基于MAS技术的生产调度方法的研究

生产调度是一类复杂的系统性问题。本文探讨了多代理系统(multi—agent system，MAS)技术在生产调度研究领域的应用。在介绍传统调度方法和多代理调度系统(multi—agent scheduling system，MASS)的基础上，提出了基于多代理机制的调度实现方案，给出了MASS的体系结构。文末对以后的研究做出了一些展望。     

基于多代理技术的申请审批系统

论述了基于多代理的申请审批系统的研究和开发，提出了基于多代理的申请审批系统结构模型和内部运作机制。通过利用MAS自治性、交互性的特点，较好地解决了传统申请审批系统中存在的问题。     

基于MAS技术的电梯群控系统建模及

将多代理系统（MAS）技术引入电梯群控系统,形成分布式电梯群控系统,以克服集中式群控系统的计算瓶颈问题.提出了基于MAS技术的网状电梯群控系统总体结构,将电梯交通系统中的轿厢和呼梯信号控制器分别映射为C-agent和HC-agent,并增加了一个M-agent映射电梯系统中的虚拟管理器.描述了基于拟市场模型的C-agent报价算法和HC-agent评价报价算法,以及分布式电梯群控系统的全局优化机制.     

基于多代理系统的软硬件协同设计

为使软硬件协同设计过程更具分布性、自主性及并行性,在软硬件协同设计中引入多代理(MAS)技术,提出软硬件协同设计的MAS模型,包括系统描述Agent、软硬件划分及映射Agent、软硬件设计Agent、协同通信Agent、性能评估Agent和硬件系统测试Agent的构建和应用。采用多个目标代理映射、协商的方法协调整个协同设计过程。实际应用表明,该方法能优化系统级芯片设计方案、软硬件结构和功能,并提高系统整体性能。     

分布式电梯群控系统多目标调度算法

基于多代理系统(MAS)的分布式电梯群控系统将电梯及群控器映射为具有不同功能的代理(agent), 呼梯信号的分派通过各agent协商解决, 使梯群调度算法的计算工作量分散到各agent. 基于拟市场模型, 分布式群控算法主要包括较厢代理(C-agent)报价算法及呼梯信号代理(HC-agent)电梯分派算法等. 设计实现了旨在同时降低平均候梯时间、平均乘梯时间及长候梯率的多目标分布式群控算法DMO. 仿真结果表明, 基于MAS的分布式电梯群控系统是可行的, 所设计的分布式群控算法能够使平均候梯时间、平均乘梯时间及长候梯率同时得到优化.     

MAS技术在ITS中的应用研究

多代理技术(MAS)是当今世界研究的热点。该文结合智能交通系统(ITS)具体应用,详细介绍了智能信息处理领域中MAS技术在ITS中的运用,提出了基于MAS技术的ITS系统构架,设计了智能Agent模块,实现了车辆检测Agent和参数Agent处理,具有较强的实用性。最后,讨论了MAS在ITS中的网络化设计。     

多代理系统下的主动配电网故障恢复博弈策略

在保障主动配电网可靠恢复前提下,为实现供电侧与用户侧利益均衡,本文提出了多代理系统(multiagent system,MAS)下的主动配电网故障恢复博弈策略.为充分考虑电网和电力用户对故障恢复的决策影响,设计了由电网代理、用户总代理和协调代理构成的MAS,建立了MAS信息传输模型.然后,构建了供电侧和用户侧在故障恢复中的利益函数、恢复策略空间,以及以电网代理和用户总代理作为参与人的合作博弈恢复模型.电网代理和用户总代理分别以改进的蚁群算法和统计方法进行分布并行计算,协调代理以双方共同的利益函数作为寻优目标通过迭代算法求得满足纳什均衡的恢复策略.本文以IEEE69节点模型为例,分别对单故障和连锁故障恢复情景求解,验证了本文所提策略的有效性.     

