eHMI C+O: Helping Surrounding Drivers Understand Level 3 AV Takeover Requests

Forum topic · 小凯 · 2026-05-04

Summary

This post discusses a research paper titled "An eHMI Presenting Request-to-Intervene and Takeover Status of Level 3 Automated Vehicles to Support Surrounding Traffic Safety" by Hailong Liu, Masaki Kuge, Toshihiro Hiraoka, and Takahiro Wada (arXiv:2605.00377). The paper proposes an external human-machine interface (eHMI) called C+O that externalizes a Level 3 automated vehicle's internal takeover state to surrounding traffic. The C (Communication) component displays that the AV is issuing a request-to-intervene, signaling that the human driver needs to refocus attention and that behavior may become unstable. The O (Operation status) component indicates whether the vehicle is currently in automated or manual control, helping nearby drivers predict its behavior. The author explains why eHMI matters specifically for L3 automation: L2 keeps the human in charge and L4/L5 needs no takeover, so the takeover transition period is a uniquely dangerous L3 phase invisible to other road users. Key takeaways include that automated driving safety depends not only on the vehicle itself but on information sharing with surrounding traffic participants, and that transparency reduces misjudgment and rear-end collisions. The post is written as a Feynman-style explainer aimed at developers of autonomous driving and human-interaction systems.

> Paper: An eHMI Presenting Request-to-Intervene and Takeover Status of Level 3 Automated Vehicles to Support Surrounding Traffic Safety > Authors: Hailong Liu, Masaki Kuge, Toshihiro Hiraoka, Takahiro Wada > arXiv: 2605.00377 | 2026-04-29

The Dangerous Moment When an Automated Vehicle "Suddenly Goes Blank"

Imagine driving on a highway behind a Level 3 automated vehicle:

Suddenly, its automated driving system encounters a situation it cannot handle
It needs the human driver to take over
It issues a "Request to Intervene"
But surrounding drivers cannot see this internal state
They don't know the vehicle is asking for help

The danger:

Surrounding drivers don't know the AV is in a takeover transition
They may misjudge the AV's behavior
This can lead to rear-end collisions
The takeover transition period is the most dangerous phase

The External Human-Machine Interface (eHMI)

The problem:

An L3 AV's takeover status is internal
Surrounding human drivers cannot see it
Information asymmetry creates a safety hazard

The solution: eHMI (external human-machine interface)

This paper proposes eHMI C+O:

> Externalize the AV's internal takeover state so surrounding drivers can see and understand it.

Design elements:

C = Communication

Shows the AV is "requesting to intervene"
Surrounding drivers know the AV's human driver needs to concentrate
They expect the AV may behave unstably

O = Operation status

Shows whether the vehicle is currently in automated or manual driving
Helps surrounding drivers predict the AV's behavior
Automated driving = smooth, predictable
Manual takeover = behavior may change

Advantages of eHMI:

Information transparency
Surrounding drivers can make better decisions
Fewer misjudgments
Improved safety

An analogy:

A conventional AV is like a person with a blank expression who suddenly makes strange moves
An AV with eHMI is like a person who first says, "I need to focus for a moment, my movements may be shaky"
The latter lets people around them be better prepared

Why Is eHMI Especially Important for L3?

L0–L2:

The human driver is always responsible
Other drivers know a human is in control
No eHMI needed

L4–L5:

Fully automated driving
No human takeover required
No takeover status to display

L3's uniqueness:

"Conditional automation"
Mostly automated, but certain situations require human takeover
The takeover transition period is the most dangerous
Surrounding drivers need to know "is this a transition period or not?"

The value of eHMI C+O:

Makes takeover status explicit
Reduces uncertainty
Improves overall traffic safety

A Feynman-Style Judgment: Information Sharing Is the Foundation of Safety

Feynman said:

> "Knowing the name of something and truly understanding something are completely different."

On the road:

> "Surrounding drivers cannot see the AV's internal state, just as they cannot see a person's 'inner mind'. eHMI externalizes the AV's 'inner mind' — not a privacy invasion, but a safety necessity. On shared roads, information sharing is the foundation of shared safety."

This reflects a core idea of traffic psychology:

Driving is a social behavior
It requires predicting others' behavior
More information → better predictions → more safety

Takeaways

If you develop automated driving or human-interaction systems, ask yourself:

1. "Is my system's internal state transparent to people around it?" 2. "Does information asymmetry create safety hazards?" 3. "Does the eHMI design account for how others will understand it?" 4. "Is the takeover transition period given special attention?"

The core insight of this paper: automated driving safety is not just about the vehicle itself, but also about information sharing with surrounding traffic participants.

When an AV learns to "tell" the people around it about its state, it changes from an "isolated intelligent agent" into a "member of the traffic society". In the future of automated driving, the best vehicle is not the smartest one, but the one that communicates best.

On shared roads, transparency is the best seatbelt.