【作者】 吴梅红；

【导师】 周昌乐；

【作者基本信息】 厦门大学 ， 人工智能基础， 2009， 博士

【摘要】 智能主体技术是人工智能领域的研究热点,智能主体系统与知识表示和知识推理有着密切的联系。在智能主体技术的研究中,主体的设计必须建立在知识表示和知识推理的坚实理论基础之上。要实现机器智能模拟人类智能,首先要实现智能系统准确清晰的知识表示以及符合人类日常思维的合理推理,从而刻画人类的认知过程,促使机器智能更好地进入知识处理和智能模拟阶段。本文在已有的知识表示和知识推理理论等相关研究工作的基础上,深入分析原有工作所存在的问题,从主体间知识的表达能力和互动能力方面探讨建立一个以动态认知逻辑理论为核心的多主体知识表示和知识推理系统,该推理系统不仅能对主体间的知识分布进行静态描述和认知推理,而且能对多主体交流过程中的知识变化进行动态刻画并推理。论文的主要工作和贡献如下:(1)对智能主体知识推理理论,包括认知逻辑理论,缺省逻辑理论,模态逻辑理论,因果推理理论进行研究分析,进一步研究智能主体的知识表示及知识更新,初步刻画人类的认知特性。(2)深入研究多主体认知逻辑理论和动态认知逻辑理论,并在动作逻辑模型的基础上扩展认知操作,提出一个新的行动扩展认知逻辑模型。(3)结合动态认知逻辑理论,提出条件约束的认知缺省理论,通过引入动作描述定义,使主体能够对自身的行为和运行机制进行描述和推理,并证明了相关性质和定理:(4)基于动态认知逻辑理论对主体的信念知识集进行刻画,并对多主体系统动态更新过程中所获得的知识进行分类,结合主体认知推理活动的非单调性和缺省性,运用条件约束认知缺省理论对多主体系统进行建模,探讨多主体信念知识集的互动性,系统地诠释了多主体认知推理过程中的知识动态更新。(5)提出了基于动态认知逻辑框架的因果推理理论,分析并证明了相关定理和性质;给出了该理论框架上的行动推理理论,并解决了基于逻辑框架的因果理论状态域刻画不一致的问题,从而进一步刻画了主体认知过程的因果关联分析。鉴于上述的工作,本文通过常识推理刻画主体的信念知识及其变化,从逻辑角度出发研究多主体系统的知识表示及更新,从而刻画人类的认知过程,研究人类的思维活动及人类思维中最能体现其能动性特征的各种不确定性推理,由此发展出更强可应用性的智能理论,具有十分重要的意义。更多还原

【Abstract】 Agent technology plays an important role in the research area of artificial intelligence, and the representation of knowledge has always been a major concern in the Agent System. In the research on agent technology, the design of agent system must be based on the solid theory of knowledge representation and knowledge reasoning. To achieve the goal of machine intelligence perfectly simulates human intelligence, intelligent systems must first realize clearly and precisely knowledge representation and reasonable commonsense inference, which could depict human cognitive processes, and promote machine intelligence access to the ideal knowledge processing and intelligent simulation more quickly.On the basis of knowledge representation and reasoning theories and the analysis of the existing problems about the related research work, we establish a knowledge representation and knowledge inference system based on dynamic epistemic logic theory, which not only describe static knowledge and cognitive reasoning, but also portrait the epistemic process of knowledge change and knowledge reasoning due to the communication of the multi-agents.The main contributions of this dissertation are as follows:(1) Give a brief introduction to the main theory of knowledge reasoning of agent technology, including the epistemic logic theory, default logic theory, modal logic theory, and causation reasoning theory, which provide an initial characterization of knowledge representation in human cognitive process.(2) On the basis of the multi-agent epistemic logic theory and dynamic epistemic logic theory, a new action-extended epistemic logical model is proposed, which extend the basic cognitive operation of epistemic model of AML. (3) Combining with dynamic epistemic theory, a constrained epistemic default theory is proposed, and some related theorems are proved.(4) On the basis of dynamic epistemic theory, we use constrained epistemic default logic to depict the interaction of the knowledge change in the multi-agent system, which elucidates the dynamic process of knowledge update.(5) Based on dynamic epistemic logic theory, a causation reasoning model is proposed and the action reasoning theory in this causation framework is also introduced, further the characterization of causation reasoning in the agent’s cognitive process is analyzed.This dissertation depicts the knowledge sets and belief changes of the agents according to common sense reasoning, and portrays the dynamic process of knowledge update in the multi-agent system via a logical way, leading to a comprehensive study of human thinking and the dynamic characteristics of its uncertainty reasoning, which is promising and meaningful.更多还原