汽车工程 ›› 2023, Vol. 45 ›› Issue (1): 9-19.doi: 10.19562/j.chinasae.qcgc.2023.01.002

所属专题: 智能网联汽车技术专题-规划&决策2023年

• • 上一篇    下一篇

基于动态博弈算法的切入场景下自动驾驶车辆运动规划研究

兰凤崇,刘迎节,陈吉清(),刘照麟   

  1. 1.华南理工大学机械与汽车工程学院,广州 510640
    2.华南理工大学,广东省汽车工程重点实验室,广州 510640
  • 收稿日期:2022-02-22 修回日期:2022-03-22 出版日期:2023-01-25 发布日期:2023-01-18
  • 通讯作者: 陈吉清 E-mail:chenjq@scut.edu.cn
  • 基金资助:
    国家自然科学基金(52175267);广东省科学计划项目(2015B010137002);国家车辆事故深度调查体系项目资助

Study on Motion Planning of Autonomous Vehicles in Cut-in Scenes Based on Dynamic Game Algorithm

Fengchong Lan,Yingjie Liu,Jiqing Chen(),Zhaolin Liu   

  1. 1.School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640
    2.South China University of Technology,Guangdong Provincial Key Laboratory of Automotive Engineering,Guangzhou 510640
  • Received:2022-02-22 Revised:2022-03-22 Online:2023-01-25 Published:2023-01-18
  • Contact: Jiqing Chen E-mail:chenjq@scut.edu.cn

摘要:

鉴于在车辆换道切入的场景中,自动驾驶车辆容易出现频繁的误减速、误避让,而造成通行能力和乘员舒适性的下降,提出一种基于主旁车动态博弈的切入场景决策规划算法。在行为决策层,根据切入场景中主旁车的冲突性关系,联立相关车辆运动方程建立整体系统的运动模型,构建考虑旁车状态的切入博弈模型,设计安全性和舒适性收益函数,进行驾驶行为博弈,输出行为决策结果。在轨迹规划层,根据车辆间距构建避障约束条件,以Sigmoid函数轨迹的变曲率和速度切向矢量的时间分量来构建车辆动力学约束。同时以加权收益方式联合考虑驾驶习惯和舒适性等需求,建立轨迹规划数学模型,求解得到满足上层博弈决策要求的运动轨迹。Carsim-Simulink联合仿真结果表明,在不同的初始条件下主车与切入的旁车能进行多种形式的合理的交互决策,准确完成切入场景下的运动规划任务,车辆能准确跟踪输出的轨迹,更符合一般驾驶习惯,提高了车辆的舒适性。

关键词: 自动驾驶车辆, 切入场景, 动态博弈, 运动规划

Abstract:

In view of that in vehicle lane change and cut-in scenes, autonomous vehicles are prone to frequent false-deceleration and false-avoidance, leading to the reduction in traffic capacity and occupant comfort, a decision-making and planning algorithm based on the dynamic gaming between ego vehicle and adjacent vehicle is proposed. In the behavior decision-making layer, according to the conflicting relationship between ego vehicle and adjacent vehicle in the cut-in scene, the motion model of the whole system is established by combining the motion equations of relevant vehicles, and the cut-in game model considering the state of adjacent vehicle is constructed, to design the gain function of safety and rid comfort, conduct driving behavior game, and output behavior decision results. In the trajectory planning layer, the constraints for obstacle avoidance is established based on vehicle spacing, and the vehicle dynamics constraints are defined by the variable curvature of sigmoid function trajectory and the time component of speed tangent vector. Meanwhile, with concurrent considerations of the requirements of driving habits and ride comfort in weighted gain mode, a mathematical model of trajectory planning is established to solve out the motion trajectory, meeting the requirements of the upper-level game decision. The results of Carsim-Simulink co-simulation show that under different initial conditions, the ego vehicle and cut-in adjacent vehicle can conduct various forms of rational interactive decision-making, accurately complete the motion planning task in cut-in scene. In addition, the vehicles can accurately follow the output trajectory, more conforming the common driving habits and enhancing vehicle comfort.

Key words: autonomous vehicles, cut-in scenes, dynamic game, motion planning