Assessing the safety criticality of driver behavior toward cyclists at intersections

Objective: Detailed analyses of car-to-cyclist accidents show that drivers intending to turn right at T-junctions collide more often with cyclists crossing from the right side on the bicycle lane than drivers intending to turn left. This fact has led to numerous studies examining the behavior of drivers turning left and right. However, the most essential question still has not been sufficiently answered: is the behavior of drivers intending to turn right generally more safety critical than the behavior of those intending to turn left? The purpose of this article is to provide a method that allows to determine whether a driver’s behavior toward cyclists can retrospectively be assessed as critical or non-critical. Methods: Several theoretical considerations enriched by findings of experimental studies were employed to devise a multi-measure method. This method was applied to a dataset containing real-world approaching behavior of 48 drivers turning right and left at four T-junctions with different sight obstructions. For each driver a behavior-specific criticality was defined based on both, their driving and gaze behavior. Moreover, based on the behavior-specific criticality of each driver, the required field of view to see a cyclist from the right was defined and was set into relation with the available field of view of the T-junction. Results: The results show that only a small portion of the drivers within the dataset would have posed an actual risk to cyclists crossing from the right side. Those situations with a higher safety criticality did not only arise when drivers intended to turn right, but also left. Conclusion: Therefore, the analysis can only provide an explanation for the higher proportion of accidents between drivers turning right and cyclists crossing from the right side in certain situations. Further research, for example analyses of exposure data regarding the frequency of turning manoeuvers at T-junctions, is needed in order to explain the higher proportion of accidents between drivers turning right and cyclists crossing from the right side.# 研究目标 针对机动车与骑行者碰撞事故的详细分析显示，在T型路口（T-junction）右转的驾驶员，与在自行车道上从右侧横穿的骑行者发生碰撞的概率，高于左转的驾驶员。这一现象促使诸多研究围绕左转与右转驾驶员的行为展开探究。然而，最核心的问题仍未得到充分解答：右转驾驶员的行为是否普遍比左转驾驶员的行为具有更高的安全临界性？本文旨在提出一种方法，可用于回溯性评估驾驶员针对骑行者的行为是否属于临界或非临界范畴。 # 研究方法 本研究结合多项理论考量与实验研究结果，构建了一套多维度测量方法。该方法被应用于一组真实世界数据集，该数据集记录了48名驾驶员在4个存在不同视线遮挡（sight obstruction）的T型路口进行左转、右转时的趋近行为。针对每位驾驶员，研究基于其驾驶行为与注视（gaze）行为，定义了行为特异性临界性指标。此外，基于每位驾驶员的行为特异性临界性，研究定义了从右侧观测到骑行者所需的视野范围（field of view），并将其与T型路口的可用视野进行关联分析。 # 研究结果 结果显示，数据集中仅有少数驾驶员会对从右侧横穿的骑行者构成实际风险。安全临界性更高的场景不仅出现在驾驶员右转的情形中，同样也存在于左转场景。 # 研究结论 因此，本分析仅能解释特定场景下右转驾驶员与右侧横穿骑行者碰撞事故占比更高的现象。未来仍需开展进一步研究，例如针对T型路口转向操作频率的暴露数据（exposure data）分析，以全面解释右转驾驶员与右侧横穿骑行者碰撞事故占比更高的现象。