DynLight: Realize dynamic phase duration with multi-level traffic signal control
College of Ecology, Lanzhou University(兰州大学生态学院)
Liang Zhang, Shubin Xie, Jianming Deng
备注:
We would like to withdraw this article for the following reasons: 1 this article is not satisfactory for limited language and theoretical description; 2 we have enriched and revised this article with the help of other authors; 3 we must update the author contribution information. PLease see: arXiv:2211.01025
We would like to withdraw this article for the following reasons: 1 this article is not satisfactory for limited language and theoretical description; 2 we have enriched and revised this article with the help of other authors; 3 we must update the author contribution information.
A Bi-level Framework for Traffic Accident Duration Prediction: Leveraging Weather and Road Condition Data within a Practical Optimum Pipeline
Ahsanullah University of Science and Technology(阿萨努拉科技大学);Bangladesh University of Engineering and Technology(孟加拉工程技术大学);Advanced Chemical Industries Limited(先进化学工业有限公司)
DynamicLight: Two-Stage Dynamic Traffic Signal Timing
College of Ecology, Lanzhou University(兰州大学生态学院);School of Artificial Intelligence, Beijing University of Posts and Telecommunications(北京邮电大学人工智能学院);Graduate School of Informatics, Nagoya University(名古屋大学信息学研究科)
A Parallel Hybrid Action Space Reinforcement Learning Model for Real-world Adaptive Traffic Signal Control
National Center For Applied Mathematics In Chongqing(重庆国家应用数学中心);Chongqing Normal University(重庆师范大学);Institute of Fundamental and Frontier Sciences(前沿与基础科学研究院);University of Electronic Science and Technology of China(电子科技大学);Chongqing Jiaotong University(重庆交通大学)
Efficient Pressure: Improving efficiency for signalized intersections
Institute of Fundamental and Frontier Sciences, University of Electronic Science and Technology of China(电子科技大学基础与前沿研究院);School of Life Sciences, Lanzhou University(兰州大学生命科学学院);School of Computing and Information Technology, University of Wollongong(伍伦贡大学计算机与信息技术学院);SMART Infrastructure Facility, University of Wollongong(伍伦贡大学SMART基础设施研究所)
Large-Scale Traffic Signal Control by a Nash Deep Q-network Approach
强化学习(Reinforcement Learning,RL)是目前最常用的交通信号控制(Traffic Signal Control,TSC)技术之一,能够根据实时交通数据自适应地调整交通信号的相位和时长。然而,一个完全集中的强化学习方法在多网络场景中遇到了困难,因为状态-动作空间随着交叉点的增加而呈指数增长。多智能体强化学习(MARL)通过对每个局部强化学习智能体的全局控制来克服高维问题,但也带来了新的挑战,如非平稳马尔可夫决策过程(MDP)导致的收敛失败。在本文中,我们介绍了一种非策略纳什深度Q网络(OPNDQN)算法,它减轻了完全集中式和MARL方法的弱点。OPNDQN算法解决了传统算法不能用于大型状态-行为空间交通模型的问题,在每次迭代中利用虚拟博弈方法找到相邻交叉口之间的纳什均衡,没有交叉口有动机单方面偏离该均衡。OPNDQN的主要优点之一是减轻非平稳性的多智能体马尔可夫过程,因为它考虑了相邻交叉口之间的相互影响,通过共享他们的行动。另一方面,对于训练一个大的交通网络,OPNDQN的收敛速度高于现有的MARL方法,因为它不包括每个代理的所有状态信息。我们进行了广泛的实验,使用模拟城市流动性模拟器(SUMO),并显示了显着的优越性OPNDQN超过现有的几个MARL方法的平均队列长度,情节训练奖励和平均等待时间。