• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2024, Vol. 60 ›› Issue (20): 217-228.doi: 10.3901/JME.2024.20.217

• 运载工程 • 上一篇    下一篇

扫码分享

面向无信号灯十字路口的自动驾驶多车协同控制方法研究

唐小林, 杨剑英, 杨凯, 李文博   

  1. 重庆大学机械与运载工程学院 重庆 400044
  • 收稿日期:2024-03-25 修回日期:2024-07-21 出版日期:2024-10-20 发布日期:2024-11-30
  • 通讯作者: 李文博,男,1991年出生,副教授。主要研究方向为汽车智能座舱人机交互、驾驶人行为状态认知与建模。E-mail:liwenbocqu@qq.com,wenbo_li@cqu.edu.cn
  • 作者简介:唐小林,男,1984年出生,教授,博士研究生导师。主要研究方向为智能网联新能源汽车决策规划与节能控制。E-mail:tangxl0923@cqu.edu.cn
  • 基金资助:
    国家自然科学基金(52222215,52302497)和中央高校基金(2023CDJXY-025)资助项目。

Research on Multi-vehicle Cooperative Control Method for Autonomous Driving at Unsignalized Intersections

TANG Xiaolin, YANG Jianying, YANG Kai, LI Wenbo   

  1. College of Mechanical and Vehicle Engineering, Chongqing University, Chongqing 400044
  • Received:2024-03-25 Revised:2024-07-21 Online:2024-10-20 Published:2024-11-30

摘要: 针对无信号灯十字路口场景下多车协同控制安全性不足、交互性差的问题,提出一种分布式自动驾驶多车协同控制方法。首先,利用轨迹预测模型获取路口内所有车辆未来3 s的运动轨迹,作为冲突筛查的输入;其次,建立基于责任敏感安全(Responsibility sensitive safety,RSS)的冲突筛查模型,通过对车辆的预测轨迹进行评估获得车辆之间的交互关系;再次,构建基于博弈论和智能驾驶员模型(Intelligent driver model,IDM)的多车协同控制框架,将冲突车辆之间的交互过程建模为纳什均衡问题,通过求解优化问题生成多车协同控制策略,无需交互的安全车辆由IDM进行控制,从而实现对所有车辆的分别控制和联合行动。最后,在无信号灯十字路口仿真场景对所提出方法进行验证。试验结果表明,与基准方法相比,所提出方法的安全性、实时性和通行效率得到有效提升。

关键词: 无信号灯十字路口, 多车协同控制, 博弈论, 运动预测, 安全评估

Abstract: Addressing the insufficient safety and poor interactivity issues at unsignalized intersections with multiple vehicles, a distributed autonomous driving multi-vehicle cooperative control method is proposed. Firstly, a trajectory prediction model is utilized to obtain the future 3-second motion trajectories of all vehicles within the intersection as input for conflict screening. Secondly, a safety assessment model based on the responsibility-sensitive safety(RSS) model is established to evaluate the interaction between vehicles by assessing their predicted trajectories. Thirdly, a multi-vehicle cooperative control framework based on game theory and the intelligent driver model(IDM) is constructed. The interaction process between conflicting vehicles is modeled as a Nash equilibrium problem, and cooperative control strategies for multiple vehicles are generated by solving optimization problems. Vehicles that do not interact with other vehicles are controlled by IDM, thereby achieving separate control and joint action for all vehicles. Finally, the proposed method is validated in a simulated scenario of an unsignalized intersection. Experimental results demonstrate that compared to baseline methods, the proposed approach effectively enhances safety, real-time performance, and traffic efficiency.

Key words: unsignalized intersections, multi-vehicle control, game theory, trajectory prediction, safety assessment

中图分类号: