期刊文献+

基于A2C的匝道相联交叉口信号控制优化

Optimization of signal control at ramp adjacent intersections based on A2C
下载PDF
导出
摘要 城市快速路的交通运行效率对于整个城市的顺畅通行至关重要,在早晚高峰期间,受限于相连接辅路的交通承载能力,快速路上较大的交通流量无法顺利从出口匝道驶入目标路段,在匝道上形成排队现象,严重时会导致匝道回溢,使快速路上车道由于被占用而产生交通瓶颈,造成较大的交通出行损失.利用深度强化学习算法进行出口匝道相关联的道路交叉口信号控制优化,将信号灯设为智能体,通过设置检测器,将快速路出口匝道及交叉口的交通运行情况作为智能体获取的状态信息,引入以辅路与出口匝道剩余通行能力之比为动态修正参数的奖励函数,在保证匝道交通运行效率下,完成交叉口信号优化过程.以中国北京市东三环快速路及某关联交叉口为例,借助交通仿真平台SUMO(simulation of urban mobility)及Traci库搭建仿真环境进行实验.结果表明,基于改进A2C(advantage actor critic)算法的信号控制方法在控制效果上优于传统信号控制以及基于深度Q网络(deep Q-network,DQN)算法的信号控制方法,在出行高峰期间能够有效降低匝道回溢的发生概率,有效改善辅道相联交叉口的通行效率. The operational efficiency of urban expressways significantly impacts citywide transportation flow.During morning and evening rush hours,the limited capacity of feeder roads to handle high traffic volumes leads to congestion on expressway exit ramps.This often results in queuing and,in severe cases,ramp spillback,causing traffic bottlenecks on the expressway lanes and substantial losses in traffic travel.This study utilizes deep reinforcement learning algorithms for traffic signal control optimization at exit ramps associated with intersection crossings.Traffic signals are treated as intelligent agents,and real-time traffic conditions of the expressway ramps and intersections are fed into the system using detectors.A dynamic reward function is introduced,adjusted based on the ratio of the remaining traffic capacity between the feeder roads and the exit ramps.The objective is to enhance ramp traffic efficiency while optimizing intersection signals.The methodology is applied to an expressway on East Third Ring Road,Beijing,and a related intersection,which utilizing the simulation of urban mobility(SUMO)traffic simulation platform and the Traci library to create a simulated environment.The results indicate that the signal control method,based on an improved advantage actor critic(A2C)algorithm,outperforms traditional signal controls and those based on the deep Q-network(DQN)algorithm.Especially during peak travel times,it effectively reduces the probability of ramp spillback and enhances the traffic efficiency of the interconnected feeder road intersections.
作者 宋太龙 郭明洋 陈一凡 贺玉龙 SONG Tailong;GUO Mingyang;CHEN Yifan;HE Yulong(Shandong Transportation Research Institute,Jinan 250031,Shandong Province,P.R.China;Beijing Municipal Professional Design Institute Co.Ltd.,Beijing 100097,P.R.China;Faculty of Architecture,Civil and Transportation Engineering,Beijing University of Technology,Beijing 100124,P.R.China)
出处 《深圳大学学报(理工版)》 CAS CSCD 北大核心 2024年第2期192-201,共10页 Journal of Shenzhen University(Science and Engineering)
基金 国家重点研发计划资助项目(2017YFC0803903)。
关键词 智能交通 匝道 交叉口 信号控制优化 深度学习 强化学习 intelligent transportation ramp intersection signal control optimization deep learning reinforcement learning
  • 相关文献

参考文献6

二级参考文献25

共引文献53

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部