基于深度强化学习的交通信号优先控制

兰州理工大学学报 ›› 2026, Vol. 52 ›› Issue (3): 91-101.

• 自动化技术与计算机技术 • 上一篇下一篇

基于深度强化学习的交通信号优先控制

王志文^*1,2,3, 于宇凌¹, 杨康康¹, 王浩旭¹, 苗葳⁴

1.兰州理工大学自动化与电气工程学院, 甘肃兰州 730050;
2.兰州理工大学甘肃省工业过程先进控制重点实验室, 甘肃兰州 730050;
3.兰州理工大学电气与控制工程国家级实验教学示范中心, 甘肃兰州 730050;
4.甘肃紫光智能交通与控制技术有限公司, 甘肃兰州 730030

收稿日期:2024-02-28 出版日期:2026-06-28 发布日期:2026-06-30
通讯作者: 王志文(1976-),男,甘肃武威人,博士,教授,博导.Email:wzw@lut.edu.cn
基金资助:
国家自然科学基金(62263019),甘肃省青年科技基金计划(21JR7A346),甘肃省重大专项(21ZD4GA028)

Traffic signal priority control based on deep reinforcement learning

WANG Zhi-wen^1,2,3, YU Yu-ling¹, YANG Kang-kang¹, WANG Hao-xu¹, MIAO Wei⁴

1. School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China;
2. Key Laboratory of Gansu Advanced Control for Industrial Processes, Lanzhou University of Technology, Lanzhou 730050, China;
3. National Demonstration Center for Experimental Electrical and Control Engineering Education, Lanzhou University of Technology, Lanzhou 730050, China;
4. Gansu Ziguang Intelligent Transportation and Control Technology Co., Ltd., Lanzhou 730030, China

Received:2024-02-28 Online:2026-06-28 Published:2026-06-30

摘要/Abstract

摘要： 现有的交通信号优先控制系统无法有效适应日益复杂多变的城市交通环境,并且在实施交通信号优先控制时,容易对非优先车辆造成不必要的延误.针对这种不足,从控制策略和优化目标两个方面着手,基于交叉路口的实时交通信息,并结合深度强化学习高效处理高维连续数据和自学习的特点,提出一种基于深度强化学习的优先控制策略(G-D3QN),改进基于竞争架构的深度双Q网络对信号优先控制策略进行求解.同时,构建双目标优化奖励函数,兼顾公交车辆和社会车辆的通行需求,在最大化交叉路口通行能力的同时,降低公交车辆延误.最后,基于SUMO搭建仿真环境,将改进后的算法与D3QN、DDQN、Q-Learning三种基类算法进行对比,结果表明改进算法能够有效协同提高公交车和社会车辆的通行效率.

关键词: 交通信号优先, 交通信号控制, 深度强化学习, SUMO仿真平台

Abstract: The existing traffic signal priority control system cannot effectively adapt to the increasingly complex and changing urban traffic environment, and is prone to cause unnecessary delays to non-priority vehicles when implementing traffic signal priority control. To address this shortcoming, this study proposes improvements from both the control strategy and optimization objectives. Based on the real-time traffic information of intersections and combining the characteristics of deep reinforcement learning to efficiently process high-dimensional continuous data and self-learning, a priority control strategy based on deep reinforcement learning is proposed to improve the deep double-Q network based on a dueling architecture to solve the signal priority control strategy. At the same time, a dual-objective optimization incentive function is constructed to balance the access needs of both public transport vehicles and general traffic,aiming to reduce delays of public transport vehicles while maximizing the capacity of intersections. Finally, a simulation environment is built using the SUMO to compare the improved algorithm with the three base class algorithms of D3QN,DDQN and Q-Learning. The experimental results show that the improved algorithm can effectively improve the access efficiency of buses and social vehicles in a coordinated manner.

Key words: transit signal priority, traffic signal control, deep reinforcement learning, SUMO simulation platform

中图分类号:

王志文, 于宇凌, 杨康康, 王浩旭, 苗葳. 基于深度强化学习的交通信号优先控制[J]. 兰州理工大学学报, 2026, 52(3): 91-101.

