Risks that induce bicycle accidents: Measurement and analysis of bicyclist behavior while going straight and turning...

In 2021, 69 % of those who died in Japan while riding a bicycle were of an older age. Preventing such accidents is an important issue because older bicyclists are more likely to be fatally injured in an accident. We developed a bicycle simulator that can safely simulate dangerous traffic environments by combining VR technology, motion capture technology, and various sensors to investigate the driving behavior of bicyclists who are involved in accidents. The simulator measures operating data such as floor reaction force, wheel rotation angle, brake lever displacement, position and posture of motion capture devices, and position and posture of the bicycle and vehicles at 60 Hz. We projected images of the virtual traffic environment to reflect the driving behavior in real-time onto three screens installed at the front, right side, and rear of the bicycle using three projectors. The constructed simulator was implemented for a "crossing inspection" where a rider checks for safety in front and behind while riding down a straight road with one lane and crossing this roadway and compared this to a "tracing inspection" where a model route is traced using a previously measured safe route taken by a younger rider. We measured and analyzed the bicyclist behaviors of nine younger and six older participants during two types of inspections as a usability study. In the crossing inspection, the crash rate for the older riders in the near lane was 5.3 times higher than that for the young. Riders collided with vehicles exhibited minimal rearward checking angles before entering a roadway, extended durations from the end of the last rearward check to roadway entry, and considerable standard deviations during road crossings. Moreover, in the tracing inspection, younger participants demonstrated superior trajectory passage rates and maintained closer proximity to the model, whereas older participants manifested diminished trajectory passage rates and remained at more extended distances from the model. Notably, a declining trajectory passage rate corresponded to an increase in the standard deviation of the crossing route and the time taken to enter the roadway.