【作者】 龚报钧；

【导师】 王树青；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2001， 博士

【摘要】 Agent技术是分布式智能研究中的热门技术，有关Agent和多Agent系统的应用研究已成为国际上一个生机勃勃的研究领域，并在复杂工业过程智能控制、电力系统管理、人机协同、新型企业组织、智能机器人、信息系统集成与协调等方面取得了成绩。本文在国家自然科学基金项目“基于多智能体的企业信息集成理论及应用研究”（No.79970035）的资助下进行的。为了克服企业信息集成平台普遍存在的适应企业业务变迁、人机协同和集成异质信息资源等方面的能力不足，论文将多Agent理论和技术应用于企业信息集成研究中，结合系统论、人机协同方法、流程再造理论（Business Process Reengineering），提出了基于多Agent的企业开放信息集成体系结构，该结构分为三层：界面层、协调层和模型层，并提出了关于各层的设计理论和实现方法： 1）提出了基于多Agent的支持人机协同的智能用户界面体系结构； 2）提出了支持动态企业模型的多Agent系统模型及其获取方法； 3）研究了模型层获取和实现方法； 4）研究了基于多Agent的企业开放信息集成平台的实现方法。 论文的主要研究工作内容如下： 1．综述了Agent和多Agent技术发展历史、多Agent技术的理论和应用发展现状，就多Agent技术在企业信息集成领域应用的意义和前景进行了探讨。 2．为了提高企业信息集成平台的信息集成能力及系统柔性，提出了基于多Agent系统的开放信息集成体系结构，并描述了在其之中的Agent的结构。该体系结构分为界面层、协调层和模型层，每层均由多Agent系统组成。由于该体系结构清晰的层次性以及多Agent系统所具有的自主能力、社交能力、反应能力和行为理性，不仅使企业信息集成平台具有清晰的结构，而且具有良好开放性、集成性和自适应性。 3．从分析实现人机协同智能用户界面需解决的若干问题入手，探讨了多Agent系统解决这些问题的技术途径；然后提出了基于多Agent系统的人机协同式智能用户界面体系结构，并重点讨论了如何解决多媒体对象展示关系、Agent之间协同、应用语义反馈、用户智能向导等问题。该体系结构具有良好的扩展性和可重用性，支持用户界面按组件（Component）方式进行组合。 4．根据基于多Agent系统的开放信息集成体系结构，协调层的任务11 浙江大学博士学位论文是负责企业业务处理，所以如何支持企业流程再造、虚拟企业、动态企业模型也自然成为该层的重点。从系统论的角度出发分析了企业组织与流程的关系，提出了一种能应用于协调层的支持动态企业模型的多Agds系统结构，该结构能根据企业组织结构、业务流程的动态变化进行动态调整。依据它构筑企业信息集成平台的协调层，能有效提高信息集成平台对动态企业组织和业务流程的适应性并延长企业信息集成平台的生命。 5．协调层多Agent系统（MAS）模型是否全面、堆确地描述了企业业务模型是决定信息集成平台是否具有适应性和通用性的关键。结合信息工程论和面向对象方法，提出了一种基于主题的获取协调层MAS模型的方法，它从企业行为的功能主题出发进行分析，获取相应的MAS模型，从而保证了被获取的协调层MAS模型能全面准确地描述企业业务。 6．为了获得构成模型层的多Ag6llt系统（MAS）樱型，以基于主题协调层MAS模型获取方祛为基础，结合模型层的特性，提出模型层MAS模型获取方法。同时，为了提高Agent技术的实用化程度，看重讨论了在关系型数据库管理系统平台上实现模型层的方法。 7．为了解决企业资源计划系统在开发和实施上面临的困境，从系统工程角度分析了企业管理变革与企业信息化的关系，剖析了企业资源计划系统面临的关键问题；针对这些关键问题，提出了采用基于多Agent系统的开放信息集成体系结构解决上述问题的相关原理和方法。 8．从研究供应链管理产生的社会、经济和技术背景入手，分析了供应链管理研究的现状和面临的问题，从而提出了开展供应链集成系统unte卵ed SUpply Chain System）研究的必要性。以大型炼油工业企业为例，全面分析了进行供应链集成系统研究所依赖的关键技术及其现状。从分析多AgCnt系统的特性入手，论证了供应链集成系统可作为多Agent系统进行企业信息集成应用的一个重要领域，最后提出相关的待研究课题。 9．探讨了采用大众化开发工具实现Agent和多Agent系统的方法、实现基于多Agent的企业信息集成系统的方法、集成异质信息源的方法，并给出实例，从而验证了各章节的研究结果。 最后总结了全文的工作，并对基于多＾gCnt技术进行企业信息集成研究进行了展望。更多还原

