智能汽车行驶风险评估综述  

A Review of Driving Risk Assessment for Intelligent Vehicles

作　　者：沈传亮[1] 肖啸 童言 胡宏宇[1] Shen Chuanliang;Xiao Xiao;Tong Yan;Hu Hongyu(State Key Laboratory of Automotive Simulation and Control,Jilin University,Changchun 130022)

机构地区：[1]吉林大学汽车仿真与控制国家重点实验室,长春130022

出　　处：《汽车文摘》2024年第8期1-8,共8页Automotive Digest

摘　　要：智能汽车的驾驶环境具有高度不确定性和复杂性等特点,容易发生碰撞事故,导致乘员受伤或死亡。为了提高智能汽车的安全性,总结了当前用于评估其行驶风险的研究方法,包括基于确定性的方法、基于概率的方法和基于机器学习的方法。基于确定性的方法是一种传统的二分类预测方法,基于概率的方法可以对各种不确定性进行建模,基于机器学习的方法可以自动学习驾驶行为,从而更准确地评估行驶风险。未来的研究方向应该是结合3种方法的优点从而开发出更加安全、可靠的自动驾驶系统。The driving environment of intelligent vehicles often has high uncertainty and complexity,which can lead to accidents and injuries to passengers.In order to improve the safety of intelligent vehicles,three major research methods are currently used to evaluate driving risks,including deterministic methods,probabilistic methods,and machine learning methods.Deterministic methods are traditional binary prediction methods,probabilistic methods can model various uncertainty,and machine learning methods can automatically learn driving behavior,making more accurate assessments of the risk of driving.Future research should combine the advantages of the three approaches to develop safer and more reliable autonomous driving systems.

关 键 词：智能汽车 风险评估 贝叶斯网络 机器学习 

分 类 号：U461.91[机械工程—车辆工程]