WorksAbout
Considerations for designing eHMI for Autonomous Vehicles: Perspectives and Color Coding for Pedestrian-AV Interaction




Overview
This research explores the design considerations for external Human-Machine Interfaces(eHMI) in autonomous vehicles, focusing on communication with pedestrians. It analyzes the effectiveness of message perspectives—egocentric versus allocentric—and color coding in conveying safe crossing information. The findings highlight the importance of pedestrian-centric messages and standardized color usage to ensure clarity and safety in pedestrian-vehicle interactions.



Project TypeAcademic Project  |  Design Research Methodologies


DurationApr  -  Jul 2024


CollaborationSolo


DownloadOpen PDF








Problem Statement and Research Question
“How could we effectively convey messages via the eHMI of AV to facilitate communication between pedestrians and AV?”

In the future with LV5 autonomous driving, drivers will be unnecessary, eliminating direct interaction between pedestrians and drivers.
As a result of this transition, eHMI will replace this interaction, taking on the role of drivers.











Research Objective

(1) Perspective of the Messages
Prepared 4 messages for each perspective, for crossing and not crossing the street,
resulting in a total of 8 messages.





(2) Color Coding of the Messages
Provided messages with four different colors (white, red, green, yellow),
which are commonly using in traffic environments.














Process

This research was proceeded with sequential approach to figure out more focused and tailored results,
starting with (1) the perspective of the messages, and then (2) the color coding of messages.
Both distributed to respondents through various channels, including social media platforms and online communities.












(1) Perspective of the Messages



Phase 1: Comparing within each perspective
Purpose
Figure out which messages are effectively deliverable within each perspective

Duration
28 May ~ 4 June (7days)

Methods
Text-based quantitative and qualitative

Survey Design
(8 Questions)
1. Choose more suitable message for each situation
- Deciding to cross the road: "WALK" vs. "GO" / “BRAKING” vs. “STOPPING”
- Deciding not to cross the road: "DON’T WALK" vs. "STOP" / “DRIVING” vs. “MOVING”

2. Select the reasons for their choices within 6 options which are made based on three criteria (Clarity, Intuitiveness, Context Appropriateness). (Multiple-choice Format)



Results of Phase 1
This phase included 48 respondents, with the majority (85.4%) between 18-34 years old.

For deciding to cross the walk, respondents answered that “WALK” and “STOPPING” are clearer and more intuitive.
For deciding not to cross the walk, they reported that “DON’T WALK” and “DRIVING” are clearer and more contextually appropriate.

Additionally, these results indicate that respondents prioritized clarity and intuitiveness for deciding to cross the road, and context appropriateness for deciding not to cross the road.






Phase 2: Comparing the perpectives based on the Phase 1

Purpose
Comparing the most articulate messages from each perspective identified in the Phase 1

Duration
7 June ~ 11 June (5days)

Methods
Text-based quantitative

Survey Design
(2 Questions)
1. Choose more suitable message for each situation (Forced-choice Format)
- Deciding to cross the road: "WALK"(egocentric) vs. "STOPPING"(allocentric)
- Deciding not to cross the road: "DON’T WALK"(egocentric) vs. “MOVING”(allocentric)



Results of Phase 2
This phase included 62 respondents, with the majority (93.5%) between 18-34 years old.

For deciding to cross the walk, 74.2% chose “WALK” as more proper for deciding to cross the road.
For deciding not to cross the walk, 71% chose “DON’T WALK” as clearer for deciding not to cross the road.

In conclusion, egocentric messages from the pedestrian’s perspective(”WALK”, ”DON’T WALK”) are more effectively deliverable than allocentric messages from the AV’s perspective(”DRIVING”, ”STOPPING”).









(2) Color Coding of the Messages



Purpose
Determine which colors with messages are most effectively conveyed the intended information

Duration
14 June ~ 18 June (5days)

Methods
Visual-based quantitative and qualitative

Survey Design
(4 Questions)
1. Choose which color with text was the most accurate for indicating situations when it is safe to cross the street and when it is not. (Forced-choice Format)
Top: Deciding to cross the road / Bottom: Deciding not to cross the road

2. Select the reasons for their choices within 3 options which are made based on formal acknowledgements and research. (Forced-choice Format): WHITE is perceived as neutral and clean and has high visibility(Lumen Learning, n.d.; Hu et al., 2023); RED is often associated with urgency and aggression, linked to its use in warning and alert signals in traffic lights and stop signs(Díaz-Román et al., 2015); GREEN is generally associated with positive actions, representing safety and permission to proceed(Proactive Creative, n.d.).; YELLOW is highly visible and easily noticeable, commonly used in traffic signs and signals to capture attention and convey caution(Lumen Learning, n.d.; LibreTexts, n.d.).





Results
This phase included 72 respondents, with the majority (95.8%) between 18-34 years old.

For deciding to cross the walk, 93.1% indicated that “WALK” with the “GREEN” color is appropriate.
For deciding not to cross the walk, 90.2% reported that “DON’T WALK” with the “RED” color is suitable

The main reason for these choices was the common understanding of traffic lights; specifically, crossing the street on a green light and stopping on a red light.











Conclusions


(1) Perspective of the Messages: The egocentric messages from the pedestrian’s point of view(“WALK” and “DON’T WALK”) are more effective than the allocentric messages from the AV’s point of view for indicating whether to cross the road.

(2) Color Coding of the Messages: GREEN color coding for messages on crossing the street and RED color coding for messages on not crossing the street are appropriate.









Limitation


1. Demographic LimitationsMost respondents were 20s, and the sample size was small, which might affect the results.
Future studies should include a broader age range and a larger sample size to enhance the reliability.

2. Cultural Variability in Traffic PerceptionThe study did not consider differences in traffic environments across various countries.
Therefore, it will be necessary to consider country-specific traffic perceptions.

3. Lack of Real-World TestingThis study was not tested in actual traffic environments, so further testing is required before direct application to actual eHMI.







▶︎  This page contains only the summary. You can view the full paper, including the literature review from here.


















© 2024. Jihyun Lee. All Rights Reserved.