Back
HarmonyOS Intelligent Cockpit — multimodal AI interface

The Car
That Knows You.

Concept direction for Huawei's HarmonyOS intelligent cockpit and DTOF parking system — where AI meets every seat, every voice, every obstacle.

Category
Automotive HMI / Cabin Experience Design
Client
Huawei Intelligent Automotive (HIMA)
Vehicle
Maextro S800 — Ultra-Luxury Flagship Sedan
Role
Creative Director — Concept & UX Storytelling
Skills
Multimodal HMI, IVI Narrative, ADAS Visualization
Methods
User Scenario Design, Feature Storyboarding

A multimodal AI cockpit that recognizes every passenger by face and voice — personalizing content, audio zones, and ambient lighting per seat. Paired with a DTOF parking system that sees what no camera can: suspended pipes, sunken stairs, and obstacles invisible to traditional sensors.

"Most cars ask you to adapt to them. What if the car adapted to you — the moment you sat down?"

— Gordon Cheng, Creative Director
The Results

The Bestselling Luxury Sedan in China.

#1
Bestselling luxury sedan in China, January 2026
10,000
Units delivered within 7 months of launch
589K
Total HIMA alliance deliveries in 2025 (+32% YoY)
99.9%
Obstacle recognition rate (GOD network)
The Challenge

Making Intelligence Feel Invisible

Huawei's HarmonyOS cockpit had dozens of AI-powered features — face recognition, voice zones, memory-based personalization, ambient control. But advanced technology means nothing if users don't feel it. The challenge: how do you communicate invisible intelligence in a way that feels human, not robotic?

For DTOF parking: Huawei's Direct Time-of-Flight sensors could detect obstacles that traditional cameras and ultrasonics miss entirely — hanging pipes, negative-height curbs, suspended barriers. But no one had seen this capability demonstrated. The challenge: how do you make the invisible visible?

The Solution

Two Films. Two Dimensions of Trust.

I directed two concept films that translated Huawei's engineering capabilities into emotionally legible stories — one for the cockpit experience (human warmth), one for DTOF parking (technical confidence).

Cockpit — Multimodal AI

Every Seat Has a Voice. Every Voice Has a Memory.

MoLA Multimodal Perception

Face Recognition → Instant Personalization

The car identifies each passenger the moment they sit down. Seat position, mirror angle, climate, and content preferences load automatically.

Preference Engine

Personalized News & Content

The system learns your interests over time. Dad gets sports news. The car remembers — no setup, no asking.

Per-Seat Voice + Memory

The Child the Car Remembers

Daughter says "Tuantuan" — the assistant recognizes her voice and face, recalls she was watching Peppa Pig last time, and asks if she'd like to continue.

Directional Audio

Navigation That Only the Driver Hears

Headrest-integrated speakers create personal sound zones. Navigation guides the driver while the child watches cartoons in back — no interference.

Ambient Intelligence

A Sky Full of Stars — By Voice

"Xiaoyi, turn the ceiling into stars." The child controls ambient lighting with natural language. The interior transforms into a planetarium.

"The best interface is one that disappears. When a 6-year-old can command the car with a sentence, the technology has become invisible."

— Gordon Cheng, on designing the child interaction scenario
DTOF Parking — Beyond What Cameras See

The Sensor That Sees the Invisible.

Direct Time-of-Flight (dToF) sensors emit 940nm infrared laser pulses and use SPAD (Single Photon Avalanche Photodiode) pixels to measure distance with centimeter precision — detecting obstacles that cameras and ultrasonics fundamentally cannot. The Maextro S800 carries 4 proprietary 192-line LiDAR sensors plus dToF arrays, enabling what Huawei calls GOD (General Obstacle Detection) with 99.9% recognition rate.

Parking HMI

Split-Screen Parking Interface

3D bird's-eye view + rear camera + real-time obstacle overlay. The driver sees exactly what the car sees — with distance markers to the centimeter.

Rear-Wheel Steering

Tight-Space Automated Parking

Rear-wheel steering enables the car to park in spaces other vehicles simply cannot fit. Fully automated — the driver watches from outside.

Negative Obstacle Detection

Seeing What Isn't There

Sunken stairs, curb drop-offs, drainage ditches — obstacles defined by absence. DTOF detects changes in ground plane that cameras interpret as flat surface.

Suspended Obstacle Detection

The Pipe the Camera Missed

Hanging water pipes, gate arms, overhead barriers — above bumper height, invisible to traditional ultrasonic sensors. DTOF scans the full vertical plane.

The Process

From Engineering Specs to Human Stories

01

Feature Deconstruction

Translated Huawei's engineering documentation into user-facing feature maps. Identified 12 cockpit features and 6 DTOF capabilities. Prioritized by emotional impact, not technical complexity.

02

Scenario Design

Created character-driven scenarios: a father driving with his daughter. Every feature demonstrated through their interaction — the car becomes a character in a family story.

03

UX Storyboarding

Storyboarded each UI moment: voice bubble positioning, face recognition feedback, parking UI transitions. Defined when the interface appears and — critically — when it stays invisible.

04

Production Direction

Directed two films: a 79-second cockpit narrative and nine DTOF capability demos. Managed the tension between engineering accuracy and emotional storytelling.

"Engineering teams build features. Storytelling makes people trust them. My job was the bridge."

— Gordon Cheng
Takeaways

What Worked

  • The father-daughter scenario made abstract AI features emotionally concrete — viewers understood personalization through a child's voice, not a spec sheet
  • Showing DTOF detecting "invisible" obstacles (sunken stairs, hanging pipes) created immediate "I didn't know that was possible" moments
  • The split-screen parking UI gave technical audiences proof while the narrative gave everyone else trust

What I Learned

  • Automotive HMI is where my CD background (brand storytelling) and IxD training (interaction logic) converge — cabin experience design spans voice, ambient light, spatial audio, and biometrics, not just screens
  • The hardest UX problem isn't making features work — it's deciding when the interface should be invisible. The child talking to the car is better UX than any touchscreen
  • Chinese OEMs are 2-3 years ahead of most Western players in software-defined cockpit experiences. Working at this frontier gives me design knowledge directly applicable to Tesla, Rivian, Apple, and Qualcomm's automotive ambitions