Abstract
The deployment of Level 4-5 autonomous vehicles (AVs) in real traffic environments presents significant challenges for pedestrian interaction, primarily due to the absence of human driver cues. To address this limitation, risk-warning external human-machine interfaces (eHMIs), which communicate dynamic crossing risks based on the AV's time-to-arrival (TTA), have received growing attention. However, given the visual complexity of real-world traffic and the potential transition from partial to full eHMI adoption within vehicle platoons, the cognitive impacts of such systems on pedestrians, particularly in terms of mental workload (MW) and situation awareness (SA), require further investigation. This study investigates the impact of three eHMI penetration rates-zero (all vehicles without eHMI), partial (half with and half without), and full (all vehicles with eHMI)-within vehicle platoons on pedestrians' MW, SA, and gap acceptance. To simulate realistic traffic conditions, vehicle speeds (30 km/h and 36 km/h) and temporal gap sizes (ranging from 2 to 6 s in 0.5-s increments) were randomized. A video-based experiment was conducted with 24 participants who simulated street-crossing decisions by selecting appropriate temporal gaps when encountering vehicle platoons. SA was assessed using the Situation Awareness Rating Technique (SART), and MW was evaluated through a concurrent oddball task, with P300 event-related potentials (ERPs) serving as indicators of residual attentional capacity. Results showed that full eHMI penetration resulted in significantly higher SA and lower MW compared to partial penetration condition. When compared to zero penetration, full eHMI penetration enhanced pedestrians' sensitivity to gap size, indicating that as the gap size increased, the probability of pedestrians crossing increased more sharply. Notably, this effect occurred without significantly altering MW or SA. In contrast, partial penetration condition led to pedestrians crossing less frequently when encountering vehicles equipped with eHMI and was associated with significantly higher MW and lower SA. These findings suggest that risk-warning eHMIs enhance pedestrian-AV interactions effectively under full penetration conditions. However, partial penetration may impair pedestrians' decision-making and cognitive performance, highlighting the importance of strategic planning during the transitional deployment phase of eHMIs.