多Agent聚合模型的研究与实现

发布时间：2018-04-18 22:21

本文选题：多Agent系统 + 多Agent聚合模型　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：分布式、异构资源的动态聚合是实现资源共享和互操作的重要内容。通过资源的动态聚合，可以使用户根据自己的聚合任务需求来发现、利用资源，进而实现其需求。分布式异构资源在时空、类型等方面的多源异构性，给资源的动态聚合带来了困难。本文通过将多Agent技术引入资源聚合过程中，研究了利用Agent技术实现资源动态聚合的模型和实现机制。本文采用多Agent的资源动态聚合流程模型描述聚合资源之间的关系。在此基础上，给出了一种支持异构资源动态聚合的分布式多Agent聚合模型框架及聚合模型中两类主要的Agent——协调Agent和执行Agent的通用结构。鉴于用户提交的聚合任务需求通常比较复杂，模型中的Agent很难独立完成，需对复杂的聚合任务需求进行分解，将分解后的任务分配给有能力执行的Agent。因此本文又给出了多Agent聚合模型中聚合任务的分解机制。为了有效支持Agent在聚合模型中的协作特性，提高聚合模型运作的效率，本文提出了一种基于FIPA合同网标准的改进合同网协议。在招标阶段引入“黄页”查找与熟人库相结合的方法，在协商过程中引入执行Agent执行某项任务的可用度(DOA)、信任度(trust)、友好度(friend)参数，并考虑了新加入系统的Agent，在中标阶段采用模糊综合评判的中标策略，并对协议流程进行了改进。在保证协商质量的基础上有效地提高了协商的效率。最后，给出了JADE平台上多Agent聚合模型及其应用的改进合同网协议的实现机制，包括多Agent聚合模型具有容错结构的分布式实施结构以及模型中几类主要功能Agent的类结构。之后，给出了一个分布式多Agent聚合模型实例——多Agent情报处理系统。以该实例为测试环境，设计了一组针对性实验对本文提出的多Agent聚合模型及模型中应用的改进合同网协议的有效性进行了验证。
[Abstract]:Dynamic aggregation of distributed and heterogeneous resources is an important part of resource sharing and interoperability.Through the dynamic aggregation of resources, users can discover and utilize resources according to their aggregation task requirements, and then realize their requirements.The multi-source heterogeneity of distributed heterogeneous resources in time and space, types and other aspects brings difficulties to the dynamic aggregation of resources.In this paper, we introduce multiple Agent technology into the process of resource aggregation, and study the model and implementation mechanism of dynamic resource aggregation using Agent technology.In this paper, the dynamic aggregation process model of multiple Agent is used to describe the relationship between aggregated resources.On this basis, a framework of distributed multi- aggregation model supporting dynamic aggregation of heterogeneous resources and two main types of Agent-coordinated Agent and general architecture of executing Agent in the aggregation model are presented.Since the requirements of aggregation tasks submitted by users are usually complex, the Agent in the model is difficult to complete independently. It is necessary to decompose the complex aggregate task requirements and assign the decomposed tasks to the agents with the ability to execute.Therefore, the decomposition mechanism of aggregation tasks in multi-Agent aggregation model is also given in this paper.In order to effectively support the collaboration of Agent in the aggregation model and improve the efficiency of the aggregation model, an improved contract net protocol based on the FIPA contract net standard is proposed in this paper.In the bidding stage, the method of "yellow page" lookup combined with the acquaintance bank is introduced, and during the negotiation process, the availability of Agent to perform a task is introduced.Considering the new AgentS, the fuzzy comprehensive evaluation strategy is adopted in the bid winning stage, and the protocol flow is improved.The efficiency of negotiation is improved effectively on the basis of ensuring the quality of negotiation.Finally, the implementation mechanism of multiple Agent aggregation model and its application on JADE platform is given, including the distributed implementation structure with fault-tolerant structure and the class structure of several main functions of Agent in the model.Then, an example of distributed multiple Agent aggregation model, multiple Agent information processing system, is presented.Taking this example as the test environment, a set of targeted experiments are designed to verify the validity of the proposed multi- aggregation model and the improved contract net protocol used in the model.
【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】：TP18;TP393.01

【参考文献】