汽车工程 ›› 2024, Vol. 46 ›› Issue (6): 975-984.doi: 10.19562/j.chinasae.qcgc.2024.06.004

• • 上一篇    下一篇

基于时空风险的智能驾驶车辆避险决策规划

杨超1,2,杨帆1,王伟达1,2(),郄天琪1,王彦松1,王宏伟2   

  1. 1.北京理工大学机械与车辆学院,北京 100081
    2.北京理工大学长三角研究院(嘉兴),嘉兴 314000
  • 收稿日期:2023-12-21 修回日期:2024-01-31 出版日期:2024-06-25 发布日期:2024-06-19
  • 通讯作者: 王伟达 E-mail:wangwd0430@bit.edu.cn
  • 基金资助:
    国家自然科学基金(52275047);重庆自然科学基金(cstc2021jcyj-msxmX0879)

Risk Avoidance Decision Planning for Intelligent Driving Vehicles Based on Spatiotemporal Risk

Chao Yang1,2,Fan Yang1,Weida Wang1,2(),Tianqi Qie1,Yansong Wang1,Hongwei Wang2   

  1. 1.School of Mechanical Engineering,Beijing Institute of Technology,Beijing  100081
    2.Yangtze Delta Region Academy of Beijing Institute of Technology,Jiaxing  314000
  • Received:2023-12-21 Revised:2024-01-31 Online:2024-06-25 Published:2024-06-19
  • Contact: Weida Wang E-mail:wangwd0430@bit.edu.cn

摘要:

为完善复杂道路场景下智能驾驶车辆的碰撞风险评估方法,实时生成有效避险轨迹,本文提出了一种基于时空风险的智能驾驶车辆避险决策规划方法。首先,采用时空耦合的多域风险度量作为评估指标,监督智能驾驶车辆横纵向碰撞风险,同时实时监测行车风险指标变化,通过与典型避险场景风险数据库间的相关性分析,判断潜在碰撞风险大小,从而提前规避风险。然后,依据行车风险场对车辆目标状态进行不均匀采样,规避行车风险较高的驾驶区域,提高避险规划的安全性和实时性。试验结果表明,所提出的避险决策规划方法可安全有效地避开横纵向碰撞风险,并且根据风险指标的时域相关性分析可提前0.5 s发现潜在碰撞风险,从而提前平稳规避风险,不均匀采样可将避险轨迹平均规划时间由0.13 缩短到0.07 s。

关键词: 智能驾驶, 决策规划, 风险评估, 行车风险场

Abstract:

In order to improve the collision risk assessment method of intelligent driving vehicles in complex road scenarios to generate effective risk avoidance trajectories in real time, in this paper a risk avoidance decision planning method for intelligent driving vehicles based on spatiotemporal risk is proposed. Firstly, the multi-domain risk measurement of time-space coupling is used as the evaluation index to supervise the horizontal and longitudinal collision risk of intelligent driving vehicles. At the same time, the change of driving risk index is monitored in real time. Through the correlation analysis with the risk database of typical risk avoidance scenarios, potential collision risk is judged, so as to avoid the risk in advance. Then, according to the driving risk field, the uneven sampling of the vehicle target state is carried out to avoid the driving area with high driving risk and improve the safety and real-time performance of the risk avoidance planning. The experimental results show that the proposed risk avoidance decision planning method can safely and effectively avoid the risk of horizontal and longitudinal collision, which can detect the potential collision risk 0.5 s in advance according to the time domain correlation analysis of the risk index, so as to avoid the risk smoothly in advance. The average planning time of the risk avoidance trajectory can be shortened from 0.13 to 0.07 s by uneven sampling.

Key words: intelligent driving, decision planning, risk assessment, driving risk field