An Eye Gaze Heatmap Analysis of Uncertainty Head-Up Display Designs for Conditional Automated Driving

Title of the Paper

An Eye Gaze Heatmap Analysis of Uncertainty Head-Up Display Designs for Conditional Automated Driving

Paper Information

• Research Area: Human-computer interaction and driver monitoring behavior in automated driving
• Keywords: Conditional automated driving, robot supervision, fallback readiness, task switching, non-driving-related activities, eye-tracking, driving simulation study, heatmap analysis, head-up display

Research Background and Problem

• In conditional automated driving (SAE Level 3), drivers often exhibit distracted visual attention during driving tasks and lack awareness of situational uncertainty, which may impair their takeover abilities and pose potential safety risks.
• A key research challenge is how to effectively communicate the automated system's uncertainty information (e.g., when the system may reach its functional limits and issue a takeover request) to drivers while balancing safety and user experience.
• This is particularly important because conditional automated vehicles require drivers to maintain "fallback readiness" to handle emergencies. However, the reduced demand for driving-related tasks leads drivers to engage in non-driving-related activities (NDRA), further diminishing their situational awareness and response capabilities.

Proposed Solution

• Proposed Methods
  • Three uncertainty-based intervention designs were proposed to support drivers' task allocation and situational awareness:
    1. T1-Display (Visual Display): Uses animated "guardian angel" visual elements to convey uncertainty information.
    2. T2-Interruption (Interruption): Temporarily interrupts the driver's NDRA when uncertainty increases.
    3. T3-Combination (Combination): Combines the strengths of T1 and T2 by incorporating visual elements and interrupting NDRA.
• Innovative Contributions
  • Compared the effects of these three interventions on eye movement patterns and behaviors (task switching and attention allocation) in a high-fidelity driving simulation experiment.
  • Investigated how the visualization of uncertainty and interruptions interact to influence user perception and behavior, as well as potential interactive effects.
• Implementation Techniques and Steps
  • Eye-tracking technology was used to record participants' attention distribution.
  • A realistic driving environment was reconstructed in an advanced driving simulator, with different levels of uncertainty in driving events designed and quantified.
  • Heatmap and statistical analysis methods were employed to observe user behavior in detail, and questionnaires were used to evaluate system usability and task performance.

Research Findings

• Specific Results
  • T2 and T3 (with interruption interventions) significantly improved users' monitoring of the driving environment, with a notable increase in eye-tracking frequency during high uncertainty.
  • The T3 combination intervention design elicited faster and more consistent task-switching behavior, with heatmaps showing more pronounced active monitoring of the driving environment.
  • T1 (uncertainty display only) may lead users to overly focus on NDRA, reducing overall situational awareness.
• Advantages Over Existing Solutions
  • The "monitoring echo" phenomenon triggered by the interruption strategy suggests it helps enhance task-switching behavior, potentially improving driving safety, rather than relying solely on uncertainty display designs.
  • T3 combines the "predictive function" of the display with the "enforcing function" of interruptions, making the system more user-guiding.
• Experimental and Evaluation Results
  • Analysis of eye-tracking behavior from 215 participants showed that T3 resulted in the longest monitoring behavior time, with NDRA engagement time effectively reduced.
  • Usability evaluations revealed no significant negative impact of T2 and T3 on user experience, indicating that users can accept partial intervention designs.
• Limitations and Future Directions
  • The experiment did not incorporate real driving motion environments, which may affect the applicability of the results to real-world scenarios.
  • The transparent overlay design on the HUD may make it difficult to distinguish whether participants were focusing on NDRA or the driving environment.
  • Future research should further investigate the long-term impact of interruption interventions on takeover performance and safety, as well as test the system's effectiveness in high cognitive load NDRA scenarios.