evARything, evARywhere, all at once: Exploring Scalable Holistic Autonomous Vehicle-Cyclist Interfaces

Research Background and Issues

• Identified Problems or Challenges
  Interactions between bicycles and autonomous vehicles (AVs) on shared roads are highly complex, especially in the absence of human drivers, where traditional social signals (e.g., gestures or facial expressions) cannot be conveyed. Existing research primarily focuses on one-to-one AV-cyclist interactions, but there is a more urgent need to address scalable multi-AV-cyclist interactions, tackling issues of information overload and signal association. Using isolated user interfaces may lead to information overload, while a lack of clear signals could increase collision risks.

• Significance of the Problem
  From 2018 to 2022, there were 75,000 collision incidents between vehicles and cyclists in the UK alone. These spatial conflicts pose significant threats to cyclist safety. In multi-user interaction scenarios, information confusion or signal association issues could further increase the likelihood of accidents. With the proliferation of autonomous driving technology, developing effective interaction interfaces to ensure cyclist safety has become a priority.

• Research Motivation and Related Work
  Traditional AV-cyclist interfaces mostly rely on single external devices (e.g., eHMIs) or AR glasses worn by cyclists. These studies fail to address the scalability of signals in multi-AV, multi-cyclist scenarios. Holistic AV-Cyclist Interfaces (HACIs) propose the concept of creating an ecosystem through multimodal interconnected interfaces, aiming to enhance information scalability and reduce user confusion. However, the practical implementation of these concepts in multi-AV scenarios remains to be empirically validated.

Solution

• Proposed Solution
  The authors developed CycleARcade, an augmented reality platform for designing and evaluating HACIs in complex scenarios. This study was conducted in two parts:

  1. Design Work: Using a participatory design approach, the authors collaborated with cyclists and HCI researchers to construct and iterate HACIs.
  2. Experimental Evaluation: Testing the designed HACIs and comparing their effectiveness to identify best practices for multi-AV information presentation.

• Innovations in the Solution

  • CycleARcade is an augmented reality platform that supports immersive multi-user experiences, allowing users to design and experience high-fidelity interfaces in real time.
  • After experiencing the real complexity of multi-AV scenarios, participants iteratively designed interfaces directly in the AR environment, resulting in more practical designs tailored to cyclists' actual needs.
  • Tight integration of multimodal signals: Visual, auditory, and tactile signals complement each other to convey precise multi-AV information.

• Implementation Steps and Key Technologies

  • Step 1: Develop three different types of HACIs using CycleARcade: RoadAlert (road-enhanced display + audio), reARview (AR rearview display), and Gem (central handlebar display + vibration alerts).
  • Step 2: Systematically evaluate the performance of each design through user experiments, collecting subjective feedback and behavioral data (e.g., shoulder checks and collision rates).
  • Key Technologies:
    • Use of Meta Quest 3 headsets and an augmented reality environment built with Unity.
    • Dynamic synchronization and validation of multimodal signals (spatial audio and visual enhancements).

Research Outcomes

• Specific Findings

  • Cyclists showed a significant preference for interfaces that allow them to keep their attention on the road while conveying clear signals through visual and auditory means (e.g., RoadAlert).
  • Environmental AR features, such as road projections, effectively reduced cognitive load for cyclists, improving user experience and safety.
  • The importance of multimodal signals, especially spatial audio, was reaffirmed, as they enhance situational awareness and reduce the frequency of shoulder checks.

• Advantages Over Existing Solutions

  • In complex multi-AV scenarios, HACIs centralize information presentation and reduce redundant signals, effectively improving cyclists' response efficiency and safety.
  • Compared to traditional single-device solutions, multimodal interfaces provide richer and complementary information, reducing confusion and anxiety for cyclists.

• Experimental or Evaluation Results

  • Quantitative data showed that RoadAlert significantly reduced the average frequency of shoulder checks (notably lower than reARview and Gem).
  • RoadAlert scored highest in perception performance, safety, and user experience, while Gem's central display increased workload and performed less effectively.
  • For AVs outside the line of sight, spatial audio demonstrated significant advantages, effectively compensating for the limitations of visual signals.

• Limitations and Future Directions

  • Limitations:
    1. The current study only addresses single-cyclist interactions with multiple AVs and does not extend to complex multi-cyclist, multi-AV interactions.
    2. The sample was limited to participants and traffic infrastructure in the UK, lacking cross-cultural applicability analysis.
  • Future Directions:
    1. Explore design methods for signal association and scalability in many-cyclist-to-many-AV interaction scenarios.
    2. Validate the performance of HACIs in other traffic environments and climatic conditions to ensure broad applicability.
    3. Develop more dynamic interfaces to adapt to changes in AV behavior or complex multi-scenario conditions.

Conclusion

This study addresses the challenge of scaling interaction interfaces between AVs and cyclists in multi-vehicle scenarios by designing and evaluating three novel interfaces using CycleARcade. The research demonstrates that multimodal signal designs combining visual enhancements and spatial audio significantly improve user experience, safety, and trust. These findings provide a critical foundation for the application of HACIs in real-world traffic, while highlighting potential future research directions, including multi-cyclist interactions and global applicability expansion.