基于Agent的自适应交通信号协同控制方法研究

发布时间：2018-02-27 15:02

本文关键词： 多智能体(Multi-Agent) 车路协同(CVIS) 自适应控制多智能体遗传算法　出处：《厦门理工学院》2015年硕士论文　论文类型：学位论文

【摘要】：目前对城市道路交叉口信号灯的优化控制的研究集中在对单个交叉口的信号配时优化,虽然仅研究单点交叉口的优化控制降低了信号配时的复杂性,但过于简化问题使研究成果不能很好地应用于实际。本文提出了基于Agent的自适应交通信号协同控制方法,该方法通过优化道路交叉口信号控制和改善车辆驾驶行为,实现区域交通系统基于Agent的协同运行。首先,本文设计了基于Multi-Agent的车路协同系统模型(Cooperative Vehicles-Infrastructure System,CVIS),为自适应交通信号协同控制提供实时可靠的交通流数据。CVIS架构中每台车和各重要路口的信号灯均为多智能体,包含路侧终端、车载终端和行人探测器,通过无线通信网络实现人-车-路的信号互联,全方位感知周围环境信息。其次,在CVIS基础上设计了一种基于Agent的自适应协同交通信号控制模型ACTAM(Adaptive and Cooperative Traffic Light Agent Model,ACTAM),该模型利用智能体的自治性和自组织性提高对道路交叉口控制的自适应性,并依据从CVIS获得的实时交通流数据信息通过增强学习实时优化交通控制策略。再次,采用多智能体遗传算法(Multi-Agent genetic algorithm,MAGA)对控制目标进行优化,这种全局优化协调控制的策略使各个路口Agent在未知其他路口信号控制方案时,彼此协调合作并实现对区域交通的自适应协同控制。最后,结合厦门市交警大队提供的数据,使用交通仿真软件Trans Modeler对厦门市政府区域和厦门市厦禾路BRT沿线进行仿真,结果表明本文方法能够有效降低交叉口的平均延误时间和停车次数,可以提高路口控制单元的适应能力并有效缓解交通拥堵的现状。
[Abstract]:At present, the research on the optimal control of signal lights at urban road intersections is focused on the signal timing optimization of single intersection, although the complexity of signal timing is reduced by only studying the optimal control of single intersection. However, the problem of oversimplification can not be applied to practice well. In this paper, an adaptive traffic signal cooperative control method based on Agent is proposed, which can optimize the traffic signal control and improve the vehicle driving behavior. To realize the cooperative operation of regional transportation system based on Agent. First of all, In this paper, a cooperative Vehicles-Infrastructure system CVIS system model based on Multi-Agent is designed, which provides real-time and reliable traffic flow data for adaptive traffic signal collaborative control. In the framework of CVIS, each vehicle and every important intersection signal light is multi-agent, including roadside terminal. Vehicle terminal and pedestrian detector, through wireless communication network to achieve human-vehicle-road signal interconnection, omni-directional perception of the surrounding environment information. Secondly, Based on CVIS, an adaptive collaborative traffic signal control model based on Agent, ACTAM(Adaptive and Cooperative Traffic Light Agent Model, is designed. The model makes use of the autonomy and self-organization of agents to improve the adaptability of road intersection control. According to the real-time traffic flow information obtained from CVIS, the real-time traffic control strategy is optimized by reinforcement learning. Thirdly, multi-agent genetic algorithm is used to optimize the control target. This global optimal coordinated control strategy enables Agent to coordinate and cooperate with each other in unknown signal control schemes of other intersections and to realize adaptive cooperative control of regional traffic. Finally, combined with the data provided by Xiamen Traffic Police Brigade, The traffic simulation software Trans Modeler is used to simulate the area of Xiamen Municipal Government and the BRT line of Xiamen Xiahe Road. The results show that this method can effectively reduce the average delay time and the number of stops at the intersection. It can improve the adaptability of the intersection control unit and effectively alleviate the current situation of traffic congestion.
【学位授予单位】：厦门理工学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.54;TP273

【参考文献】