With the advent of autonomous driving, the seamless integration of autonomous vehicles (AVs) with humandriven vehicles (HDVs) poses significant challenges. Particularly, at unsignalized intersections lacking explicit coordination, safely predicting and accommodating the unprotected left-turn behaviors of drivers with varying preferences is challenging. This study introduces a novel framework that integrates Social Value Orientation (SVO) with Deep Reinforcement Learning (DRL) and a Motion Safety Controller (MSC), targeting the enhancement of AVs safety and efficiency in these complex scenarios. Integrating the SVO into the DRL reward function encourages socially considerate behaviors among AVs to minimize potential incidents. Equipped with a track supervisor and a motion controller, the MSC assists in foreseeing and mitigating hazards, thus enhancing road safety. Our study trains, evaluates, and validates the efficacy of our proposed method on a gym-like highway simulator. The method surpasses existing state-of-the-art (SOTA) baseline models in reducing collision rates and improving vehicular behavior. This study represents a significant advancement in autonomous driving, addressing complex unprotected left turns at intersections and aligning technological innovations with the societal need for safer roads. The video demo of the evaluation process can be found at: https://youtu.be/H4E_qUT15Qgcom


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Integrating Social Value Orientation and Motion Safety Controller in Autonomous Driving: Improving the Safety of Unprotected Left-Turn Behaviour


    Beteiligte:
    Zhang, Chuancheng (Autor:in) / Wang, Zhenhao (Autor:in) / Lv, Chenyang (Autor:in) / Cui, Yanhao (Autor:in) / Jiang, Bin (Autor:in) / Guo, Qiang (Autor:in)


    Erscheinungsdatum :

    24.09.2024


    Format / Umfang :

    942781 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Unprotected left-turn driving control method based on deep reinforcement learning

    ZHAO MIN / SUN DIHUA / CHEN JIN | Europäisches Patentamt | 2021

    Freier Zugriff

    APPARATUS AND METHOD FOR PROVIDING WARNING IN CASE OF UNPROTECTED LEFT TURN OR UNPROTECTED U-TURN OF AUTONOMOUS VEHICLE

    YUN ILSOO / KIM JINYOUNG / PARK SUNGHO et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    Analysis of Unprotected Intersection Left-Turn Conflicts Based on Naturalistic Driving Data

    Wang, Xinpeng / Zhao, Ding / Peng, Huei et al. | British Library Conference Proceedings | 2017