基于合同网机制的柔性智能车间调度系统建模与仿真
发布时间：2018-11-28 11:46
【摘要】：车间调度问题由于其NP-Hard特性，传统集中式的解决方案由于其求解消耗随着问题规模的日渐增长指数级上涨而显得臃肿不堪，而在更切合实际环境的柔性制造环境下更是无法满足其复杂性，动态性，随机性的需求。在这样的背景下，很多学者都开始研究引进更适合问题模型的基于自治与协调的自适应系统。本文在结合分析近代国内外相关研究成果的基础上，以具有柔性生产加工的离散作业制造车间作为研究对象，建立了基于Agent单元和合同网协商机制的自治与协商模型，并为系统模型引入了Q-Learn算法和工序价值评估算法进行局部优化，为柔性车间调度问题提供了有效实用的解决方案。 本文主要内容可概括为： 1．首先针对柔性车间调度问题，对近年来的相关研究成果与发展现状进行简单概括与分析，进而提出本文研究的出发点与意义。 2．针对问题的复杂性、不确定性、多约束多资源互相协调的特点，提出以复杂适应系统（CAS）理论为基础，Agent为基本单元，基于合同网协商机制的柔性车间调度问题的实时调度框架模型。对系统框架进行利弊分析，指出系统骨架的合同网缺乏优化和动态学习的问题。 3．将Q-Learn强化学习算法引进系统模型中的标书评估决策，并根据算法的具体要求对Q-Learn算法的模型进行了详细定义和描述，赋予合同网协议动态智能学习能力。 4．在标书评估过程中提出了工序价值的概念，将系统目标量化为“价值”的形式附加到子任务中，对系统进行局部优化。 5．最后，使用Java+Swarm+Matlab工具，设计和开发了综合的仿真系统，对文中的设计与观点进行仿真测试，为该理论在问题的实际应用上做出可行性尝试。本文对解决柔性制造环境下的车间调度问题提出了一个合理的通过仿真证明其可行性的基于Agent的自治与协商调度方案，在一定程度上推动了该方向对问题的求解研究。
[Abstract]:Because of its characteristic of NP-Hard, the traditional centralized solution of job shop scheduling problem appears to be bloated because of its solution consumption increasing exponentially with the increasing scale of the problem. And in the more practical environment of flexible manufacturing environment is not able to meet its complexity, dynamic, random needs. In this context, many scholars have begun to study the introduction of adaptive systems based on autonomy and coordination that are more suitable for problem models. Based on the analysis of relevant research results at home and abroad in modern times, this paper takes the discrete job shop with flexible production and processing as the research object, and establishes the autonomy and negotiation model based on Agent unit and contract net negotiation mechanism. The Q-Learn algorithm and the process value evaluation algorithm are introduced to the system model for local optimization, which provides an effective and practical solution for the flexible job shop scheduling problem. The main contents of this paper can be summarized as follows: 1. Firstly, this paper summarizes and analyzes the related research results and development status in recent years, and then puts forward the starting point and significance of this study. 2. 2. In view of the complexity, uncertainty, multi-constraint and multi-resource coordination of the problem, this paper proposes a complex adaptive system based on (CAS) theory and Agent as the basic unit. A real-time scheduling framework model for flexible job shop scheduling problem based on contract net negotiation mechanism. This paper analyzes the advantages and disadvantages of the system framework and points out the lack of optimization and dynamic learning of the contract network of the system skeleton. 3. The Q-Learn reinforcement learning algorithm is introduced into the bidding evaluation decision of the system model, and the model of the Q-Learn algorithm is defined and described in detail according to the specific requirements of the algorithm, which endows the contract net protocol with dynamic intelligent learning ability. 4. In the process of bid evaluation, the concept of process value is put forward, and the system objective is quantified as "value", which is attached to the sub-task, and the system is locally optimized. 5. Finally, a comprehensive simulation system is designed and developed by using Java Swarm Matlab tool. The design and viewpoint of this paper are simulated and tested, which makes a feasible attempt for the practical application of the theory. In this paper, a reasonable Agent based autonomy and negotiation scheduling scheme is proposed to solve the job shop scheduling problem in flexible manufacturing environment, which is proved to be feasible by simulation. To some extent, it promotes the research of solving the problem in this direction.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH186;N945.12
