CASIA OpenIR  > 毕业生  > 博士学位论文
面向Agent的开放巨型智能系统设计中的若干问题
操龙兵
学位类型工学博士
导师戴汝为
2003
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
关键词开放巨型智能系统 综合集成研讨厅 人机结合从定性到定量的综合集成 面向agent的计算 Multi-agent社会 社会抽象 决策 Agent构件/mas组织框架 Open Giant Intelligent Systems Hall For Workshop Of Metasynthetic Engineering Metasynthesis From Qualitative To Quantitative
摘要1990年,著名科学家钱学森院士等提出开放的复杂巨系统及其方法论-人机 结合从定性到定量的综合集成体系,揭开了一个科学的新领域。十多年来,这一 思想得到人工智能、系统科学与管理科学界的普遍重视。本文以此为理论指导, 将综合集成研讨厅视为开放巨型智能系统的一个实例,采用交叉学科的知识与多 学科交叉的方法,经验性地探讨了“面向Agent的开放巨型智能系统设计”中的 一些重要理论与实践问题。 从考察人工智能与智能系统的演化历程与特点出发,讨论了综合集成所涉及 的关键环节、综合集成与决策之间的关系;提出综合集成过程是一个形成知识生 产与服务体系的过程;综合集成研讨厅的实质是一个人机协作的智能信息系统。 就决策而言,研讨厅是决策支持系统的高级形式。 以系统观点,讨论了开放巨型智能系统的系统层次结构与复杂性;在对主流 计算模式进行对比与可行性论证的基础上,认为Agent是目前处理这种问题的最 合适的方法;在考察主流Agent理论的局限性之后,提出了Agent新的理论基础 及社会抽象机制,讨论了多Agent社会以及MAS组织与环境之间的关系,认为社 会性交互是MAS社会的统一机制。 提出了基于Internet的研讨厅等设计思想;讨论了面向Agent的开放巨型智 能系统的设计方法、MAS系统的组织结构与组织关系、系统体系结构,以及基于 Agent的分布式计算环境;提出基于Agent构件/MAS组织框架的计算模式,将开 放巨型智能系统建成一个人工Agent社会生态系统;并结合原理性实验,讨论了 面向Agent的研讨厅的Agent类型、体系架构、计算模型、工作机制、分布式资 源集成等问题。 尽管目前的经验性探讨还只是全部工作的开始;但是,作者认为有可能构建 基于Agent的综合集成研讨厅,并建立“面向Agent的开放巨型智能系统”的学科体系。
其他摘要In 1990, Chinese famous scientist Xuesen Qian proposed a new science, which is named Open Complex Giant Systems and its methodology - metasynthesis from qualitative to quantitative. This theory has caught widespread attention in areas of artificial intelligence, system sciences, and management science during the recent ten years. In this paper, we take the engineering technology of this theory, a hall for workshop of metasynthetic engineering(HWME), as an instance of open giant intelligent systems(OGIS), and empirically discuss several important theoretical and practical problems, which are closely related to the design of agent-oriented OGISs, in the light of interdisciplinary knowledge and methods. On the basis of a review of the evolutionary processes and their characteristics of artificial intelligence and intelligent systems, we discuss some key problems related to metasynthesis, and the relationship between metasynthesis and decision-making. We argue that the process of metasynthesis from qualitative to quantitative is a system of knowledge production, management and services. The essence of an HWME is a human-computer-cooperated intelligent information system. In terms of decision-making, an HWME can be viewed as the most advanced form of a decision support system. Then, we discuss the hierarchical model and the complexity of an open giant intelligent system from the viewpoint of systematism. We argue that the agent-oriented paradigm is the most suitable one for building an OGIS based on the comparison of mainstream computing models and the reasoning of feasibility. Furthermore, we propose a new theory foundation and a societal abstraction for agent-based computing for coping with the limitations of the mainstream agent-oriented abstraction. We discuss the conception of multi-agent society and the relationship between multi-agent organization and its environment in term of the new agent theory. We believe societal interactions among MAS organization and among MAS organization and its environment is a unified mechanism in multi-agent society. With the construction of an HWME, we propose some new ideas, for instance an Internet-based HWME; we further discuss design methods of an agent-oriented OGIS, and the organization structures and organization relationships, software architectures of multi-agent society. Finally, we propose an agent-oriented computing model based on the idea of agent component patterns and multi-agent organizational schemes, according to this model we view an agent-oriented OGIS as an artificial agent ecosystem. We present a prototype of agent-oriented HWME showing how to design the types of agents, software framework and their working mechanisms, and agent- based distributed resource integration based on the proposed computing paradigm. Although our empirical research work is to some degree a small part of the completely ideal solution, we believe it is possible to build a pra
馆藏号XWLW660
其他标识符660
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5740
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
操龙兵. 面向Agent的开放巨型智能系统设计中的若干问题[D]. 中国科学院自动化研究所. 中国科学院研究生院,2003.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[操龙兵]的文章
百度学术
百度学术中相似的文章
[操龙兵]的文章
必应学术
必应学术中相似的文章
[操龙兵]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。