汽车工程 ›› 2024, Vol. 46 ›› Issue (3): 383-395.doi: 10.19562/j.chinasae.qcgc.2024.03.002

• • 上一篇    

面向结构化道路的智能驾驶轨迹规划一致性研究

吴晓建1,2(),廖平伟1,4,雷耀2,江会华2,王爱春2,胡家琦3   

  1. 1.南昌大学先进制造学院,南昌 330031
    2.江铃汽车股份有限公司,南昌 330001
    3.南昌航空大学测试与光电工程学院,南昌 330038
    4.比亚迪股份有限公司,深圳 518119
  • 收稿日期:2023-07-09 修回日期:2023-09-08 出版日期:2024-03-25 发布日期:2024-03-18
  • 通讯作者: 吴晓建 E-mail:saintwu520@163.com
  • 基金资助:
    国家自然科学基金(52262054)

Research on Consistency of Intelligent Driving Trajectory Planning for Structured Road

Xiaojian Wu1,2(),Pingwei Liao1,4,Yao Lei2,Huihua Jiang2,Aichun Wang2,Jiaqi Hu3   

  1. 1.College of Advanced Manufacturing,Nanchang University,Nanchang 330031
    2.Jiangling Motors Corporation,Ltd. ,Nanchang 330001
    3.School of Measuring and Optical Engineering,Nanchang Hangkong University,Nanchang 330038
    4.BYD Auto Co. ,Ltd. ,Shenzhen  518119
  • Received:2023-07-09 Revised:2023-09-08 Online:2024-03-25 Published:2024-03-18
  • Contact: Xiaojian Wu E-mail:saintwu520@163.com

摘要:

智能车在动态环境中的轨迹规划须具备良好的舒适性及安全性,离散采样轨迹规划算法具有实时性高、多目标最优等优点而被广泛研究和应用,但在仿真及实车测试中发现,典型的基于多项式优化求解的离散采样局部轨迹规划结果在换道等瞬态过程存在一致性较差的问题。本文针对性提出一种考虑一致性评价的“拼接+强规划”轨迹规划算法。具体而言,根据自车状态截取历史轨迹为当前周期拼接轨迹,结合拼接轨迹和轨迹末状态采样点生成基于多项式的候选轨迹簇作为轨迹强规划阶段,再基于轨迹横向偏差设计轨迹一致性评价函数以从轨迹簇中选取较高一致性的最优行驶轨迹。经仿真和真实道路场景实车验证,表明所提出的轨迹规划方法在满足轨迹安全性、平顺性、舒适性要求的基础上提高了智能驾驶车辆行驶轨迹的整体一致性。

关键词: 智能驾驶, 局部轨迹规划, 轨迹一致性, 避障

Abstract:

Trajectory planning for intelligent vehicles in dynamic environment needs to have good comfort and safety. Discrete sampling trajectory planning algorithms have been widely studied and applied due to high real-time performance and multi-objective optimality. However, it is found in simulations and real vehicle tests that the results of local trajectory planning using typical methods such as polynomial optimization suffer from poor consistency during transient processes like lane changing. In this paper, a "splice+rigid planning" trajectory planning algorithm that considers consistency evaluation is proposed. Specifically, historical trajectories are spliced with the current cycle trajectory based on the vehicle's state. Polynomial-based candidate trajectory clusters are generated by combining the spliced trajectory with sampled points from the trajectory end state, which serves as the rigid planning phase. Then, a trajectory consistency evaluation function is designed based on the lateral deviation of the trajectory to select the optimal driving trajectory with higher consistency from the trajectory cluster. The results of simulation and real road scenario tests show that the proposed trajectory planning method improves the overall consistency of the intelligent driving vehicle's trajectory while meeting requirements for trajectory safety, smoothness, and comfort.

Key words: intelligent driving, local trajectory planning, trajectory consistency, obstacle avoidance