Running into Traffic: Investigating External Human-Machine Interfaces for Automated Vehicle-Runner Interaction

Paper Title

Running into Traffic: Investigating External Human-Machine Interfaces for Automated Vehicle-Runner Interaction

Publication Info

• Topic area: Human-computer interaction in automated vehicle-pedestrian communication.
• Keywords: Automated vehicles, external human-machine interfaces, pedestrians, runners, road safety, embodied cognition, augmented reality, usability, trust, inclusivity.

Background and Problem

• Problem / challenge: Existing eHMIs have been primarily designed for walking pedestrians and lack consideration for runners, who have distinct movement and perceptual demands. This raises doubts about the generalizability of walker-centric eHMIs to faster-paced pedestrians.
• Significance: Runners represent a significant subgroup of pedestrians, with over 621 million regular participants globally. Ensuring their safety and inclusivity in AV interactions is critical as AVs become more prevalent in urban traffic.
• Motivation and related work: Prior research has focused on eHMIs for walkers, cyclists, and drivers, but has overlooked runners. Studies have shown that runners face unique physical constraints and time pressures, which influence their interaction with interfaces. This paper addresses the gap by investigating eHMI usability for runners and comparing it to walkers.

Solution

• Proposed approach: Development of an augmented reality simulator (ARcade) to study AV-runner interactions using three eHMI conditions: LightRing (colour-changing lights), CyanBand (animated cyan lights), and No-eHMI (baseline).
• Novelty:
  1. The first user study to differentiate runners from walkers in AV interactions.
  2. Empirical evidence showing how movement patterns influence eHMI use across AV-yielding behaviours.
  3. Design guidelines for inclusive eHMIs accommodating diverse pedestrian movement behaviours.
  4. Introduction of DualBeam, a novel eHMI designed for inclusivity across pedestrian types.
• Procedure and key techniques: Participants navigated a virtual crossing while walking or running, encountering AVs with different eHMI conditions. Movement behaviours (speed, head rotations, stopping frequency) and perceptions (trust, safety, workload, usability) were measured using simulator logs and questionnaires.

Results

• Concrete findings:
  • Runners moved faster (mean speed: 2.62 m/s) than walkers (mean speed: 1.75 m/s) and relied more on eHMIs for rapid crossing decisions.
  • LightRing outperformed CyanBand and No-eHMI across trust (mean score: 4.26), perceived safety (mean score: 4.30), and usability (mean score: 0.94).
  • CyanBand was effective for walkers but hindered runners due to its directional animations requiring continuous monitoring.
  • No-eHMI consistently underperformed, leading to higher perceived risk (mean score: 3.09) and workload (mean score: 4.58).
• Advantage over baselines:
  • LightRing provided faster interactions and higher confidence in AV intentions compared to CyanBand and No-eHMI.
  • Colour-changing signals were more distinguishable and inclusive across walking and running activities.
• Experiments / evaluation:
  • 24 participants (mean age: 25 years) completed 12 trials each, testing all eHMI conditions while walking and running.
  • Measures included crossing speed, stopping frequency, head rotations, collision frequency, trust, safety, workload (NASA-TLX), and usability (UEQ-S).
• Limitations and future work:
  • AR simulation may not fully replicate real-world AV interactions; future studies should explore safe methods for testing non-yielding AVs with real vehicles.
  • Broader demographic studies are needed, including older adults, children, and individuals with accessibility needs.
  • Cross-cultural replications and testing with larger vehicles or multi-vehicle scenarios are recommended.

Summary

This study highlights the importance of designing eHMIs that accommodate the distinct movement behaviours and perceptual demands of walkers and runners. LightRing, with its colour-changing signals, emerged as the most inclusive and effective eHMI across both pedestrian types, while CyanBand showed limitations for runners due to its reliance on directional animations. The findings emphasize the need for explicit AV signals to ensure safe and informed crossing decisions, particularly for faster-paced pedestrians. Building on these insights, the paper introduces DualBeam, a novel eHMI designed for inclusivity across diverse road user types. These contributions support the safe integration of AVs into urban environments and promote active mobility.