From Awareness to Intent: Mitigating Silent Driving System Failures through Prospective Situation Awareness Enhancing Interfaces

Paper Title

From Awareness to Intent: Mitigating Silent Driving System Failures through Prospective Situation Awareness Enhancing Interfaces

Publication Info

• Topic area: Enhancing driver performance and safety in silent failures of partially automated driving systems.
• Keywords: Silent failures, situation awareness, augmented reality, head-up display, automated driving systems, trust, physiological indicators, EEG, EMG, human-machine interaction.

Background and Problem

• Problem / challenge: Silent failures in partially automated driving systems (ADS) occur when the system fails to detect hazards without issuing a warning, leaving drivers unprepared to intervene. Existing research focuses on takeover requests (TORs) but neglects silent failure scenarios, particularly how to support drivers in such situations.
• Significance: Silent failures pose critical safety risks in SAE Level 2 vehicles, where drivers must remain vigilant and ready to intervene. Addressing these failures is essential for improving safety and trust in automated systems.
• Motivation and related work: Prior studies have explored explanations and transparency in ADS but assume accurate system perception. Few studies have investigated how perception and maneuver planning information influence driver performance during silent failures under varying environmental conditions.

Solution

• Proposed approach: Prospective Situation Awareness Enhancement (PSAE) interfaces delivered via augmented reality head-up displays (AR-HUDs) to improve drivers’ situation awareness (SA) and takeover performance during silent failures.
• Novelty:
  1. Quantitative modeling of PSAE impacts on drivers’ psychological and physiological states in silent failures.
  2. Examination of hierarchical pathways linking driver psychology, physiology, and takeover performance.
  3. Identification of neurophysiological correlates of SA and their implications for adaptive human-machine interfaces (HMIs).
• Procedure and key techniques:
  • Designed three PSAE interfaces: Environment Perception (EP), Planned Maneuver (PM), and Combined (EP+PM).
  • Conducted a driving simulator study with 48 participants under varying lighting conditions (day/night) and hazard visibility (visible/invisible).
  • Measured psychological states (SA, trust, perceived safety), physiological indicators (EEG, EMG), and takeover performance (success rate, lead time).
  • Analyzed data using mixed-effects regression and structural equation modeling (SEM) to identify causal pathways.

Results

• Concrete findings:
  • EP improved SA significantly compared to Baseline and EP+PM, while PM and EP+PM increased trust.
  • Higher SA was associated with earlier neural activation (alpha suppression) and improved takeover success and lead time.
  • Night-time conditions increased vigilance (higher SA, muscle preparation) but reduced trust.
• Advantage over baselines:
  • PSAE interfaces (EP, PM, EP+PM) enhanced SA and trust compared to Baseline, with EP showing the strongest effect on SA.
  • Path analysis revealed that PSAE effects on takeover performance were mediated through SA, not direct physiological changes.
• Experiments / evaluation:
  • Mixed 4×2×2 factorial design with PSAE type, lighting condition, and hazard visibility as factors.
  • Data collected via psychological surveys, EEG, EMG, and behavioral logs.
  • Regression and SEM analyses confirmed the mediating role of SA in PSAE effectiveness.
• Limitations and future work:
  • Small sample size per PSAE condition limits statistical power.
  • Simulator-based study may not fully replicate real-world driving conditions.
  • Future research should explore diverse silent failure scenarios, post-takeover driving quality, and adaptive HMIs using real-world validation.

Summary

This study investigates the role of Prospective Situation Awareness Enhancement (PSAE) interfaces in mitigating silent failures in SAE Level 2 automated driving systems. Results show that PSAE interfaces improve drivers’ situation awareness (SA), trust, and takeover performance, with SA acting as a key mediator. EP interfaces were most effective for SA, while PM and EP+PM increased trust. Neurophysiological findings suggest alpha suppression as a potential marker of SA. The study highlights the importance of clear, context-aware HMIs and provides insights for designing adaptive systems to enhance safety in automated driving.