Calibrating Pedestrians' Trust in Automated Vehicles: Does an Intent Display in an External HMI Support Trust Calibration and Safe Crossing Behavior?

Title of the Paper

Calibrating Pedestrians’ Trust in Automated Vehicles: Does an Intent Display in an External HMI Support Trust Calibration and Safe Crossing Behavior?

Paper Information

• Subject Area: Autonomous driving technology and human-computer interaction
• Keywords: Autonomous vehicles, automation trust, pedestrian behavior, external human-machine interface (eHMI), system transparency, status display, intent display, misuse, safety, trust calibration

Research Background and Issues

• Identified Problems or Challenges:

  • The widespread adoption of autonomous driving technology has transformed transportation systems, making interactions between pedestrians and autonomous vehicles an urgent challenge.
  • Current external human-machine interface (eHMI) designs typically provide only information about the autonomous driving mode status. Such designs may lead pedestrians to overtrust autonomous systems, potentially causing safety risks.
  • There is a lack of systematic research on how pedestrian trust is affected and restored when autonomous systems fail (e.g., when vehicles fail to yield to pedestrians as per traffic rules).

• Significance:

  • Pedestrian trust determines whether they engage in safe street-crossing behaviors, directly impacting the safety of interactions between pedestrians and vehicles in mixed traffic environments.
  • Addressing this issue is crucial for designing trust mechanisms in autonomous vehicles and ensuring future traffic safety.

• Research Motivation and Related Work:

  • The study aims to explore the impact of external HMI displays (status + intent vs. status-only) on pedestrian trust and crossing behavior in scenarios where autonomous systems fail.
  • Previous research has found that providing vehicle intent information via eHMI can enhance pedestrian trust, but detailed exploration of failure scenarios in autonomous vehicles has been lacking.

Solution

• Methods and Solutions:

  • An enhanced external human-machine interface (eHMI) was proposed: a “status + intent” display method. Status information is indicated by continuous lighting that shows the vehicle’s autonomous driving mode, while intent information is displayed via slow flashing blue-green lights above the windshield to indicate whether the vehicle plans to yield to pedestrians.
  • A simulated experiment was designed with pedestrians as participants to observe dynamic trust changes and crossing behavior.

• Innovations:

  • This study is the first to systematically explore the impact of external HMI design on pedestrian trust in failure scenarios (when vehicles fail to yield).
  • It provides an effective trust calibration mechanism by displaying vehicle yielding intent, enhancing system transparency, reducing pedestrian misuse, and improving traffic safety.

• Implementation Steps and Key Technologies:

  1. Experimental Design:
     • A video simulation experiment involving 67 participants who simulated street-crossing behavior.
     • Experimental groups were designed by crossing two eHMI designs (status-only eHMI, status + intent eHMI) with two system states (no failure, single failure).
  2. Data Collection:
     • Quantitative data included trust scores and crossing onset time (COT).
     • Qualitative data were gathered through semi-structured interviews with pedestrians post-experiment.
  3. Technical Architecture:
     • Videos were displayed in a highly immersive format, and pedestrian crossing behavior was measured using automated pressure sensors.
     • Real-time data recording was implemented using Arduino and control systems.

Research Findings

• Specific Findings:

  • Dynamic patterns of trust:
    • Trust gradually builds during early interactions without failures, leading to earlier street crossings (reduced COT).
    • Trust drops sharply when the system fails but recovers quickly during subsequent interactions without failures.
  • Advantages of the “status + intent” design:
    • Significantly reduces the decline in pedestrian trust, helping pedestrians detect failures and make safer decisions.
    • Experiments showed that in failure scenarios, the collision rate between pedestrians and vehicles was 31.3% for the status-only group, compared to just 5.9% for the status + intent group.

• Advantages Over Existing Solutions:

  • Enhances system transparency, reducing pedestrian overtrust and misuse (e.g., entering the street without observing vehicle dynamics).
  • Provides theoretical support, demonstrating that pedestrians’ dynamic trust mechanisms are similar to drivers’ trust in autonomous systems, which can be calibrated through transparent information.

• Limitations and Future Directions:

  • Limitations:
    • The experiment was conducted in a simulated environment rather than real traffic scenarios.
    • Participants were limited to employees from a specific company, restricting sample representativeness.
    • Complex traffic environments, such as scenarios with dense interactions between vehicles and pedestrians, were not explored.
  • Future Directions:
    • Extend research to long-term real traffic scenarios to observe the dynamic changes in trust over time.
    • Utilize technologies like eye-tracking to analyze pedestrian crossing decision-making processes in greater depth.
    • Investigate the impact of design variations (e.g., intent signal colors, flashing patterns) on pedestrian trust.

Practical Implications

• Proposes that adding “intent” signals significantly improves the effectiveness of eHMI design, promoting safer pedestrian crossing behavior.
• Suggests manufacturers combine educational activities to enhance public understanding of autonomous vehicles and the meaning of eHMI signals, reducing safety risks caused by overtrust.