【作者】 郭栋栋；

【导师】 王喜富；

【作者基本信息】 北京交通大学 ， 系统工程， 2007， 硕士

【摘要】 近年来,智能公交调度的发展对传统的手工车辆调度提出了强有力的挑战。实时公交调度专家系统是人工智能(AI)在公交领域的实际应用成果,是公交智能调度系统的重要组成部分。它将管理信息系统、软件工程、系统工程、神经网络、遗传算法以及专家系统等理论与公交调度作业有机结合在一起,创造性的应用对于改善城市公共交通、提高公交作业管理水平以及公交客运智能化调度有着特殊重要的意义。在调研当前公交调度专家系统研究现状的基础上,本文对实时公交调度专家系统相关的基本概念、工作原理、组织机制、相关算法和实现的技术方法等较全面和系统的研究,继而设计了实时公交调度专家系统模型,最后对模型的可行性进行了实例验证。本文的主要研究内容如下:1.针对静态调度存在的灵活性差、无法适应动态多变的道路环境等弊端,阐述了研究动态调度的必要性。基于公交调度工作的发车原理,本文从多个角度分析影响实时调度的诸多因素,彼此之间相互影响结果,继而提出相应的解决办法和策略。2.针对公交调度领域知识的特点,在研究不同种类的知识模型和系统推理机制特点的基础上,设计了公交调度专家系统的知识表示模型---产生式知识表示与面向对象思想相结合的知识模型。同时提出公交调度专家系统推理模型,以及相关推理策略。3.根据公交调度的实际情况,划分了公交状态异常类。为了达到实际公交系统运行异常的快速诊断以及辅助专家系统建立知识库的需要,设计了一种聚类算法用以检测公交系统中的异常类别及建立新的异常类。4.另外在动态实时的生成调度方案方面,本文给出了优化调度的一个数学规划模型。并利用遗传算法可以快速的给出规划模型的近似优化解的优点,有效的避免了传统算法难以快速收敛的问题,解决了公交调度动态实时性的问题。更多还原

【Abstract】 Real-time bus dispatching expert System of is an application fruit of Artificial Intelligence and an indispensable part of Public Traffic Dispatching System. It has integrated with Information Management System, Software Engineering, System Engineering, Neutral Network, genetic algorithm Expert System theory, public traffic dispatching. This creative appliance is of high importance to better urban traffic conditions, improve the level of public traffic management and intelligent dispatching. On the basis of deep and widespread research of domestic and foreign document, we have made an all-rounded and systematic research on the basic concepts, theory, concepts, working principles and techniques of Expert System. The main contents are as follows:1. Considering the disadvantages of static dispatching like low flexibility and inability to adapt to changeable road environment, we analyze the necessity of the research of dynamic dispatching. Based on the principle of bus dispatching, the dissertation analyzes the effects of the factors on real-time and their interactive affections, then put forward the relating method and strategy.2. According to the characteristics of the knowledge of the bus dispatching area, we design the knowledge model of the Bus Dispatching Expert System—the model of the integration of production rule and object-oriented. Meanwhile we put forward the reasoning model of the system and relating reasoning strategy.3. Based on the practical conditions of real-time dispatching, we design abnormity classes. In order to quickly diagnose abnormities during real operation and meet requirements to set up a knowledge database for expert system, we design a clustering algorithm to check abnormity classes and create new abnormity classes.4. In terms of dynamical creating real-time dispatching schemes, we put forward a mathematics model to optimize dispatching. We also effectively avoid slow convergence of traditional algorithm by the way of genetic algorithm.更多还原