STUDY 01 / 02

INDEPENDENT STUDY · LH STUDIO · AD + ADAS HMI

Calm

Handover

.

ADAS awareness interactions — when, where and how a car

that's driving itself should ask for the driver's attention, and

when it shouldn't.

TYPE

ROLE

TIMELINE

DELIVERABLES

Self-initiated concept study

Designer — research, interaction design,

prototyping

2025

Interaction spec, HTML prototype,

takeover film, validation plan

QUESTION

FIVE STATES

ESCALATION

MONITORING

VALIDATION

Never a surprise takeover.

From pilot driving to you're driving.

Calm first, urgent only if needed.

Where, never whether.

How I'd test it.

01 A car that drives itself will eventually need the driver

back. The risk isn't the request — it's a request the

driver didn't see coming.

QUESTION

I set one concrete scenario. A Level 3 Highway Pilot is driving a mapped divided highway at 62

mph. The driver is watching media on the centre screen. A construction zone ahead ends the

lane markings, so the pilot has to hand back control — within 10 seconds.

DECIDED

No request without a reason the driver can see

DECIDED

Advisories before any request

DECIDED

Media pauses the moment a takeover starts

DECIDED

The pilot never re-engages by itself

02 The handover is five states and one fallback. Each

defines when it starts and ends, what every channel

shows, and the edge cases that change it.

FIVE STATES

Pilot driving

Situation awareness

Extended awareness

You can look away. You'll get a heads-up before you need to drive.

Construction zone ahead. Pilot continues. No action needed yet.

Radar and lidar see a work truck you can't see yet. Slowing to 45 mph.

Teal band, steady wheel light, no sound. Media allowed.

One soft chime. Two advisories within 10 s merge into one

banner — chimes never stack.

Explains what the car sees and what it's doing. Low sensor

confidence skips straight to F4 — never show an unsure

hazard as certain.

FALLBACK

Takeover request

Minimal risk maneuver

Manual driving

Take over.

Driver not responding. Pulling over.

You're driving.

Full-width request with the reason and a 10 s countdown.

Ends only with eyes on road and hands on wheel. The

countdown never resets.

After 10 s with no response: hazards on, the car stops in its

lane, doors unlock, emergency call offered.

Lane keeping and collision warning stay on. When the pilot is

available again, it waits to be turned on.

FIG. 02A — F1: THE PILOT IS DRIVING; THE DRIVER IS WATCHING A SHOW

FIG. 02B — F3: THE CAR SHOWS A TRUCK THE DRIVER CAN'T SEE YET

03 Each step adds a channel instead of replacing one —

so a driver who missed step one still gets everything

from step one later.

ESCALATION

STEP

TIME SINCE REQUEST

CLUSTER + HUD

SOUND

HAPTIC

WHEEL LIGHT

1 0–4 s

Amber "Take over" + reason + countdown

Two-tone chime, once

None

Slow amber pulse

2 4–7 s

Red "Take over now"

Urgent tone, 1 per second

Seat and wheel pulses

Fast red pulse

3 7–10 s

Red "Take the wheel now"

Spoken request + faster tone

Continuous seat and wheel

Fast red pulse

Stop

after 10 s

"Pulling over" + hazards

Spoken status, once

Short brake pulse

Solid red

FIG. 03 — THE ESCALATION LADDER, FROM THE TAKEOVER FILM

WHY THESE NUMBERS

WHY AMBER FIRST

10 seconds, split in three

Act now, calmly

The 10 s window follows the transition-demand approach in UN R157 (ALKS). The 4 s and 7 s

steps give a calm prompt first and haptics only when needed. They're hypotheses, not

validated values.

Starting red trains drivers to treat every request as an emergency. Amber says "act now,

calmly"; red is kept for a driver who hasn't reacted.

04 The driver-monitoring camera changes

appears — never

driver has.

whether

it appears, or how long the

where

an alert

MONITORING

That keeps the safety-critical path independent of an estimate that can be wrong.

GAZE ROUTING

CONFIRMATION

SPECULATIVE

Shown where you're looking

Eyes and hands, both

Comfort index

In F2 and F3, if the driver is looking at

the centre screen, the heads-up

appears there as well as in the cluster.

Eyes on road from the camera; hands

on wheel from the wheel's touch or

torque sensor. One without the other

keeps escalating.

A stress-adaptive layer may soften or

add cues, but never changes the 10 s

window or skips a step — until there's

evidence it predicts takeover quality.

FIG. 04 — F2 HEADS-UP, ROUTED TO THE CENTRE SCREEN WHERE THE DRIVER IS LOOKING

05 This is a concept, not a tested system. The next step

is a small simulator study to turn the timing from a

hypothesis into evidence.

VALIDATION

6–8 licensed drivers

ADAS USERS + NON-USERS

Watch a video while the pilot drives

3 RUNS, UNANNOUNCED

Amber-first vs. starting red

BETWEEN SESSIONS

Time to eyes on road, hands on wheel, first input

RAW TIMES

Can the driver say why the car asked?

AFTER EACH RUN

Success looks like:

every participant takes over before step 3, and every participant can

name the reason for the request.

NEXT STUDY — 02 / 02 · EXTENDS THE INFINITI QX WORK

Consent-first voice →

Next step.

10 s

TAKEOVER WINDOW - THREE CALM-TO-URGENT STEPS, THEN THE CAR STOPS ITSELF

OPEN QUESTIONS

What the study would answer

Does the F2 advisory shorten takeover time? Does gaze routing help or distract? Is 4s too long for a calm first step? And how should monitoring behave when the camera is blocked or the driver wears sunglasses?