An eHMI Presenting Request-to-Intervene and Takeover Status of Level 3 Automated Vehicles to Support Surrounding Traffic Safety
URL: https://arxiv.org/abs/2605.00377v1
archive: archived pipeline: cataloged
Summary
Driving-simulator study testing an external HMI ('eHMI C+O') that displays Level-3 AV takeover status to surrounding manually-driven vehicle (MV) drivers via cyan and orange light bars indicating Request-to-Intervene (RtI) and takeover progress. Forty MV-driver participants followed a Level-3 AV across conditions (no-eHMI, ADS-status-only, eHMI C+O) in a within-subject design, including normal takeover and AV-accident scenarios. Compared with no-eHMI and ADS-status-only baselines, eHMI C+O improved understanding of AV intent, prediction of AV behavior, and perceived information sufficiency; reduced hesitation; increased confidence; and produced earlier and larger increases in time headway after RtI. In the AV-accident scenario the eHMI cut following-MV accident odds by 76.8% versus no-eHMI. Exploratory path analysis links the safety benefit to improved situation awareness and earlier defensive responses.
Key finding
Externalizing Level-3 AV takeover status to surrounding drivers via a cyan/orange eHMI yielded a 76.8% reduction in following-MV accident odds in an AV-accident scenario and produced earlier, larger increases in time headway after the RtI, mediated by improved situation awareness.
Methodology
Driving-simulator experiment, within-subject design. 40 MV-driver participants (ages 22–24) followed a Level-3 AV across three eHMI conditions (no-eHMI, ADS-status-only, eHMI C+O) crossed with takeover scenarios (normal takeover, AV-accident). A priori power analysis indicated N=28 sufficient for the 2×3 within-subject design. Three 55-inch 1920×1080 displays presented the simulated environment; participants drove an automatic-transmission MV. Mixed-effects ART ANOVAs and logistic models analyzed questionnaire ratings, time-headway AUC over 5 s and 10 s post-RtI windows, and accident odds; exploratory path analysis tested mediation through situation awareness.
Sample size: N=40 MV-driver participants (ages 22–24)
Quality score: 5 / 5