周车轨迹预测不确定性智能车避撞策略研究
Research on intelligent vehicle collision avoidance strategy based on uncertainty of surrounding vehicle trajectory prediction
陈龙 1王歆叶 1熊晓夏 2蔡英凤 1刘擎超 1王海2
作者信息
- 1. 江苏大学汽车工程研究院,江苏镇江 212013
- 2. 江苏大学汽车与交通工程学院,江苏镇江 212013
- 折叠
摘要
提出了一种基于周车轨迹预测不确定性的智能汽车避撞策略研究方法.轨迹预测模块,将基于物理的轨迹预测模型和数据驱动模型相结合构建物理引导的轨迹预测模型(PG-LSTM),模型输出关于周车预测轨迹的二维高斯分布参数,以表征驾驶员行为的不确定性;风险评估及避撞策略模块,结合轨迹预测模型的输出结果,提出一个新的风险度量——预测驾驶风险PDR和预测相对驾驶风险指数PRDRI作为评估未来风险的参考指标,建立紧急工况下避撞决策机制.通过Carsim搭建复杂紧急工况场景进行仿真实验.仿真结果表明:所提出的驾驶风险评估模型可以准确地辨识复杂行车场景未来驾驶风险,同时基于驾驶风险所提出的避撞决策机制能够提升智能汽车的避撞安全性.
Abstract
This paper proposes a research method for an intelligent vehicle collision avoidance strategy based on the uncertainty of trajectory prediction of surrounding vehicles.The trajectory prediction module combines physics-based trajectory prediction models with data-driven models to construct a physics-guided trajectory prediction model(PG-LSTM).The model outputs parameters of a two-dimensional Gaussian distribution for the predicted trajectories of surrounding vehicles to represent the uncertainty of drivers'behaviors.The risk assessment and collision avoidance strategy module,leveraging the output of the trajectory prediction model,introduces a new risk metric-Predictive Driving Risk(PDR)and Predictive Relative Driving Risk Index(PRDRI)as reference indicators for assessing future risks,establishing a collision avoidance decision-making mechanism for emergent situations.Complex emergency scenarios are simulated using Carsim.Our results indicate the proposed driving risk assessment model accurately identifies future driving risks in complex driving scenarios.Moreover,the collision avoidance decision mechanism based on driving risk enhances the collision avoidance safety of intelligent vehicles.
关键词
智能汽车/驾驶风险/轨迹预测/避撞策略Key words
intelligent vehicle/driving risk/trajectory prediction/collision avoidance strategy引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据