基于自然驾驶数据的跟车场景潜在危险估计模型

Potential risk assessment model in car following based on naturalistic driving data

为了改善主动安全系统的性能,提出一种跟车场景潜在危险估计模型.将前车制动时为避免碰撞留给本车制动之前的时间作为跟车场景潜在危险估计指标,即时间裕度.将TTC(time to collision)定义为明显危险估计指标.使用自然驾驶数据来确定时间裕度中的目标车和本车的减速度.为了得到真实可信的驾驶员制动行为特性,讨论了驾驶员制动行为的收敛性.使用核密度估计来描述驾驶员的纵向加速行为,使用相对熵来表征2个不同数据集之间的差异.使用稳定收敛的数据集提取了驾驶员制动行为的特征参数,并根据这些特征参数得到了考虑驾驶员制动极限的时间裕度.最后使用跟车危险工况制动开始时刻的时间裕度构建了跟车场景潜在危险估计模型.分析稳定跟车工况中的明显危险和潜在危险表明,潜在危险估计模型可以描述两车相对速度较小情况的危险等级.真实跟车危险工况验证发现,使用潜在危险估计模型可以更早地发现危险发生的可能.

A potential risk assessment model is proposed to improve the performance of the active safety system.The time before the driver of the host vehicle has to brake to avoid the collision when the target vehicle brakes is defined as a potential risk measure,i.e.time margin.Whereas time to collision (TTC) is defined as an obvious risk assessment measure.The naturalistic driving data was employed to determine the brake deceleration of the host vehicle and target vehicle in the time margin.To obtain the authentic braking behavior characteristics,the convergence of the braking behavior of the driver was discussed.The kernel density estimation was used to describe the accelerating behavior of the driver.And the relative entropy (Kullback-Leibler divergence)was applied to represent the distinctions between two different datasets.The braking behavior characteristic parameters were extracted by using the dataset from which a convergent braking behavior can be obtained.And,the time margin considering the limitation of the driver was obtained based on the braking behavior of the driver.Finally,the time margins at the braking starting time in the dangerous car following scenarios were employed to construct a potential risk assessment model.Analyses of the potential risk and obvious risk in the steady car following scenarios show that the potential risk assessment model can evaluate the risk level when the relative speed between the host vehicle and the target vehicle is small.Verifications based on the dangerous car following scenarios reveal that the potential risk assessment model can help to detect the possibility of the danger earlier.