From Awareness to Intent: Mitigating Silent Driving System Failures through Prospective Situation Awareness Enhancing Interfaces
URL: https://arxiv.org/abs/2604.18449v1
archive: archived pipeline: cataloged verified
Summary
Multimodal driving-simulator study (Wang, Yan, Yang, He, Wang, Liu, Chen, Wang, He; HKUST-Guangzhou + NUS; arXiv:2604.18449) on supporting takeover from silent failures in SAE Level-2 automation, where the ADS continues operating but its internal world model is misaligned with reality and no TOR is issued. Forty-eight participants drove an AR-HUD-equipped simulator under different Prospective Situation Awareness Enhancement (PSAE) interfaces — varying perceptual cues (what the system 'sees') and intent cues (planned maneuver) — crossed with lighting (day/night) and hazard visibility (visible/invisible). Behavioral, subjective (SA, trust, perceived safety), EEG (ERSP), and EMG measures were collected.
Key finding
PSAE interfaces improved takeover performance via situational awareness as a mediator; perceptual cues were most effective at enhancing SA, while intent cues better supported trust calibration; an EEG ERSP correlate of SA was identified, supporting transparency-oriented HMI design for silent failures.
Methodology
Driving-simulator within-subjects experiment with 48 participants, AR head-up-display delivery of PSAE conditions (perceptual vs. intent cues), under crossed lighting (day/night) and hazard visibility (visible/invisible) conditions; measures: takeover performance (behavioral), SA, perceived safety, trust (subjective), EEG event-related spectral perturbations and EMG motor-preparation timing.
Sample size: N=48 participants
Quality score: 5 / 5