Abstract: To help drivers correctly understand their safe-driving ability, and to reveal the 'driver-vehicle-road' interaction relationship and measure the impact of various factors on driving safety, this paper proposed an integrated risk field model considering driving style factors. In view of the insufficient description of human factors in the existing risk field model, the vehicle acceleration variance and steering wheel angle variance were transformed into driving style factors to characterize the potential driving habits. According to the sudden and short-term features of the accident risk, the Hilbert-Huang Transform (HHT) method was used to transform the driving entire-process risk into the risk signal energy of the stationary phase and the local mutation phase, which was the key basis for evaluating the intrinsic attributes of driving ability. The driving simulation experiment was carried out with pedestrians crossing the street. The results show that the driving risk signal energy presents an 'oval' circle distribution in the spatial form, and it is decreasing from the center of the physical contour to the edge. The signal energy climbs quickly from the view of risk evolution process. Risk signal energy clustering can obtain more detailed and quantifiable classification results than subjective questionnaire. By comparison, it is found that drivers have 'cognitive-manipulation' bias (questionnaire safe-model dangerous/tendency), and rich driving experience can compensate for weak safety awareness (questionnaire dangerous-model safe/tendency). The research results provide a new method for identifying dangerous driving groups and improving their safe-driving ability.

Key words: traffic engineering, integrated risk field, driving simulation, driving style, hilbert-huang transform (HHT)

摘要： 为帮助驾驶人正确认知自身的安全驾驶能力，揭示“人-车-路”相互作用关系并衡量各要 素对车辆驾驶安全性的影响，构建了考虑驾驶风格因子的综合风险场模型。首先，针对现有风险 场模型在人因要素刻画不足，将车辆加速度方差和方向盘转角方差转化为驾驶风格因子，表征驾 驶人潜在驾驶习惯；然后，根据事故风险的突发性和短时性特征，利用希尔伯特-黄变换(HilbertHuang Transform, HHT)将驾驶全域风险分解为平稳阶段和局部突变阶段风险信号能量，作为评 价驾驶能力内在属性的关键依据；最后，开展行人横穿风险场景的驾驶模拟试验验证。结果表 明：驾驶风险能量在空间形态上呈“椭圆形”圈层分布，沿物理轮廓中心高边缘低，风险演变过程 能量大小“爬山式”递增效应显著。对风险能量聚类可获得比主观问卷更加细致可量化的分类结 果，对比发现，驾驶人存在“认知-操纵”偏差(问卷安全型-模型危险型/危险倾向)同时丰富的驾驶 经验对薄弱的安全意识有补偿作用(问卷危险型-模型安全型/安全倾向)。研究成果可以为识别 危险驾驶人群，提高安全驾驶能力提供一种新方法。

关键词: 交通工程, 综合风险场, 驾驶模拟, 驾驶风格, HHT变换