节点文献
智能混杂系统建模、分析理论及应用的研究
Research on Theory and Application for Modeling and Analysis of Intelligent Hybrid System
【作者】 叶阳东;
【导师】 贾利民;
【作者基本信息】 铁道部科学研究院 , 交通信息工程及控制, 2002, 博士
【摘要】 本文以铁路系统中的“铁路智能运输系统技术研究”为背景,在混杂系统建模、分析理论的基础上,全面分析了基于智能的混杂系统研究所面临的问题,完整地给出了智能混杂系统的定义,并归纳出其相应的特性。论文主要是围绕混杂系统中知识的处理、不确定性问题的处理、系统的自治能力等方面,基于Petri网的理论和方法进行了一类智能混杂系统的建模、分析理论及应用的研究,工作本身是具有开创性质的,也是基础性的。 论文在全面归纳Petri网的基本概念、特性、分析方法的基础上,从理论上详细地研究了带有时间约束的Petri网,形成了一系列与时间约束变迁/库所有关的结论和定理,扩充了TCPN’s现有的理论。面向智能混杂系统的建模,论文首次定义了一种分层混合对象Petri网(HHPNO),定义中充分体现出O-O方法和Petri网的一致性和优势互补性,网中综合考虑了系统的背景知识表述、系统不同属性的处理、多种对象Petri网的集成等问题;基于系统的分层特性,HHPNO在高层采用了带有对象的Petri网(PNO);面向底层混杂系统的描述采用对象Petri子网。论文对HHPNO中的PNO、规范化接口、库所子网、变迁子网、层次化求精等问题的相关理论进行了系统的研究。 文章在详细研究了现行混合Petri网的建模和分析理论的基础上,面向智能混杂系统定义了一种智能条件/事件混合Petri网,该网包含赋时离散Petri网、连续Petri网和智能决策接口,离散网和连续网之间是通过条件和事件信号进行交互,智能决策接口可构成和其它智能系统或智能决策对象的互连,使得一类混杂系统具有智能决策的能力,同时又保持系统的相对独立性。建模实例表明所定义的混合Petri网具行模块性优良和接口的一致性等特征。论文针对线性混合Petri网,用两种方法进行状态可达性分析:谓词转换方法和基于路径的方法。本文的分析方法相对于已有的方法来说,引入了外部决策因素,该研究既丰富了现有混杂系统的分析理论。 在HHPNO的建模体系下,针对列车群行为构建了两种实验模犁TOPNO和TGOSOPS,实验表明模型便于分布式系统的实施并具有很好的开放性。论文构建了一种基于多种对象Petri网的列车群运行仿真系统及分布式仿真系统的基本结构,并设计实施了基本的实验模型,实验模型在设计方面立足于通用性好、扩充性优良、操作便利等特点。该模型可构成研究一类智能混杂系统建模及分析理论的实验平台。针对智能混杂系统的知识处理问题,在已经建立的列车群运行系统的对象Petri网模犁的基础上,本文利用Petri网对知识的处理能力,给出了一种有效的时间Petri网的时间知识推理算法,由于算法中引入模糊时间区间的操作、保留了最少数量的生成图节点,所以该算法比已有的算法效率高并可处理时间区间的不确定性问题。首次将该算法用于铁路智能运输系统中时间类知识的表示和推理,理论和实验表明:本算法可验证时间约束的列车运行调整方案的可行性;能验证铁路专家系统中时间知识的一致性等问题。
【Abstract】 With the background of research on Railway Intelligent Transportation System(RITS) technology in railways and on the basis of theories for modeling and analysis of Hybrid System(HS) the dissertation comprehensively analyzes the issues on hybrid system oriented to intelligence, presents the definition of Intelligent Hybrid System(IHS) and concludes its properties. In other words, based on knowledge processing, uncertainty processing, system autonomy and so on in HS, the theories and schemes that relate to Petri nets are employed for the establishment of the theoretical and applied methodology for the modeling & analysis of IHS in this dissertation.After totally inducing the basic concepts, properties and the methods of analysis in Petri nets, and minutely studying relative theories about Timing Constraint Petri Nets(TCPN) , the dissertation presents several conclusions and theorems about place/transitions with time constraint and extends existing theories about TCPNHS. In modeling of IHS, the formal definition of Hierarchical Hybrid Petri Nets with Objects(HHPNO) is inaugurated, in which Object-Oriented approaches and Petri nets are indicated to be consistent and complementary in system modeling, and the issues about domain knowledge presentation, different kinds of attributes processing, the cooperation of varied kinds of Object-Oriented Petri nets etc. are dealt with synthetically. As concerning the hierarchical property of IHS modeling, Petri Nets with Objects(PNO) is used to analyze high-level problems in HHPNO, but on the processing of low-level issues Object Petri Subnets(OPS) is utilized. The theories about PNO, the standardization of interface, the place subsets, the transition subsets and hierarchical refining, etc, concerned with HHPNO are discussed.On the basis of the theories for the modeling and analysis of hybrid Petri Nets, the dissertation gives out the formal definition of a kind of Intelligent Condition/Event Hybrid Petri Net(ICEHPN) oriented to IHS, which includes the Discrete Timed Petri Net(DTPN) , the Continuous Petri Net(CPN) and the Intelligent Decision Interface(IDI). DTPN and CPN act on each other through condition and event signals. The IDI is used to interconnect with other Intelligent Systems or Intelligent Decision Objects, makes HSs have intelligent decision ability and keeps the independent property of HS. The modeling example indicates that the ICEHPN has superior properties for modularization and the consistency of interface, over existing schemes. Aiming at linear hybrid Petri nets two methods are adopted to analyze state reachability of hybrid system, which are predicate-transformation method and path-based method. The outside enable predicates for discrete transitions are utilized to introduce external decision factors of the system in those methods compared with existing schemes, so dissertation extends existing analysis methods about HSIncluded in the modeling methodology of HHPNO, two experimental models of train group operation, Train Operation Petri Net with Objects (TOPNO) and Train Group Operation System with Object-orientedPetri Subnet(TGOSOPS), are constructed. The simulation experiments indicate that the models feature with convenient realization for distributed systems and good properties for interconnection. The basic structure of simulation system and distributed simulation system about train group operation are built according to the varied kinds of object-oriented Petri nets, and the fundamental experiment model featuring universality, expansibility, manipulability, etc, is implemented. The model can be used to construct the platform for studying the modeling and analysis theories for one kind of IHS. Oriented to the knowledge processing in IHS, in order to design a time Petri nets efficient algorithm for temporal knowledge reasoning the dissertation makes full use of the efficiency of Petri nets for knowledge processing and the model of train group operation. Because fuzzy time interval operator is utilized and the least sprouting graph nodes are ad
【Key words】 artificial intelligence; hybrid system; Petri nets; continuous variable dynamic system; discrete event dynamic system; railway intelligent transportation system; object-oriented; agent; simulation; uncertainty; time interval;