Scholarlink

With the advancement of vehicle-to-vehicle (V2V), vehicle platooning has emerged as a promising approach to enhance traffic efficiency and safety on freeways. Even though the longitudinal car-following strategy within platoons has been widely studied, the lateral lane-changing strategy remains a challenging area, where an efficient method is required to cooperate platoon members to execute lane-changing. To this end, this paper proposes a novel hybrid control model based on reinforcement learning (RL) for platoons at mixed-traffic freeways. In specific, a double deep Q network (DDQN) deployed on the platoon leader produces the lane-changing strategy disseminated to platoon followers via V2V devices. Subsequently, the lane-changing mode of each platoon member is activated, which enables the entire platoon to execute lane-changing action simultaneously. Moreover, adaptive cruise control (ACC) and cooperative adaptive cruise control (CACC) are deployed on platoon leader and platoon followers for car-following control, respectively. Simulation results demonstrate that the proposed model outperforms rule-based baseline methods by enhancing 10% in terms of traffic efficiency.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Hybrid Control Model for Platoons at Mixed-Traffic Freeways Based on Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Han Li,
  • Xiangkai Zhou,
  • Sheng Liu,
  • Tao Wang,
  • Linlin You

摘要

With the advancement of vehicle-to-vehicle (V2V), vehicle platooning has emerged as a promising approach to enhance traffic efficiency and safety on freeways. Even though the longitudinal car-following strategy within platoons has been widely studied, the lateral lane-changing strategy remains a challenging area, where an efficient method is required to cooperate platoon members to execute lane-changing. To this end, this paper proposes a novel hybrid control model based on reinforcement learning (RL) for platoons at mixed-traffic freeways. In specific, a double deep Q network (DDQN) deployed on the platoon leader produces the lane-changing strategy disseminated to platoon followers via V2V devices. Subsequently, the lane-changing mode of each platoon member is activated, which enables the entire platoon to execute lane-changing action simultaneously. Moreover, adaptive cruise control (ACC) and cooperative adaptive cruise control (CACC) are deployed on platoon leader and platoon followers for car-following control, respectively. Simulation results demonstrate that the proposed model outperforms rule-based baseline methods by enhancing 10% in terms of traffic efficiency.