Cyclist and pedestrian trust in automated vehicles: An on-road and simulator trial

Abstract

Automated vehicles (AVs) need to be trusted by cyclists and pedestrians where they will share the road. To test trust, cyclists and pedestrians, and a comparison cohort of drivers, observed trials of both a road and simulator AV undertaking three common priority-based maneuvers: a right turn into a side road, overtaking a parked car, and passing over a pedestrian priority (zebra) crossing. The AV made the maneuvers either giving way, or not giving way, to a pedestrian or cyclist. One hundred and thirty-four participants aged 18 to 79 years were recruited based on being predominantly either a pedestrian, cyclist, or driver in their regular road use. For the on-road trials, the cyclist and pedestrian participants observed the AV maneuvers from the adjacent footway, and the driver participants observed from inside the AV. In the simulation environment, all participants were inside the automated vehicle. Trust scores were higher when participants observed a maneuver where the AV had to give way to a cyclist, and this can be linked with the re-assurance provided by the behavior of the AV in such an encounter. There was no significant difference in trust by road user type (cyclist, pedestrian or driver), age or driving experience, suggesting messaging to road users about the impacts of automated vehicles need not be differentiated by road user type. There was some evidence of differences in trust, especially for more complex maneuvers, between the on-road trial and the simulator, suggesting a need for caution in reliance on simulation-only experiments.