基于自动机学习的黑盒软件系统的监督控制研究

发布时间：2021-08-05 14:52

　　复杂软件系统的开发中,为节省开发成本,开发者常常会复用一些已有的系统组件或者第三方提供的构件。这些组件大都是黑盒系统。对于使用者来说,他们通常并不知道这些被复用组件的逻辑模型和源代码。在软件工程实践中,很多系统需求描述均使用自然语言或半形式化语言。开发者想得到这些需求的正确形式化模型十分困难。本论文考虑系统的形式化模型未知或者系统需求的形式化模型未知或者这两种模型均未知等情况,将离散事件系统的监督控制理论和自动机学习算法结合起来,为系统设计控制器,并用它对实际系统实施在线监督控制,一定程度上使被复用的组件在不更改源代码的情况下满足系统需求。本论文主要贡献如下。当系统需求的形式化模型未知但受控对象的形式化模型易于获得时,本文通过对经典的L^*学习算法进行扩展,结合离散事件系统的监督控制理论,提出了一种可为受控系统生成非阻塞的且具有最大可允许行为的控制器的方法。该方法分为两个步骤。首先,利用学习算法学习得到一个临时正确的控制器。其次,若学习到的控制器是非阻塞的,那么该控制器就是我们想要得到的具有最大可允许行为的非阻塞控制器。否则,该方法将这个阻塞控制器看作一个新的受控系... 

【文章来源】：西安电子科技大学陕西省 211工程院校 教育部直属院校

【文章页数】：124 页

【学位级别】：博士

【文章目录】：
摘要
ABSTRACT
List of Symbols
List of Abbreviations
Chapter1 Background and Motivation
    1.1 Supervisory Control Theory
    1.2 Automaton Learning Algorithm
    1.3 Integration of SCT and Learning Algorithms
    1.4 Organization and Contributions
        1.4.1 Organization
        1.4.2 Contributions
Chapter2 Preliminaries
    2.1 Automaton.
    2.2 Supervisory Control Theory
    2.3 Learning-based Testing and LBTest
    2.4 L*Learning Algorithm
        2.4.1 Brief Description about the L*Learning Algorithm
        2.4.2 A Small Example.
Chapter3 Supervisory Control without Formal Models of Requirements
    3.1 Supervisor Synthesis Approach
        3.1.1 The Proposed Learning Approach.
        3.1.2 Computing Supervisors Using SCT
        3.1.3 Simplifications of Requirement Membership Queries
        3.1.4 Illustrative Examples
    3.2 Experimental Studies
        3.2.1 A Manufacturing System Using AGVs
        3.2.2 Small Factory
    3.3 Conclusions and Discussion
Chapter4 Supervisory Control without Formal Models of Systems
    4.1 The Integration of LBT and SCT
        4.1.1 Integration Framework
        4.1.2 Example:A Simplified Cruise Controller
    4.2 Experimental Studies
        4.2.1 Testing Performing
        4.2.2 Supervisory Control of the System
        4.2.3 Confirmation of the Experiment Results by Simulation for the BBW
        4.2.4 Scalability Study.
    4.3 Conclusions and Discussion
Chapter5 Supervisory Control without Formal Models of Systems and Requirements
    5.1 Learning and Control Approach
        5.1.1 Testing the System.
        5.1.2 System Abstraction
        5.1.3 Learning Moore Automaton
        5.1.4 Computing Supervisors
        5.1.5 Supervisory Control of the System
    5.2 Experimental Studies
        5.2.1 BBW System
        5.2.2 Adaptive Cruise Control in A Platooning Program
    5.3 Conclusions
Chapter6 Conclusions and Future Works
    6.1 Conclusions
    6.2 Future Works
Reference
Acknowledgement
Biography



本文编号：3323949