面向多障碍物场景的车辆紧急避撞耦合决策与轨迹规划方法

打开文本图片集

1²，²1²，¹，1；2.智能网联汽车与动力系统江西省重点实验室；3.厦门理工学院机械与汽车工程学院）

关键词：汽车主动安全；复杂场景；紧急避撞决策；轨迹规划与跟踪；深度学习框架PyTorch 中图分类号：TP273；U461.91 文献标识码：A DOI: 10.3969/j.issn.1674-8484.2025.06.014

A coupled decision-making and trajectory planning approach for vehicle emergency collision avoidance in multi-obstacle scenarios

GUAN Yongxue1²,LIU Senhai1,²,HAN Yong³,XULi1²,SHUWeibin1,FANChenxu*1

(1.ProductR&AcademyJanglingotorsCo.td.,Nnchang33oo,China;.JangxiProvincialKey LaboratoryofIntelligentConnectedNew Energy VehiclesandPowerSystems,Nanchang33oo,China; 3.SchoolofMechanicalandAutomotiveEngineering,Xiamen UniversityofTechnology,Xiamen36loo,China)

Abstract:An integrated framework coupling decision-making with trajectory planning was proposed to enhance the emergency colisionavoidancecapabilityof vehicles in high-speed multi-bstacle scenarios and address the chalenge of real-timeresponsiveness indecision-makingand planning due to computational complexity.The high-dimensional game problem was simplied intoasequence of single-obstacle interaction processes byestablishinga multi-vehicle non-cooperative game modeltodescribedynamic interactionsand designingasequential decision-making mechanism based on threat assessment.A graphics procesing unit (GPU)-accelerated trajectoryoptimizationalgorithmwas implemented using the opensource machine learningframework PyTorch,generating safeand comfortablecolisionavoidance trajectories whilesatisfying vehicledynamic constraints.The results show that the average decision-making computation time of the proposed method in typical high-speed scenarios is 20～50ms ,and trajectory planning takes 33.1~149.1ms, outperforming traditional model predictive control(MPC)methods.The lateral velocityandaccelerationof the planned trajectories are controlled within 4.0m/s and 4.0m/s2 ,respectively,meetingsafetyand comfort requirements.When tracking theplanned trajectories,themaximum lateral tracking errorand speed error are 0.22m and 0.59m/s ,respectively,fulfiling therequirements for high-speed emergency collisionavoidance.In CARLA simulations,successfulcollsionavoidance isachieved inallscenarios.The conclusiondemonstrates thattheproposed framework effectively balances decision-making optimalityand real-time performance, providing a reliable solution for vehicle active safety in complex scenarios.

Key words:automotive active safety; complex scenarios; emergency colision avoidance decision-making; trajectory planning and tracking; open source machine learning framework PyTorch

车辆主动安全技术作为高级驾驶辅助系统的主要开发对象之一，能够在车辆遇到危险场景时进行主动紧急避撞，避撞形式包括车辆的纵向紧急制动系统(autonomousemergencybraking，AEB)和横向紧急转向系统(automatic emergency steering，AES)[1-2]。（剩余21224字）