东北大学学报(自然科学版) ›› 2022, Vol. 43 ›› Issue (7): 930-936.DOI: 10.12068/j.issn.1005-3026.2022.07.003

• 信息与控制 • 上一篇    下一篇

考虑交互轨迹预测的自动驾驶运动规划算法

刘启冉1, 连静1,2, 陈实1, 范蓉1   

  1. (1.大连理工大学 汽车工程学院, 辽宁 大连116024; 2.大连理工大学 工业装备结构分析国家重点实验室, 辽宁 大连116024)
  • 发布日期:2022-08-02
  • 通讯作者: 刘启冉
  • 作者简介:刘启冉(1995-),男,河北唐山人,大连理工大学硕士研究生; 连静(1980-),女,吉林公主岭人,大连理工大学副教授,博士生导师.
  • 基金资助:
    国家自然科学基金资助项目(51775082,61976039,52172382);中央高校基本科研业务费专项资金资助项目(DUT20GJ207); 大连市科技创新基金资助项目(2021JJ12GX015).

Motion Planning Algorithm of Autonomous Driving Considering Interactive Trajectory Prediction

LIU Qi-ran1, LIAN Jing1,2, CHEN Shi1, FAN Rong1   

  1. 1. School of Automotive Engineering, Dalian University of Technology, Dalian 116024, China; 2. State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian 116024, China.
  • Published:2022-08-02
  • Contact: LIAN Jing
  • About author:-
  • Supported by:
    -

摘要: 针对自动驾驶汽车运动规划中预测周围交通态势的问题,提出一种考虑周围车辆间交互轨迹预测的运动规划算法.首先,针对结构化道路信息,构建改进社会力模型,对自动驾驶汽车周围的车辆行驶轨迹进行预测.其次,在Frenet坐标系下采样生成轨迹集合,将轨迹集合和预测轨迹投影到时空占用图上,计算投影点之间的最短距离进行碰撞检查,并结合加速度、曲率检查对轨迹进行筛选以得到候选轨迹.然后,构建代价函数对筛选过的候选轨迹进行评估得到最优运动轨迹.最后,不同行驶场景中的仿真结果表明,该运动规划算法能提前决策驾驶行为,规划出的速度曲线更加平稳,运动轨迹的安全性、舒适性和行驶效率更高.

关键词: 自动驾驶;运动规划;交互轨迹预测;社会力模型;代价函数

Abstract: Aiming at the problem of predicting the surrounding traffic situation in autonomous vehicle motion planning, a motion planning algorithm considering the interactive trajectory prediction between surrounding vehicles is proposed. Firstly, for structured road information, an improved social force model is constructed to predict the trajectory of vehicles around autonomous vehicles. Secondly, the predicted trajectory and the trajectory set that is generated in the Frenet coordinate system are projected on the space-time occupancy map, and the shortest distance between the projection points is calculated for collision checking. To obtain candidate trajectories, the trajectories are selected by collision, acceleration and curvature checking. Then, the cost function is constructed to evaluate the candidate trajectories and obtain the optimal motion trajectory. Finally, the simulation results in different driving scenarios show that the motion planning algorithm can make decisions about driving behavior in advance. The planned speed curve is more stable and the safety, comfort and driving efficiency of the motion trajectory are better.

Key words: autonomous driving; motion planning; interactive trajectory prediction; social force model; cost function

中图分类号: