基于Multi-Agent的城市交通信号控制研究

发布时间：2018-03-20 03:37

本文选题：人工智能　切入点：多Agent系统　出处：《长沙理工大学》2008年硕士论文　论文类型：学位论文

【摘要】： 随着国民经济的不断增长,人民生活水平日益提高,汽车保有量也随之增长,随之而来的城市交通问题则不断突显出来。城市交通系统是一个具有随机性、不确定性、实时变化的复杂大系统,采用以往的交通控制方式,已经无法解决日益严峻的交通问题,采用现代科技手段及智能方法来解决城市交通问题成为当前研究的热点。随着分布式人工智能技术的发展,Agent技术和多Agent系统理论成为研究分布式计算环境下软件智能化的基础,由于城市交通控制固有的分布性,采用多Agent技术研究城市交通信号控制问题具有很好的前景。本文结合Agent技术及多Agent系统理论,提出了以单个路口交通信号控制Agent为基础的多Agent分布式协调控制系统。首先,对系统中单路口的交通信号控制Agent进行了结构设计,对其工作过程进行了描述,并对其学习单元的学习模式进行了设计,在此基础上,对基于多Agent的交通信号控制系统的结构进行了设计,并对系统进行了形式化描述。接着,对单个路口的交通状态加以选择,采用模糊聚类方法对车辆到达及信号显示状态进行了定量描述,建立了信号控制规则集,以总停车延误为控制目标,采用改进的Q-学习算法对Agent进行训练,以改进信号控制规则,通过仿真,对文中提出的单路口控制方法进行了仿真。仿真结果表明,该方法优于传统的定时控制和感应控制方式。最后,着重对基于多Agent的分布式信号控制系统中,各信号控制Agent间的信息交互,协调方式,及系统的学习方式进行了描述,并对系统协调的实现进行了仿真,仿真结果表明,文中方法能明显的减少车辆的总延误时间。
[Abstract]:With the continuous growth of the national economy, people's living standards are improving day by day, the number of cars is also increasing, and the following urban traffic problems are constantly highlighted. The urban transportation system is a kind of random and uncertain. The complex large-scale system with real time change has been unable to solve the increasingly serious traffic problems by using the former traffic control mode. It has become a hot spot to solve the urban traffic problems by modern scientific and technological means and intelligent methods. With the development of distributed artificial intelligence technology, agent technology and multi-#en0# system theory become the basis of studying software intelligence in distributed computing environment, because of the inherent distribution of urban traffic control. It is very promising to study the problem of urban traffic signal control by using multi-Agent technology. Based on the Agent technology and the theory of multiple Agent system, this paper presents a multi-#en3# distributed coordinated control system based on a single intersection traffic signal control Agent. Firstly, the traffic signal control Agent of a single intersection in the system is designed. The working process of the system is described, and the learning mode of the learning unit is designed. On this basis, the structure of the traffic signal control system based on multiple Agent is designed, and the system is formalized. The traffic state of a single intersection is selected, the vehicle arrival and signal display states are described quantitatively by fuzzy clustering method, and the signal control rule set is established, with the total parking delay as the control target. The improved Q- learning algorithm is used to train the Agent to improve the signal control rules. The simulation results show that the proposed single intersection control method is simulated. This method is superior to the traditional methods of timing control and induction control. Finally, the information exchange, coordination mode and learning mode of each signal control Agent in the distributed signal control system based on multiple Agent are described. The simulation results show that the method can obviously reduce the total delay time of the vehicle.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：U491.51

【参考文献】