This study investigated pedestrians’ crossing behaviour in a virtual reality environment. One aim was to develop a framework for evaluating external Human-Machine Interfaces (eHMI) used by automated vehicles for future studies. Pedestrians were provided with a series of two approaching cars, which were travelling at either 25mph, 30mph, or 35mph, with eight manipulated time gaps in between cars, where the second car either decelerated or kept pace. These stimuli were presented in 3 blocks. Pedestrians’ task was to cross (or not) naturally between the approaching cars. Data from decelerating trials were analysed. Results showed 51% of crossings happened before deceleration, 31% of crossings after the car had stopped and only 18% of the crossings during deceleration, leaving a great margin for evaluating the effect of eHMI and changing pedestrian crossing behaviour during deceleration. A learning effect was found, demonstrating a shift of decision making across blocks, whereby crossing increasingly occurred during the approaching vehicle’s deceleration, rather than after it had come to a full stop. Further analyses were conducted to investigate the effect of speed on initiation time, crossing time and safety margin. This study provides guidelines in choosing the appropriate time gaps and speeds that may influence pedestrians’ crossing decisions and behaviour while presenting different designs of eHMI in future studies. The results also provide guidance on how to evaluate safety, efficiency/receptivity and learning effects, when comparing different eHMI designs in VR experiments.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)