One Size Does Not Fit All: Designing and Evaluating Criticality-Adaptive Displays in Highly Automated Vehicles

Title of the Paper

One Size Does Not Fit All: Designing and Evaluating Criticality-Adaptive Displays in Highly Automated Vehicles

Paper Information

• Research Area: Human-Computer Interaction and User Interfaces for Autonomous Vehicles
• Keywords: Autonomous vehicles, adaptive displays, criticality levels, individual differences, traffic density, situational awareness, trust, usability

Research Background and Issues

• Problems or Challenges Identified by the Authors:

  1. Current research on highly automated vehicles (HAVs) primarily focuses on conditionally automated vehicles (Level 2-3). The display design for HAVs still lacks sufficient solutions for addressing criticality levels, i.e., distinguishing between critical and non-critical objects.
  2. Display designs may lead to visual clutter, reducing drivers' ability to identify important information.
  3. Individual differences (e.g., age, trust propensity) influence users' cognition and preferences regarding information displays, making "one-size-fits-all" solutions inadequate for diverse user needs.

• Significance: Balancing drivers' situational awareness (SA) and performance in non-driving-related tasks (NDRTs) is key to enhancing the driving experience in HAVs. Trust and display design are crucial for the widespread acceptance of autonomous vehicles.

• Motivation and Related Work: Providing personalized information displays tailored to user needs can not only improve driving safety but also optimize NDRT performance, thereby promoting the acceptance of autonomous driving technology. A literature review indicates that different types of displays and key factors (e.g., traffic density and interaction object categories) significantly affect user experience and cognition. However, current HAV display designs fail to adequately consider time sensitivity and dynamic changes in user needs.

Solution

• Proposed Solution: The authors designed three types of criticality-adaptive displays based on object criticality levels:

  1. IO Display: Highlights influential objects (IOs) and critical objects (COs).
  2. CO Display: Displays only critical objects.
  3. ICO Display: Gradually distinguishes between influential objects and critical objects.

• Innovations:

  1. Introduced the concept of criticality levels, categorizing traffic objects into non-critical, influential, and critical for display design.
  2. Proposed dynamic object display methods using color changes and temporal patterns to reduce visual clutter.
  3. Incorporated individual differences (e.g., age and trust propensity) into display evaluation to optimize user experience.

• Implementation Steps and Techniques:

  1. Display Design:
     • Object criticality measured by distance and operational signals.
     • Display format uses highlighted borders and color differentiation for different object categories (vehicles, static objects, VRUs).
     • Candidate designs selected through focus group discussions.
  2. Experiment Design:
     • Virtual driving videos simulate two traffic densities (low and high) and three interaction object categories (vehicles, static objects, VRUs).
     • Online survey experiment with 295 participants completing video viewing and auditory 1-Back task experiments.
  3. Data Analysis:
     • Linear mixed model analysis of the effects of display type, traffic density, interaction object category, and participants' age and trust propensity on trust, SA, NDRT performance, and usability ratings.

Research Outcomes

• Specific Findings:

  1. Trust propensity significantly influenced usability evaluations: participants with low trust propensity found ICO displays more useful, while those with high trust propensity preferred CO displays.
  2. Interaction with VRUs resulted in higher SA but poorer NDRT performance; interactions with static objects and vehicles showed better NDRT performance.
  3. Age and CO displays both led to slower NDRT reaction times, but older participants found the displays more helpful.

• Comparison with Existing Methods: The proposed ICO and CO displays better matched the needs of users with low and high trust propensity, respectively. The dynamic display design significantly reduced information redundancy issues.

• Experimental Evaluation Results:

  1. Trust propensity and age played important roles in display evaluations, with significant interactive effects on task performance and system evaluation.
  2. Displays supported drivers in identifying critical objects in complex traffic scenarios while alleviating cognitive load from other elements in new environments.

• Limitations and Future Directions:

  1. The experiment was conducted online, limiting its applicability to real driving scenarios. Further validation through driving simulators or field tests is needed.
  2. Scenario design could be expanded to include more complex traffic dynamics (e.g., animal interactions or vertical interactions).
  3. Future research could explore more advanced adaptive display solutions that dynamically respond to users' real-time trust and situational awareness levels.

Through this study, the authors proposed design principles for criticality-adaptive displays, offering new directions for user interface design in autonomous vehicles while optimizing the driving experience for drivers of different ages and trust levels.