Analysis of car driver responses to avoid car-to-cyclist perpendicular collisions based on drive recorder data and driving simulator experiments.

The reduction of cyclist injuries and fatalities is one of the important issues in traffic safety. To establish countermeasures to avoid collisions, it is necessary to understand driver's responses in such critical situations. From videos of drive recorders, the behavior of cyclists and drivers that led to car-to-cyclist collisions can be analyzed objectively. In this research, the drivers' responses to avoid car-to-cyclist perpendicular collisions were examined using videos of drive recorders, and through using a driving simulator. First, drivers' responses to avoid collisions were compared between near-miss incidents and actual collisions using videos from drive recorders. In a statistical analysis of selected parameters of the drivers' responses, the average values of different parameters such as the time-to-collision (TTC), the deceleration at the time the cyclist was first visible to the driver, and the braking reaction time (BRT) were significantly different between near-miss incidents and collisions. Applying logistic regression to the near-miss and collision data revealed that the BRT at the time the cyclist was first visible had the largest influence on collision occurrences. The velocity-TTC, and the velocity-distance graphs at the time of braking-onset was plotted; the car deceleration of 5.2 m/s was the threshold which discriminated near-miss and collisions. Second, the responses of people to avoid collisions against cyclists were investigated in a driving simulator. Based on real-world collision videos from the drive recorders, two scenarios were reconstructed where the crossing cyclist appeared suddenly from behind view-blocking obstructions at an intersection. Scenario A has a relatively large TTC of 1.9 s, and scenario B had a small TTC of 0.5 s. The BRT was significantly smaller in the group that avoided collisions than in the group that collisions occurred in scenario A, whereas all parameters (TTC, BRT, and car deceleration) were all significantly smaller in the group that avoided collision in scenario B. The driver response parameters (TTC, BRT and car deceleration) were comparable between the drive recorder data and the driving simulator experiments. It was demonstrated that the BRT is the most important parameter toward avoiding collisions. Some drivers who accelerated the cars at the intersections had large BRT, and this led to collisions. Additionally, it was observed that swerving of cars without braking was not effective for collision avoidance.