The invention relates to a method for constructing a deep reinforcement learning vehicle following model fused with a driving style, which comprises the following steps of: actually measuring a high-precision traffic flow data set based on an NGSIM (Next Generation Subscriber Identity Module) project, setting a corresponding rule to screen original data to obtain following pair data meeting conditions, and dividing the following pair data into training data and verification data; the driving characteristics of a driver are analyzed based on training data, and a driving style identification model is built based on a support vector machine algorithm. Based on car-following behavior analysis results of drivers with different characteristics, car-following spacing design reinforcement learning reward functions of different driving styles are fused to build a car-following model based on a depth deterministic strategy gradient algorithm. The model is trained in the established training environment to obtain deep reinforcement learning car-following models of different driving styles, and the established car-following models can reflect different driving characteristics on the premise of ensuring safety and comfort based on simulation result display of verification data.

    本发明涉及一种融合驾驶风格的深度强化学习车辆跟驰模型的构建方法,基于NGSIM项目实测高精度交通流数据集,设定相应的规则对原始数据进行筛选,得到符合条件的跟驰对数据,将跟车对数据分为训练数据和验证数据。基于训练数据对驾驶人的驾驶特性进行分析,同时基于支持向量机算法搭建了驾驶风格辨识模型。基于对不同特性驾驶员的跟驰行为分析结果,融合不同驾驶风格的跟车间距设计强化学习奖励函数搭建基于深度确定性策略梯度算法的跟驰模型。在所搭建的训练环境中对模型进行训练得到不同驾驶风格的深度强化学习跟驰模型,基于验证数据的仿真结果显示,所搭建的跟驰模型在保证安全性和舒适性的前提下能够体现不同的驾驶特性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Construction method of deep reinforcement learning vehicle following model fused with driving style


    Weitere Titelangaben:

    融合驾驶风格的深度强化学习车辆跟驰模型的构建方法


    Beteiligte:
    WANG KEYIN (Autor:in) / SHI ZHEN (Autor:in) / ZHANG JIANHUI (Autor:in) / YANG ZHENGCAI (Autor:in)

    Erscheinungsdatum :

    04.11.2022


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion



    Autonomous Vehicle Driving Path Control with Deep Reinforcement Learning

    Tiong, Teckchai / Saad, Ismail / Teo, Kenneth Tze Kin et al. | IEEE | 2023


    Vehicle following speed control method based on deep reinforcement learning

    FEI RONG / YANG LU / QIU YUAN et al. | Europäisches Patentamt | 2023

    Freier Zugriff


    Driver longitudinal car-following behavior model construction method based on deep reinforcement learning

    GUO JINGHUA / LI WENCHANG / WANG JINGYAO et al. | Europäisches Patentamt | 2021

    Freier Zugriff

    Autonomous Driving with Deep Reinforcement Learning

    Zhu, Yuhua / Technische Universität Dresden | SLUB | 2023