Enhancing Pedestrian Hazard Awareness with Augmented Reality Warnings: Effects of Warning Variant, Cue Placement, and Directional Alignment

At occluded urban intersections, pedestrians have little time to localize out-of-sight vehicles under divided attention. In this situation, peripheral augmented reality (AR) warnings could help redirect attention toward hidden hazards. However, it remains unclear which visual forms and spatial presentations support rapid localization. To address this issue, we conducted two virtual reality studies using an occluded-intersection task with a secondary task. Study 1 tested the baseline benefit of AR warnings, and Study 2 manipulated warning variant, cue placement, and directional alignment. Results showed that peripheral warnings shortened reaction time and reduced collision rate, while secondary-task accuracy did not differ significantly. Within warning-present conditions, the tested gaze-locked/lower-eccentricity implementation yielded faster localization and lower early gaze entropy, whereas directional alignment yielded faster localization and earlier entry into the initially signaled sector. In the warning-variant comparison, the tested Gradient implementation produced slower responses and delayed sector entry, whereas Looming received the most favorable subjective evaluations. Overall, results support the tested gaze-locked/lower-eccentricity implementation and directional alignment for faster localization and identify Looming as the most favorably rated warning variant.