Intranet环境下基于Multi—agent机制的生产调度系统研究

提出一种基于多代理机制的系统实现方案。系统中各代理对等分布，通过有机协作来完成调度问题的求解。主管代理的引入使系统的功能可配置，状态可监测。为保证应用系统的安全运行，提出了一些安全措施。     

基于多代理的炼钢-连铸生产调度方法

多代理机制的分布式结构和实时处理能力为炼钢-连铸生产动态调度系统的实现提供了可行的技术方法。本文通过对炼钢-连铸生产调度问题的简单分析,提出炼钢-连铸生产调度的多代理体系框架、各个代理的功能及主要代理的详细设计,探讨了基于多代理的炼钢-连铸生产调度机制,并给出了一个简单的炼钢。连铸生产调度的实例分析结果。     

基于MAS的动态企业建模方法

动态企业建模技术是为适应企业多变的市场环境和经常的企业业务过程重构而发展起来的技术;而Agent技术和MAS系统具有自治性、社会性、理性等突出特点。对运用具有不同性质的Agent构成三层企业模型MAS的方法进行了深入的研究,它在不同层次上支持企业业务、组织结构的重组,增加了企业模型的柔性。     

基于消息报告机制的多Agent调度系统研究

针对生产车间调度系统中因资源庞大以及动态变化而引起的监管困难问题,构建了一种 基于消息报告机制的多代理调度系统分层体系结构。根据Agent在调度系统中的作用和角色的不 同,将Agent划分为管理Agent、任务Agent和资源Agent,并通过一个具有消息报告机制的管理Agent 来有效管理和监控系统的运行。实例分析表明,消息报告机制不仅能保证系统的有效运行,而且能提 高系统的工作效率。     

The application of a system-based risk management framework and social network analysis to the Maritime Autonomous Surface Ship system: Who are the decision-makers in the wider system?

Maritime Autonomous Surface Ships (MASS) currently have no formal regulations developed specifically for their operation, as their regulatory framework is still under development. Rasmussen's Risk Management Framework has been used to develop an actor map of the current MASS system in the UK, to show who the actors, decision-makers, and planners are within the wider sociotechnical system and the level at which they sit. From the actor map, two social networks were created, one to show the connections that currently exist between the actors within the MASS system and another to show what a future MASS system could look like if regulations and standards were put in place for MASS. Social Network Analysis was then used to investigate the wider MASS system's dynamics, to understand which actors currently have a high degree of influence within the UK MASS system, and where the shortfalls are in the current MASS system. The analysis showed that the industry and end user levels lacked support from the higher system levels, and the addition of formal regulations and standards in the future MASS system would increase the MASS system's resilience. System recommendations for each level in the Risk Management Framework were then made to suggest ways to increase the influence of the regulators and promote the safe operation of MASS.     

Multiagent based dynamic resource scheduling for distributed multiple projects using a market mechanism

The resource scheduling problem in a multi-project environment extends job-shop scheduling problems by allowing for task dependency and multiple self-interested entities. In this paper we deal with short-term scheduling of resources, which are shared by multiple projects. In specific, we address the dynamic nature of the situation. We model this as a dynamic economy, where the multiple local markets are established and cleared over time, trading resource time slots (goods). Due to the dynamic and distributed nature of the economy, through our approach we can achieve higher levels of flexibility, scalability and adaptability. Unlike most market-based mechanisms, which are based on equilibrium concepts and iterative adjustment of resources prices, we propose a novel market mechanism called precedence cost tâtonnement (P-TâTO), which solves individual resource-constrained local resource scheduling in an optimal way, and searches for a precedence conflict-free schedule through a tâtonnement type procedure. In this paper, we discuss our dynamic economy model and some details of the market mechanism along with empirical analysis results.     

Integrated process planning and scheduling by an agent-based ant colony optimization

This paper presents an ant colony optimization (ACO) algorithm in an agent-based system to integrate process planning and shopfloor scheduling (IPPS). The search-based algorithm which aims to obtain optimal solutions by an autocatalytic process is incorporated into an established multi-agent system (MAS) platform, with advantages of flexible system architectures and responsive fault tolerance. Artificial ants are implemented as software agents. A graph-based solution method is proposed with the objective of minimizing makespan. Simulation studies have been established to evaluate the performance of the ant approach. The experimental results indicate that the ACO algorithm can effectively solve the IPPS problems and the agent-based implementation can provide a distributive computation of the algorithm.     

