| 题名 | 用复杂性科学来理解与设计智能系统 |
| 作者 | 周登勇 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2000-06-01 |
| 授予单位 | 中国科学院自动化研究所 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 戴汝为 |
| 关键词 | 复杂性 计算机仿真 群体行为 动物行为学 软件代理人 人工生命 人工智能 complexity computer simulation collective behavior ethology software agent artificial life artificial intelligence |
| 学位专业 | 控制理论与控制工程 |
| 中文摘要 | 本世纪80年代，一场跨学科、交叉学科融汇的科学革命正在兴起，孕育着2l世纪一 门新科学的诞生。这门科学不依赖牛顿式的宇宙观——隐喻世界如钟表的一般规律而可预测：它所要探讨的是复杂系统中，各组成部分之间相互作用所突现出的特性。这就是所谓的复杂性科学(the science of complexity)。 本文的研究工作以复杂性科学为根本出发点，研究了多Agent之间的协作，自治的适应性Agent的构造，以及能在网络上漫游的Mobile Agent系统。 首先，论述当前复杂性科学的主要的三种观点，即I．Prigogine的非平衡系统的自组织，SFI的Holland提出的复杂适应性系统理论以及国内的科学家提山的开放的复杂巨系统理论。在对这些观点分析的基础上，设计了基于复杂性观点的计算机仿真。文中详细论述了采用跨平台的Java语言自行设计的基于Muti-Agent的计算机仿真系统的设计动机、基本结构、部分代码实现以及进一步的工作，并与相关的工作，特别是SFI的SWARM系统进行了的对比。 接着，在白行设计的平台上所进行了些列的动物群体行为的仿真。这些仿真既反映了仿真平台的功能，又生动地揭示了复杂性科学中的一些基本思想。同时，这些仿真事例无论是对我们如何从复杂性科学的角度把握动物集体行为的本质，还是对我们从复杂性科学的角度在工程上设计多Agent(如多机器人)协作系统，都有巨人的理论与实践上的意义。 最后，通过这些仿真事例，作者还总结了设计复杂性仿真应用的非常有用的基本原则。 再接着，采用复杂性科学的观点统一阐述了人工智能中的“旁门左道”，深刻地刻画了这些理论的共同特色，从而使这些理论的基本观念变得非常的清晰，同时也总结出设计智能系统的一套基本原则。这些原则无论是对理解人的智能的本质，还是设计一个工程上的智能系统都有重要的意义。而且，按照这些原则设计了一种白治Agent的控制结构，并在一个仿真环境中检验了该结构的合理性。 最后，将自治Agent与Muti-Agent协作两方面的技术应用于当前软什Agent中的一个新的新星——Mobile Agent技术。本文论述了我们自行设计的Mobile Agent系统的基本结构与实现以及Mobile Agent在我们研究项目中的应用。我们的Mobile Agellt系统强调的是该技术导致的软件设计上的柔性。 本文的工作的创新性在于从复杂性科学的角度并结合实际的问题出发探索智能现象以及智能系统的设计，同时给予当前的一些人工智能、方法给予精确的理论定位。 |
| 英文摘要 | A revolution of interdiscipline in science started in I980s this century gives birth to a new science, which does not rely on a Newtonism view of the world, treating this world like a clock both precise and predictable. What it studies is the characteristics that the composing parts in a complex system interact with each other. This is what is called, the science of complexity. My researches are muti-agent coordination, the architecture of autonomous adaptive agent and mobile agent roaming on Internet. The science of complexity is the basis of my researches. First, three main theory on the science of complexity are discussed, which are I. Prigogine's Self-organization of Non-balance System, Holland's Complex Adaptive System(CAS) and Open Complex Giant System(OCGS) founded by the scientists in China. A computer simulation system is developed for complexity, which is implemented by Java language. The motivation of design, architecture, and some codes are discussed. Related works, especially SWAARM, are compared with this simulation system. Then, some simulations are studies based on this simulation platform, which proof that our simulation platform is useful and display the basic ideal in the science of complexity. At the same time, these simulation examples are valuable for us whether to understand the essence of animal collective behaviors or to design multiple robots coordination system in engineering. Some experiences are concluded. Then, the non-main theories in Artificial Intelligence are studies from the ideal of the science of complexity, and the common places of these theories are discovered. My researches make them more clearly than before. A set of principles is supposed, which is useful whether to understand the essence of mind or to design an intelligent system. Other more. A kind of control architecture for autonomous adaptive agents is suggested based on these principles, which is tested in a simulating environment. Last. we apply the technology about autonomous agents and muti-agent to mobile agent. A mobile agent system has been developed in our lab. The design and implement of our mobile agent system are discussed. We think that the most important character of mobile agent ~s flexible for software development. The innovative character of this paper is shown in the way that it studies the artificial intelligence and the design of an artificial intelligence system bv adopting both the science of complexity and some practical problems. In the meantime it gives precise theoretical explanations for some present artificial intelligence theories. This work is not only meaningful to artificial intelligence, cognitive science, and artificial life. but also meaningful to the science of complexity itself. |
| 语种 | 中文 |
| 其他标识符 | 546 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5711]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 周登勇. 用复杂性科学来理解与设计智能系统[D]. 中国科学院自动化研究所. 中国科学院自动化研究所. 2000. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
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。
修改评论