A systematic human-Centered development and evaluation method of eHMI icons to enhance pedestrian experience and safety

External Human-Machine Interfaces (eHMIs) have emerged as a key focus in both the autonomous driving industry and research communities. However, current eHMI studies often face significant challenges, including the lack of comprehensive interaction scenarios, effective control baselines, and reliable subjective metrics for evaluating eHMI effectiveness. To address these issues, this study aims to systematically construct and validate a comprehensive framework for icon-based eHMIs, including identifying core interaction scenarios, designing and testing icon sets, and evaluating their comprehensibility and memorability compared to traffic signs. The study comprises three experiments. In Study 1, 13 core interaction scenarios were identified through literature review, expert interviews, and surveys. For each scenario, three sets of static icons were designed, and common traffic signs were selected as effective controls. Study 2 involved the development of a tool for measuring icon comprehensibility, where 39 eHMI icons were tested across 13 scenarios, including objective accuracy and subjective evaluation dimensions. The results indicated that 13 optimal icons were statistically comparable to traffic signs in objective measures, scored lower in subjective comprehension, but higher in intelligence-related dimensions, thus serving as a benchmark for eHMI icon comprehension. In Study 3, the memorability of the 39 eHMI icons was tested, with recall accuracy assessed after 20 min, 1 day, and 7 days. This study evaluated pedestrians' understanding and memory of the icons over time, providing insights into the long-term educational impact and communicative efficacy of the icon designs. Results showed that while time affected the recall of the 39 icons, it did not significantly impact the 13 optimal icons or traffic signs, though the recall performance of the optimal icons was consistently lower than that of traffic signs. In summary, this research establishes a comprehensive and systematic set of interaction scenarios and develops and validates a complete evaluation framework for eHMI icons. • Highlight 1: Comprehensive identification of key interaction scenarios this study systematically identifies 13 core human-vehicle interaction scenarios through literature review, expert interviews, and surveys, providing a structured foundation for eHMI design and evaluation • Highlight 2: Development and validation of eHMI icons A set of 39 eHMI icons was designed and assessed using traffic signs as a control baseline. The study establishes a rigorous evaluation framework integrating objective accuracy and subjective perception measures • Highlight 3: Comprehensibility assessment with objective and subjective metrics A novel measurement tool was developed to assess eHMI comprehensibility, demonstrating that optimal icons are comparable to traffic signs in objective understanding but evoke stronger emotional and intelligence-related perceptions • Highlight 4: Memorability analysis and long-term impact the study examines the recall performance of eHMI icons over 20 min, 1 day, and 7 days, revealing stable recall for the best-performing icons while highlighting the need for further optimization to enhance long-term retention