The invention discloses a deep reinforcement learning traffic light control method. The method comprises the following steps: (1) preprocessing traffic network data in a city; (2) constructing a model by utilizing a Multi-step DQN algorithm according to the preprocessed data; (3) accumulating the experiences of the n single steps, and learning by using the accumulated experiences; (4) the network parameters of the Multi-step DQN are updated; (5), combining the Attention expansion play with the DQN network, and constructing a deep reinforcement learning model; (6) importing the traffic data set and the traffic flow data set into the deep reinforcement learning model for training, and recording an experimental result; (7) comparing experimental results in the step (2) and the step (5); (8) carrying out visual display; compared with a traditional control method, the MALlight has the advantage that the MALlight is better in reducing the average passing time of the vehicles and improving the average throughput of the intersection.

    本发明公开了一种深度强化学习交通灯控制方法,包括以下步骤:(1)对城市内交通网络数据进行预处理;(2)根据预处理后的数据利用Multi‑step DQN算法构建模型;(3)将n个单步的经验进行累加,再利用累加后的经验进行学习;(4)更新Multi‑step DQN的网络参数;(5)将Attentive experience replay与DQN网络相结合,构建深度强化学习模型;(6)将交通数据集、车流数据集导入深度强化学习模型,进行训练,并记录实验结果;(7)比较步骤(2)和步骤(5)中的实验结果;(8)进行可视化展示;本发明与传统控制方法相比,MALight在减少车辆的平均通行时间和提高路口的平均吞吐量做得更好。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Deep reinforcement learning traffic light control method


    Weitere Titelangaben:

    一种深度强化学习交通灯控制方法


    Beteiligte:
    KONG YAN (Autor:in) / LI YING (Autor:in) / CHIH-CHAO YANG (Autor:in)

    Erscheinungsdatum :

    2024-03-15


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS



    Traffic light control method based on deep reinforcement learning and inverse reinforcement learning

    ZHANG YA / GU SHIYI / CHEN GUOXI | Europäisches Patentamt | 2023

    Freier Zugriff

    Adaptive traffic light control method based on deep reinforcement learning

    KONG YAN / CHO JUN-HO | Europäisches Patentamt | 2023

    Freier Zugriff

    Deep Reinforcement Learning for Autonomous Traffic Light Control

    Garg, Deepeka / Chli, Maria / Vogiatzis, George | IEEE | 2018


    Deep Reinforcement Learning-based Traffic Signal Control

    Ruan, Junyun / Tang, Jinzhuo / Gao, Ge et al. | IEEE | 2023


    Intelligent Traffic Light via Policy-based Deep Reinforcement Learning

    Zhu, Yue / Cai, Mingyu / Schwarz, Chris W. et al. | Springer Verlag | 2022