基于蚁群Stigmergy协作机制和ANN的多智能制造主体协同方法研究

发布时间：2018-02-13 20:30

本文关键词： 多主体协同感知 Storm-Hadoop ANN多物理域信息融合 Stigmergy协作机制　出处：《深圳大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着消费需求日益多样化与个性化,传统的刚性制造系统已经越来越难以满足个性化定制、小批量甚至单件生产等产品制造要求。并且随着制造业、制造技术和制造系统的不断发展、制造过程的复杂性等要求制造系统实体资源/设备等制造主体具备一定程度的自动化甚至智能化,因此就需要实现制造主体的自主感知、判断、推理决策和自主学习等等。当前传统制造系统正逐渐向着智能制造系统方向迈进,而智能制造系统是由多制造主体分布式耦合而成的有机制造系统,各制造主体间存在着高度的耦合关系,主体间通过相互协作、群体协同决策以完成系统任务。因此这就要求制造系统中各制造主体能够更好地根据各自状态以及系统整体状态进行相互协作,也就是在多制造主体协同感知的前提下实现制造系统各制造主体的协同。本文对多智能制造主体及其协同感知相关问题进行了探讨研究,从智能制造主体的定义与建模研究、CPS-Agent模型以及智能制造系统模型等方面研究其基本理论,并对多制造主体协同感知问题进行了描述,从而为其建模的实现奠定基础。针对多智能制造主体协同感知问题的建模,本文对比分析了集中式系统与分布式系统中的各种机制,重点研究了基于蚁群Stigmergy的多智能制造主体协同感知机制,并将其应用到解决多智能制造主体群体协同决策问题,从而引出分布式协同感知问题。在对目前流行的分布式系统调研的基础上,对分布式系统Storm和Hadoop进行了充分的分析和研究,并将两者进行融合,形成一种混合分布式系统,即Storm-Hadoop分布式集群系统,得到基于多CPS-Agent和Storm-Hadoop的多智能制造主体协同感知框架。为解决多智能制造主体多物理域信息融合问题,对基于ANN的智能制造主体多物理域信息融合方法进行了分析研究,对智能制造主体的多物理域信息进行了分析与描述,讨论了智能制造主体多物理域信息的获取与处理方法,并对基于ANN的多物理域信息融合进行仿真研究,为多智能制造主体群体协同决策任务分解奠定基础。针对多智能制造主体协同群体决策问题如任务协同分解与分配,着重研究了基于Stigmergy协作机制的多智能制造主体群体协同决策方法,对算法设计、原理进行了详细阐述,并进行了仿真研究,得到较为满意的结果。最后,对以上提出的相关理论和算法,搭建了Storm-Hadoop集群,利用Visual Studio2012设计开发了原型系统,对多物理域信息融合、系统群体协同决策任务分解进行了仿真试验验证,得到了相应的结果,并对结果进行了总结分析,实现了本文所提出的基于蚁群Stigmergy和ANN多智能制造主体的协同状态感知、任务协同分配。
[Abstract]:With the increasing diversification and individuation of consumer demand, the traditional rigid manufacturing system has become more and more difficult to meet the requirements of personalized customization, small batch and even single-piece production. The continuous development of manufacturing technology and manufacturing system and the complexity of manufacturing process require manufacturing agents such as manufacturing system physical resources / equipment to have a certain degree of automation or even intelligence, so it is necessary to realize the independent perception of manufacturing agents. Judgment, reasoning, decision making, autonomous learning, etc. At present, traditional manufacturing systems are gradually moving towards intelligent manufacturing systems, and intelligent manufacturing systems are organic manufacturing systems, which are made up of distributed coupling of multiple manufacturing agents. There is a high degree of coupling among manufacturing agents. Therefore, the manufacturing agents in manufacturing system should be able to cooperate with each other according to their respective states and the overall state of the system. That is to realize the coordination of manufacturing agents in manufacturing system under the premise of multi-agent collaborative perception. In this paper, the problems related to multi-intelligent manufacturing agents and their cooperative perception are discussed. The basic theory of intelligent manufacturing agent model and intelligent manufacturing system model are studied from the definition and modeling of intelligent manufacturing agent, and the cooperative perception problem of multiple manufacturing agents is described. In order to establish the foundation for its modeling. Aiming at the modeling of multi-intelligent manufacturing agent cooperative perception, this paper compares and analyzes various mechanisms in centralized system and distributed system. The cooperative awareness mechanism of multi-intelligent manufacturing agents based on ant colony Stigmergy is studied and applied to solve the cooperative decision-making problem of multi-intelligent manufacturing agents. On the basis of the research on the popular distributed system, Storm and Hadoop are fully analyzed and studied, and the two systems are combined to form a hybrid distributed system. In order to solve the problem of multi-intelligent manufacturing agent multi-physical domain information fusion, a multi-intelligent manufacturing agent cooperative awareness framework based on multiple CPS-Agent and Storm-Hadoop is obtained in Storm-Hadoop distributed cluster system. This paper analyzes and studies the multi-physical domain information fusion method of intelligent manufacturing agent based on ANN, analyzes and describes the multi-physical domain information of intelligent manufacturing agent, and discusses the method of obtaining and processing multi-physical domain information of intelligent manufacturing agent. The simulation research of multi-physical domain information fusion based on ANN is carried out, which lays a foundation for task decomposition of multi-intelligent manufacturing agent group collaborative decision making, aiming at multi-intelligent manufacturing agent collaborative group decision making problem such as task cooperative decomposition and assignment. The collaborative decision method of multi-intelligent manufacturing agent group based on Stigmergy cooperation mechanism is studied emphatically. The algorithm design and principle are described in detail, and the simulation results are given, and the results are satisfactory. Based on the above theories and algorithms, the Storm-Hadoop cluster is built, and the prototype system is designed and developed by using Visual Studio2012. The simulation results of multi-physical domain information fusion and group cooperative decision task decomposition of the system are verified, and the corresponding results are obtained. The results are summarized and analyzed, and the cooperative state awareness and task co-allocation based on ant colony Stigmergy and ANN are realized.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TH16

【相似文献】