考虑车道剩余容量的区域交通信号控制方法  

Regional Traffic Signal Control Methods Considering Lane Remaining Capacity

作　　者：代亮[1] 黄自彬 张中昊 李臣富 DAI Liang;HUANG Zibin;ZHANG Zhonghao;LI Chenfu(School of Electronics and Control Engineering,Chang'an University,Xi'an 710064,China;Shaanxi Fast Gear Co Ltd,Xi'an 710119,China)

机构地区：[1]长安大学,电子与控制工程学院,西安710064 [2]陕西法士特齿轮有限责任公司,西安710119

出　　处：《交通运输系统工程与信息》2025年第2期108-118,共11页Journal of Transportation Systems Engineering and Information Technology

基　　金：陕西省交通运输厅交通科研项目;长安大学中央高校基本科研业务费专项资金。

摘　　要：平面交叉口是城市路网整体通行能力的瓶颈,是城市路网交通组织、交通渠化和交通治理的重点。深度强化学习通过智能体与环境交互寻找目标策略,契合交通环境复杂多变的特点,被广泛应用于平面交叉口交通信号控制领域。本文提出考虑车道容量的区域交通信号协同控制方法,通过建模上下游交叉口协作关系,在最大压力方法中引入交叉口下游车道容量信息设计奖励函数,同时,基于多智能体强化学习算法提出分布式区域交通信号协调控制方法。通过使用济南与杭州真实路网和交通流数据集进行性能验证,与现有区域交通信号控制方法相比,平均行程时间降低6.05%,平均延误降低18.39%,平均排队长度降低21.86%,吞吐量提升0.24%。Intersection is the bottleneck of the overall traffic capacity of urban road networks,are the focal points of traffic organization,channelization,and management within the networks.Deep reinforcement learning is widely used in the field of traffic signal control at intersections,as it interacts with the environment to find target strategies,which aligns well with the complex and dynamic characteristics of traffic environments.This paper proposes a regional traffic signal coordination control method that considers lane capacity.By modeling the cooperation relationship between upstream and downstream intersections and introducing downstream lane capacity information into the maximum pressure method to design the reward function,a distributed regional traffic signal coordination control method is proposed based on multi-agent reinforcement learning algorithm.Performance verification is carried out using real road networks and traffic flow datasets from Jinan and Hangzhou.Compared with existing regional traffic signal control methods,the proposed method reduces the average travel time by 6.05%,the average delay time by 18.39%,the average queue length by 21.86%,and increases the throughput by 0.24%.

关 键 词：智能交通 交通信号控制 深度强化学习 多智能体 特征融合 

分 类 号：U491[交通运输工程—交通运输规划与管理]