摘要
传感器感知盲区是造成智能汽车交通事故的主要原因之一。为了降低传感器感知盲区对智能汽车主动安全性能的影响,对传感器感知盲区条件下的智能汽车主动制动系统控制进行了研究。首先,建立感知盲区数据库,并搭建卷积神经网络对其进行识别;其次,根据其运动特征进行分类,建立感知盲区条件下的安全距离模型;最后,基于上述安全距离模型对感知盲区内的潜在障碍物进行自车速度控制,达到主动避撞的目的。仿真和实车试验表明,提出的传感器感知盲区分类可以较好地表述感知盲区的运动特征,传感器感知盲区条件下的主动避撞安全距离模型对潜在障碍物有较好的预防作用,主动避撞算法提高了智能汽车在传感器感知盲区内的主动安全性能。
Sensor occluded scenes is one of the main causes of intelligent vehicle traffic accidents.In order to reduce the impact of sensor occluded scenes on active safety performance of intelligent cars,the control of intelligent car active braking system under sensor occluded scenes conditions was studied.Firstly,a database and a convolution neural network were specially established to recognize sensor occluded scenes.Secondly,different types of sensor occluded scenes were classified according to their motion characteristics.Then,a safe distance model under the condition of sensor occluded scenes was established.Finally,based on the above-mentioned safe distance model,the speed of potential obstacles in sensor occluded scenes were controlled to achieve the purpose of active collision avoidance.According to simulation test and real vehicle test,the sensor occluded scenes classification proposed can better express the motion characteristics of potential obstacles in sensor occluded scenes.The active collision avoidance safety distance model under sensor occluded scenes conditions had good preventive effects on potential obstacles,and the active safety performance of intelligent vehicle in sensor occluded scenes was improved.
作者
袁朝春
王桐
何友国
SHEN Jie
陈龙
翁烁丰
YUAN Chaochun;WANG Tong;HE Youguo;SHEN Jie;CHEN Long;WENG Shuofeng(Automotive Engineering Research Institute,Jiangsu University,Zhenjiang 212013,China;University of Michigan-Dearborn,Dearborn MI 48128,USA)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2020年第2期363-373,共11页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(51775247、51305167)
江苏省普通高校研究生科研创新计划项目(KYCX18_2230)
关键词
智能汽车
传感器感知盲区分类
潜在障碍物
卷积神经网络
安全距离模型
intelligent vehicle
classification of sensor occluded scenes
potential obstacles
convolution neural networks
safety distance model