本文编号：2362739
[Abstract]:Because of its characteristic of NP-Hard, the traditional centralized solution of job shop scheduling problem appears to be bloated because of its solution consumption increasing exponentially with the increasing scale of the problem. And in the more practical environment of flexible manufacturing environment is not able to meet its complexity, dynamic, random needs. In this context, many scholars have begun to study the introduction of adaptive systems based on autonomy and coordination that are more suitable for problem models. Based on the analysis of relevant research results at home and abroad in modern times, this paper takes the discrete job shop with flexible production and processing as the research object, and establishes the autonomy and negotiation model based on Agent unit and contract net negotiation mechanism. The Q-Learn algorithm and the process value evaluation algorithm are introduced to the system model for local optimization, which provides an effective and practical solution for the flexible job shop scheduling problem. The main contents of this paper can be summarized as follows: 1. Firstly, this paper summarizes and analyzes the related research results and development status in recent years, and then puts forward the starting point and significance of this study. 2. 2. In view of the complexity, uncertainty, multi-constraint and multi-resource coordination of the problem, this paper proposes a complex adaptive system based on (CAS) theory and Agent as the basic unit. A real-time scheduling framework model for flexible job shop scheduling problem based on contract net negotiation mechanism. This paper analyzes the advantages and disadvantages of the system framework and points out the lack of optimization and dynamic learning of the contract network of the system skeleton. 3. The Q-Learn reinforcement learning algorithm is introduced into the bidding evaluation decision of the system model, and the model of the Q-Learn algorithm is defined and described in detail according to the specific requirements of the algorithm, which endows the contract net protocol with dynamic intelligent learning ability. 4. In the process of bid evaluation, the concept of process value is put forward, and the system objective is quantified as "value", which is attached to the sub-task, and the system is locally optimized. 5. Finally, a comprehensive simulation system is designed and developed by using Java Swarm Matlab tool. The design and viewpoint of this paper are simulated and tested, which makes a feasible attempt for the practical application of the theory. In this paper, a reasonable Agent based autonomy and negotiation scheduling scheme is proposed to solve the job shop scheduling problem in flexible manufacturing environment, which is proved to be feasible by simulation. To some extent, it promotes the research of solving the problem in this direction.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH186;N945.12
【参考文献】
相关期刊论文 前10条
1 谭跃进,邓宏钟;复杂适应系统理论及其应用研究[J];系统工程;2001年05期
2 黄炳强;曹广益;王占全;;强化学习原理、算法及应用[J];河北工业大学学报;2006年06期
3 马鑫;梁艳春;;基于GPGP协同机制的多Agent车间调度方法研究[J];计算机研究与发展;2008年03期
4 胡舜耕;张莉;钟义信;;多Agent系统的理论、技术及其应用[J];计算机科学;1999年09期
5 张海俊,史忠植;动态合同网协议[J];计算机工程;2004年21期
6 杨萍;孙延明;刘小龙;车兰秀;;基于细菌觅食趋化算子的PSO算法[J];计算机应用研究;2011年10期
7 高阳,陈世福,陆鑫;强化学习研究综述[J];自动化学报;2004年01期
8 熊锐,吴澄;车间生产调度问题的技术现状与发展趋势[J];清华大学学报(自然科学版);1998年10期
9 刘大有,杨鲲,陈建中;Agent研究现状与发展趋势[J];软件学报;2000年03期
10 李冬梅,陈卫东,席裕庚;基于强化学习的多机器人合作行为获取[J];上海交通大学学报;2005年08期
相关博士学位论文 前2条
1 鞠全勇;智能制造系统生产计划与车间调度的研究[D];南京航空航天大学;2007年
2 王世进;基于自治与协商机制的柔性制造车间智能调度技术研究[D];上海交通大学;2008年
相关硕士学位论文 前1条
1 王雪辉;基于多智能体的车间调度系统的研究[D];河北工业大学;2005年
，本文编号：2362739
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2362739.html