WANG Zhi-wen, YU Yu-ling, YANG Kang-kang, WANG Hao-xu, MIAO Wei. Traffic signal priority control based on deep reinforcement learning[J]. Journal of Lanzhou University of Technology, 2026, 52(3): 91-101.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: https://journal.lut.edu.cn/CN/

https://journal.lut.edu.cn/CN/Y2026/V52/I3/91

参考文献

[1] QADRI S S S M,ALI GO··KCE M,O··NER E.State-of-art review of traffic signal control methods:challenges and opportunities[J].European Transport Research Review,2020,12:55.
[2] AHMED F,HAWAS Y E.An integrated real-time traffic signal system for transit signal priority,incident detection and congestion management[J].Transportation Research Part C:Emerging Technologies,2015,60:52-76.
[3] IMRAN N T,GIRIJAN A,VANAJAKSHI L D.A review on bus signal priority systems[J].Transportation in Developing Economies,2022,8:2.
[4] GHAFFARI A,MESBAH M,KHODAII A.Designing a transit priority network under variable demand[J].Transportation Letters,2020,12(6):427-440.
[5] TRUONG L T,CURRIE G,SARVI M.Analytical and simulation approaches to understand combined effects of transit signal priority and road-space priority measures[J].Transportation Research Part C:Emerging Technologies,2017,74:275-294.
[6] HAMILTON A,WATERSON B,CHERRETT T,et al.The evolution of urban traffic control:changing policy and technology[J].Transportation Planning and Technology,2013,36(1):24-43.
[7] 马万经,杨晓光.基于车道的单点交叉口公交被动优先控制模型[J].中国公路学报,2010,23(5):96-101.
[8] 林永杰,杨险峰,邹难,等. 城市交通干道上被动式公交信号优先控制[J].东北大学学报,2013,34(9):1227-1231.
[9] CHRISTOFA E M,SKABARDONIS A.Traffic signal optimization with conditional transit signal priority for conflicting transit routes[C]//12th World Conference on Transprt Research.Lisbon:[s.n. ],2010:11-15.
[10] YU Z,GAYAH V V,CHRISTOFA E.Implementing phase rotation in a person-based signal timing optimization framework[C]//2018 21st International Conference on Intelligent Transportation Systems(ITSC).Maui,HI:IEEE,2018:20-25.
[11] 刘昱岗,李佳励,张擎,等. 基于动态权重的公交优先信号配时优化策略[J].交通运输系统工程与信息,2016,16(3):53-59.
[12] 于泽,宁念文,郑燕柳,等. 深度强化学习驱动的智能交通信号控制策略综述[J].计算机科学,2023,50(4):159-171.
[13] WANG Z,YANG K,LI L,et al.Traffic signal priority control based on shared experience multi-agent deep reinforcement learning[J].IET Intelligent Transport Systems,2023,17(7):1363-1379.
[14] BERTSEKAS D P.Feature-based aggregation and deep reinforcement learning:a survey and some new implementations[J].IEEE/CAA Journal of Automatica Sinica,2019,6(1):1-31.
[15] LI L,LV Y,WANG F Y.Traffic signal timing via deep reinforcement learning[J].IEEE/CAA Journal of Automatica Sinica,2016,3(3):247-254.
[16] LIANG X,DU X,WANG G,et al.A deep reinforcement learning network for traffic light cycle control[J].IEEE Transactions on Vehicular Technology,2019,68(2):1243-1253.
[17] ZHONG N,LIU K,LI Y.Deep Q-Learning network model for optimizing transit bus priority at multiphase traffic signal controlled intersection[J].Mathematical Problems in Engineering,2023,2023:9137889.
[18] MOUSAVI S S,SCHUKAT M,HOWLEY E.Traffic light control using deep policy-gradient and value-function-based reinforcement learning[J].IET Intelligent Transport Systems,2017,11(7):417-423.
[19] WU C,KIM I,MA Z.Deep reinforcement learning based traffic signal control:a comparative analysis[J].Procedia Computer Science,2023,220:275-282.
[20] ZENG J,HU J,ZHANG Y.Training reinforcement learning agent for traffic signal control under different traffic conditions[C]//2019 IEEE Intelligent Transportation Systems Conference(ITSC).Auckland,New Jealand:IEEE,2019:4248-4254.
[21] FANG S,CHEN F,LIU H.Dueling double deep Q-network for adaptive traffic signal control with low exhaust emissions in a single intersection[J].IOP Conference Series:Materials Science and Engineering,2019,612(5):052039.
[22] HUANG Y,WEI G L,WANG Y X.V-D D3QN:the variant of double deep Q-learning network with dueling architecture[C]//2018 37th Chinese Control Conference(CCC).Wuhan,China:IEEE,2018:9130-9135.

基于深度强化学习的交通信号优先控制

Traffic signal priority control based on deep reinforcement learning

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 3

编辑推荐

Metrics

本文评价

[1]	田春婷, 赵宁, 秦建伟, 孟晓凤. 基于机器学习的红外光谱数据鉴别中药材性能方法[J]. 兰州理工大学学报, 2023, 49(3): 55-59.
[2]	许小芾. 剩余型直觉模糊推理的模糊取式反向三Ⅰ算法的性质研究[J]. 兰州理工大学学报, 2023, 49(3): 83-87.
[3]	陈鹏, 赵小强, 朱奇先. 基于VMD-MPE-KPCA特征提取与MRVM相混合的滚动轴承故障诊断方法[J]. 兰州理工大学学报, 2020, 46(5): 92-99.