汽车工程 ›› 2023, Vol. 45 ›› Issue (7): 1099-1111.doi: 10.19562/j.chinasae.qcgc.2023.07.001

所属专题: 智能网联汽车技术专题-规划&决策2023年

• 专题:汽车智能化关键技术 •    下一篇

面向变曲率道路的自动驾驶汽车换道博弈运动规划与协同控制研究

林程1,汪博文1,吕沛原1,宫新乐1,2(),于潇1   

  1. 1.北京理工大学,电动车辆国家工程研究中心,北京  100081
    2.清华大学车辆与运载学院,北京  100084
  • 收稿日期:2023-01-03 修回日期:2023-02-05 出版日期:2023-07-25 发布日期:2023-07-25
  • 通讯作者: 宫新乐 E-mail:Xinlegong@gmail.com
  • 基金资助:
    国家自然科学基金(51975049)

Research on Motion Planning and Cooperative Control for Autonomous Vehicles with Lane Change Gaming Maneuvers Under the Curved Road

Cheng Lin1,Bowen Wang1,Lü Peiyuan1,Xinle Gong1,2(),Xiao Yu1   

  1. 1.Beijing University of Technology,National Engineering Research Center for Electric Vehicles,Beijing  100081
    2.School of Vehicle and Mobility,Tsinghua University,Beijing  100084
  • Received:2023-01-03 Revised:2023-02-05 Online:2023-07-25 Published:2023-07-25
  • Contact: Xinle Gong E-mail:Xinlegong@gmail.com

摘要:

当多辆自动驾驶车辆在结构化道路上执行换道合流任务时,需要综合考虑转向和合流动作以避免事故的发生,同时曲率和周车车速的变化也增大了协同控制的难度。本文聚焦上述问题,提出了面向变曲率道路的多车换道博弈运动规划与协同控制方法。首先,建立曲率坐标系下的多车模型来解析车间安全距离及动力学状态。其次,通过系统地考虑道路曲率变化及周围车辆信息,提出基于博弈的多车换道运动规划算法,采用分布式框架快速求解兼顾个性化驾驶的最优速度轨迹及换道时机。最后,基于B样条曲线高效识别道路曲率及规划轨迹,构建了自适应时变预测控制算法实现轨迹跟踪,其特点在于单步参数矩阵实时更新,消除车速和曲率频繁变化带来的控制偏差累积。实验结果表明,相比斯坦利(Stanley)方法,能降低58%的跟踪误差;相较协同自适应巡航方法,能减少74%的合流时间;另外计算求解效率也仅为集中式模型预测控制方法的10%。

关键词: 自动驾驶汽车, 多车博弈, 模型预测控制, 换道运动规划

Abstract:

When multiple autonomous vehicles perform lane change and merging tasks on structured road, steering and merging actions need to be comprehensively considered to avoid potential accidents. Meanwhile, the changing road curvature and surrounding vehicle speed also increase the difficulty of cooperative control. Focusing on the above issues, this paper proposes a multi-vehicle lane change gaming motion planning and cooperative control method facing variable curvature road. Firstly, a multi-vehicle model in curvature coordinate system is developed to determine the inter-vehicle safety distance and dynamics state. Then, by systematically considering the road curvature variation and surrounding vehicle information, a game-based multi-vehicle lane change motion planning algorithm is proposed, which uses a distributed framework to quickly solve the optimal speed trajectory and lane change timing considering personalized driving. Finally, the road curvature and planning trajectory are identified effectively based on B-sample curve, and an adaptive time-varying model predictive control algorithm is constructed to achieve trajectory tracking. Specifically, the control parameters are updated in real time under the single-step prediction domain to eliminate the control deviations caused by frequently various vehicle speed and curvature. The co-simulation results show that the proposed method can reduce the tracking error by 58% compared to the Stanley method, with reduction of the merging time by 74% compared to the cooperative adaptive cruise control method. Moreover, the computational solution efficiency is only 10% of the centralized method.

Key words: autonomous vehicles, multi-vehicle gaming, model predictive control, motion planning for lane changing