Light it Up: Evaluating Versatile Autonomous Vehicle-Cyclist External Human-Machine Interfaces

Document Title

Light it Up: Evaluating Versatile Autonomous Vehicle-Cyclist External Human-Machine Interfaces

Document Information

• Subject Area: Research in human-computer interaction and autonomous driving technology, specifically focusing on the design and evaluation of interfaces for interaction between autonomous vehicles and cyclists.
• Keywords: Autonomous vehicle-cyclist interaction, external human-machine interfaces (eHMIs), cyclist behavior analysis, traffic scenarios, user experience design, red-green signals, virtual reality evaluation, iterative design, onboard display interfaces, road safety improvement

Research Background and Issues

• Issues or Challenges:

  • As autonomous vehicles gradually enter real-world roads, traditional social interaction signals provided by human-driven vehicles (e.g., eye contact, gestures) will disappear.
  • Current research on external human-machine interfaces (eHMIs) primarily focuses on vehicle-pedestrian interaction, without adequately addressing the needs of cyclists.
  • Cyclists face faster vehicle speeds and more complex positional relationships in diverse traffic scenarios, making spatial conflicts with vehicles harder to resolve.

• Importance:

  • Statistics show that the risk of cyclist-vehicle conflicts is high, with numerous accidents caused by misunderstandings of behavioral intentions on UK roads alone.
  • Ensuring cyclist safety and reducing traffic conflicts are critical tasks in developing autonomous driving traffic environments.

• Research Motivation and Related Work:

  • Autonomous vehicles need an alternative method to clearly communicate with cyclists, indicating the vehicle's intentions and attention to cyclists.
  • Existing studies have proposed early design concepts such as halos, roof-mounted emoji displays, and road projections, but their actual effectiveness and user experience have not been thoroughly tested.

Solution

• Methods or Solutions:

  • Employ a two-stage evaluation method to design, optimize, and test external human-machine interfaces (eHMIs) for interaction between autonomous vehicles and cyclists.
  • The first stage involves testing three interface designs using a virtual reality cycling simulator to analyze cyclists' perception and user experience of the signals.
  • The second stage uses the "Wizard-of-Oz" method for real-world validation, evaluating the practical implementation of optimized eHMIs in real scenarios.

• Innovations:

  • Multi-scenario Compatibility: Propose and optimize designs that adapt to various traffic scenarios, addressing uncertainty and enhancing safety.
  • Iterative Design and User Feedback: Integrate user interaction feedback into design improvements through a two-stage iterative design process.
  • Distinctiveness and Usability: Explore simple encoding methods based on red-green signals, combined with animations to enhance information presentation.

• Implementation Steps and Key Technologies:

  • Stage One: Use a VR cycling simulator to analyze interface usability and cycling behavior across five critical road scenarios (e.g., roundabouts, lane merging, bottlenecks).
  • Stage Two: Evaluate optimized interfaces in real-world tests using virtual "autonomous vehicles" to validate cyclists' perception of information in real environments.
  • Technologies include Meta Quest Pro virtual reality headsets for head tracking and gaze tracking, LED light strips, and display matrices for interface presentation.

Research Outcomes

• Specific Findings:

  • Cyclists preferred easily distinguishable red-green signals, which quickly conveyed whether the autonomous vehicle would yield or continue driving.
  • Cyclists found vehicle-wide signal displays to be more readable and trustworthy than single-location signals.
  • Successful eHMI implementation in real environments reduced shoulder-checking and increased cycling speed.

• Advantages Compared to Existing Solutions:

  • Systematic and user feedback-driven design makes current cyclist interface designs more aligned with practical needs.
  • Provides a universal design approach for different traffic environments and potential expansion across diverse user types.

• Experimental or Evaluation Results:

  • In VR simulator tests, the "Safe Zone" signal mode performed best, with the lowest workload and the highest participant confidence in vehicle awareness and intentions.
  • In real-world tests, cyclists achieved the fastest speeds under the "LightRing" condition and rated their sense of safety highest, though red-green signals remained a core element.

• Limitations and Future Directions:

  • Limitations:

    • Tests were limited to UK urban traffic environments, without studying compatibility across different countries and driving cultures.
    • Validation conditions did not involve complex multi-user scenarios, and vehicles consistently operated under yielding logic.
    • Results on long-term user learning of signals and high-speed vehicle testing remain unexplored.

  • Future Directions:

    • Expand evaluations of eHMI applicability to different countries and traffic scenarios.
    • Investigate interoperability of signals with other user groups (e.g., pedestrians, drivers).
    • Conduct long-term tracking studies in real large-scale traffic environments to observe dynamic changes in cyclist behavior and signal effectiveness.

This research is significant for advancing autonomous driving technology to improve road safety and provides theoretical support and practical validation for future diversified traffic interface designs.