【Abstract】 Agent has become an important technology in both artificial intelligence and mainstream computer science. Agent and multi-agent systems (MAS) technology is a rapidly developing area of research. Agent and MAS are being used in broad field: complex industry control system, electricity system management, enterprise information integration and new enterprise organization. This work is support by National Natural Science Foundation of P. R. China (No.79970035). This dissertation is devoted to theoretical and practical study of enterprise information integration based on MAS. The main contents are as follows:1 .The historical development of agent and MAS is reviewed from theory and practice. Great emphasis is place on prospect of integrating enterprise information using MAS.2. To improve Enterprise Information Integrating Platform’s (EIIP) ability of integrating and flexibility, this paper proposes an open architecture of information integration based on multi-agent (OAIIA) and describes the structures of the agent in OAIIA. OAIIA consists of three tiers: interface tier, coordination tier and model tier, and every tier is composed with multi-agent. Due to the explicit three-tier structure of OAIIA and the autonomy, sociability, reactivity and rationality of multi-agent, EIIP based on OAIIA has many advantages such as good open-ability, integration-ability and self-adaptability. At last, the implementation technique and a case of supporting dynamic business process based on OAIIA are also put forward.3. This paper proposes the intelligent interface architecture for human-computer cooperation named IIAMA (Intelligent Interface Architecture based on Multi-Agent), and discusses in detail how to deal with temporal relation of multimedia object, intelligent guiding and application semantic in IIAMA. IIAMA is based on multi-agent model in which each visual element cooperates with each other for providing continuous support to a user. The control agent coordinates communication among visual elements, and the user agent provides user interest model and reasoning service. Besides, IIAMA supports larger grain reuse and flexible interface system composition.4. The advent of the new enterprise paradigms of extended, virtual andagile enterprises causes new requirement on organizational concepts and supporting technologies. Inter- and intra- organizational relations need enterprise information system (EIS) to support their establishment, deployment and discontinuation. Due to dynamic business process is one of the most important features of virtual enterprise, we propose the model of multi-agent system (MAS) for supporting dynamic enterprise model (DEM). It separates the business process from the organizational structure (organizational structure tier and business process tier), models each of them as MAS, and coordinates agents by "yellow page" mechanism. This model not only can regulate itself in terms of DEM, but also centered on the coordination strategies between agents composing it. It is believed that the model of MAS is a practical way to build flexible EIS.5. The flexibility of enterprise information platform is decided by that the model of multi-agent system is the image of the model of enterprise business or not. By combining information engineering with object-orient, a method based on subject is present, which getting model of multi-agent in control tier in terms of the analysis of function subject of enterprise. An example is also discussed.6. Method for getting MAS model in model tier is preset, which by analyzing function subject of enterprise behavior to get object model, and transfer object model into MAS model. In order to enhance the practicality of agent technology, how to realize model tier based on MAS is discussed. The implementation technique and a case of model tier based on MAS are also put forward.7. The relation of management between enterprise information system is studied from the system point of view and some key problems of Enterprise Resource Planning (ERP) are present. I更多还原