服务聚合的方法。本文给出了一种基于多Agent系统的资源服务聚合模型,以解决分散资源聚合问题。论文分析了模型的层次结构、各个层次的职能分工、层次之间的关系以及模型中各类Agent的结构;针对用户提交的聚合任务通常难以由Agent独立完成的问题,给出了一种任务分解方法,使得分解后的任务能够由模型中的Agent来执行,实现分解的复杂任务可以由多个Agent之间的协作来完成。针对资源服务聚合模型的特点,本文在FIPA合同网协议的基础上进行了扩展,来支持模型中Agent之间的协作,及任务的分配。在给出的扩展合同网协议中,招标阶段使用基于公共信息板的招标策略,显著减少了系统的通信量;投标阶段使用自主投标的策略,Agent可以根据自身的实时状态来决定是否投标;中标阶段引入案例推理思想,结合历史时期内解决问题的经验和反映Agent实时能力的信任度、活跃度和可用度系数来对当前问题进行决策,并对任务执行失败后的处理机制进行了改进,提升了协作的效率。最后,论文在JADE平台上给出了一个战场目标处理系统实例,对论文所给的聚合模型和扩展合同网协议进行验证,实验结果表明:本模型及其应用的扩展合同网协议能够完成资源服务的动态聚合,对相似任务具有分辨能力,在任务执行失败后能重新指定中标Agent来避免重新发起任务招标,并且完成相同任务的通信量和时间都少于FIPA合同网。
[Abstract]:With the development and wide application of computer technology and network technology, there are more and more resources scattered in the network. Users can reorganize these distributed and heterogeneous resources through the aggregation of resources. However, large data resources, heterogeneous distribution and other factors bring difficulties to the aggregation of resource services. In this paper, the problem of resource service aggregation in multiple Agent systems is studied. In the process of resource service aggregation, this paper analyzes and discusses the method of implementing resource service aggregation by using multiple Agent technology. In this paper, a model of resource service aggregation based on multi-#en1# system is presented. In order to solve the problem of decentralized resource aggregation, this paper analyzes the hierarchical structure of the model, the function division of each level, the relationship between the different levels and the structure of various Agent in the model. In order to solve the problem that aggregation tasks submitted by users are usually difficult to be completed by Agent alone, a task decomposition method is proposed to enable the decomposed tasks to be executed by the Agent in the model. The complex task of implementing decomposition can be accomplished by the cooperation of multiple Agent. According to the characteristics of resource service aggregation model, this paper extends the FIPA contract net protocol to support the cooperation between Agent in the model. In the proposed extended contract net protocol, the bidding strategy based on public information board is used in the bidding stage, which significantly reduces the communication volume of the system. In the bidding stage, the strategy of independent bidding can be used to decide whether to bid or not according to its own real-time state, and the case-based reasoning (CBR) is introduced in the bidding stage, which combines the experience of solving problems in the historical period and the trust degree reflecting the real-time capability of Agent. The active degree and availability coefficient are used to make decision on the current problem, and the processing mechanism after the failure of task execution is improved to improve the efficiency of cooperation. Finally, an example of battlefield target processing system is given on JADE platform. The experimental results show that this model and the extended contract net protocol can achieve the dynamic aggregation of resources and services, and have the ability to distinguish similar tasks. After the task execution fails, the Agent can be reassigned to avoid relaunching the task bidding, and the amount of communication and time to complete the same task is less than that of the FIPA contract net.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】