求解模糊柔性Job-shop调度问题的多智能体免疫算法

考虑实际纸盆车间调度中模具、机器、操作人员等资源约束,以及加工时间和交货日期的不确定性,建立了批量可变的模糊柔性Job-shop调度问题模型.结合多智能体系统以及免疫信息处理机制,构造了一种求解实际Job-shop调度问题的多智能体免疫算法.该方法通过竞争、自学习、自适应疫苗接种、模拟退火等操作,更新每个智能体在解空间的位置,从而能精确地收敛到全局最优解.纸盆车间调度实例的求解结果验证了该算法的有效性.     

基于优先数的智能生产调度系统

生产调度是企业生产中的关键环节,也是实施CIMS的重点。该文介绍了一种基于优先数的动态、多目标优化生产调度方法,并构造了一个基于优先数的智能生产调度系统。系统通过Agent对生产调度问题中的各个目标优化,给出不同目标的指标值,同时给出企业在不同时候对不同目标的偏好值,计算各个生产任务的优先数,根据任务的优先数安排生产。     

基于Agent图的三级多Agent系统组织模型

多Agent系统的组织结构是Agent个体之间交互的框架。对分布式多Agent系统的组织方式、协作机制进行了简要讨论,提出了Agent域及Agent图的概念。根据不同Agent之间的地理位置和通信代价,由Agent个体、Agent组及Agent域三级组织结构形成一个Agent图,并借鉴计算机网络的分布式自适应路由选择策略进行多Agent系统的协作组织。分析表明,该模型具有高效、健壮、通信开销较小等优点。     

Dynamic parts scheduling in multiple job shop cells considering intercell moves and flexible routes

Aiming at the problem of scheduling with flexible processing routes and exceptional parts that need to visit machines located in multiple job shop cells, a pheromone based approach (PBA) using multi-agent is presented in this paper, in which various types of pheromone inspired by ant colony optimization (ACO) are adopted as the basis of negotiation among agents. By removing redundant routes and constructing coalition agents, communication cost and negotiation complexity are reduced, and more importantly, the global performance of scheduling is improved. The performance of the PBA is evaluated via experiments with respect to the mean flow time, maximum completion time, mean tardiness, ratio of tardy parts, and ratio of intercell moves. Computational results show that compared with various heuristics, the PBA has significant advantages with respect to the performance measures considered in this paper.     

基于多Agent协商的柔性车间调度系统

针对柔性作业车间调度的特点,提出了一种基于多agent协商的柔性作业车间调度系统。系统由工件agent,机器agent和工序agent组成。Agent之间通过相互发送消息和响应消息进行交互,并且通过消息相应函数按照各agent局部的信息、同时兼顾系统的性能进行决策。工件agent通过招标的方式,选择合适的机器完成加工任务,机器agent通过竞争来获得工件的加工权。最后用Java语言在Eclipse平台上进行程序设计,对柔性作业车间调度的平均滞后问题进行仿真实验,并与传统的分派规则比较,结果显示所提方法的优越性。     

随机博弈框架下的多agent强化学习方法综述

多agent学习是在随机博弈的框架下,研究多个智能体间通过自学习掌握交互技巧的问题.单agent强化学习方法研究的成功,对策论本身牢固的数学基础以及在复杂任务环境中广阔的应用前景,使得多agent强化学习成为目前机器学习研究领域的一个重要课题.首先介绍了多agent系统随机博弈中基本概念的形式定义;然后介绍了随机博弈和重复博弈中学习算法的研究以及其他相关工作;最后结合近年来的发展,综述了多agent学习在电子商务、机器人以及军事等方面的应用研究,并介绍了仍存在的问题和未来的研究方向.