VW
Drive Mentor+ Eco-Drive
Drive together. Grow together.
Volkswagen CarBrain Student Competition

From monitor to mentor

Drive Mentor+ Eco-Drive turns the car from a reactive nanny into an active co-pilot — intuitive, empathetic, a partner on the road.

Problem

The "Nanny" Effect

Current Advanced Driver Assistance Systems (ADAS) are primarily reactive and intrusive. Rather than collaborating with the driver, the vehicle acts as a digital nanny — jarring beeps, abrupt braking, flashing alerts. The result: cognitive load, frustration, and drivers who disable safety features entirely.

The vision

Three pillars of the Active Co-Pilot

Proactive Guidance

Anticipates the driver's needs and intervenes early — gentle steering nudges, smooth speed modulation — before any crisis.

Empathetic Interaction

Replaces harsh "beeps" with a calm, human voice that adapts tone to the driver's state.

Skill Elevation

Doesn't drive for you — helps you become a better driver via real-time, low-stress coaching.

Architecture

Four layers, zero latency

Cameras and sensors stream into the edge, where safety decisions happen locally. The cloud handles only what can wait until the trip ends.

Layer 1

Sensors

DMS cameras read blink rate, grip and facial tension. External cameras read traffic, lanes, pedestrians. Wheel speed, steering angle, brakes — all streamed.

Layer 2

Edge Perception

Stress × road-complexity × grip evaluated locally on the NPU. Zero-latency response so high-anxiety moments never wait on a cloud round-trip.

Layer 3

Decision Layer

Capability vs threat. Caps power, suppresses non-essential UI, or escalates to the Parental Loop when safety overrides are challenged.

Layer 4

Output

Empathetic voice, soft tactile assist, cabin ambient mute during stress. The mentor never beeps — it speaks, glows, or guides the wheel.

Edge (in-car)

Zero-latency decisions

  • Real-time driver stress detection
  • Immediate voice coaching
  • ESC + pedal calibration
  • Audio mute during stress
Cloud (post-drive)

Video-assisted debrief

  • Stitches cabin + external cameras
  • Long-term driver profile
  • Track Golden Model updates
  • Parent summaries
Adaptive persona system

Four personas, one mentor

Research personas

Three personas, built from 111 interviews

N

Noa Shapira

· "The Confidence Builder"
Age: 19Experience: 0–5 yearsSurvey: 34 respondentsGoal: Confidence on the road
The hardest part is making quick decisions when something happens on the road. I need the car to guide me in real time and help me feel less alone out there.
Goals
  • Get real-time, calm guidance under pressure
  • Build confidence drive by drive
  • Feel supported, not watched
  • Avoid being shamed for small mistakes
Frustrations
  • Harsh beeps that escalate panic
  • Systems that take over instead of teaching
  • Parents stress-monitoring every drive
  • Generic alerts that don't match the moment
During the drive
  • · Empathetic voice guidance, no harsh alerts
  • · HUD stripped to essentials (speed, route, distance)
  • · Soft parking lines + audible support
Post-drive insights
  • · Personal progress recap, not a scorecard
  • · Driver-pattern memory across drives
  • · Highlights, not gotchas
Design note:Noa is DriveMentor's emotional core. Design for warmth — every interaction must lower her cognitive load and reinforce small wins.
D

David Levi

· "The Experienced Self-Improver"
Age: 38Experience: 10+ yearsSurvey: 59 respondentsGoal: Quiet competence
I want a car that knows when I'm fine and when I'm not. Don't talk to me unless it matters — and when it does, get it right.
Goals
  • Receive meaningful post-drive feedback on performance
  • Stay aware of road risks without feeling nagged
  • Improve driving quality without being lectured
  • Understand long-term trends in his driving
Frustrations
  • 34% have already disabled a safety system — felt too intrusive
  • Safety alerts that feel repetitive and unnecessary
  • No clear way to know if driving has improved over time
  • Systems that over-intervene and reduce driving enjoyment
During the drive
  • · Subtle hazard alerts — gentle cues, no beeps
  • · Real-time instructions only when genuinely critical
  • · Smart route and fuel efficiency suggestions
Post-drive insights
  • · Post-drive summary: what went well / what didn't
  • · Long-term improvement tracking over weeks
  • · Driving quality score with trend over time
Design note:David is DriveMentor's largest audience — competent, confident, skeptical of systems that over-intervene. Design respectful, data-driven, lets him feel in control.
T

Tal Mizrahi

· "The Performance Seeker"
Age: 27Experience: 2–14 yearsSurvey: 18 respondentsGoal: Speed & performance
I want to use everything my car has to offer — but safely. If you can unlock more performance once I prove I can handle it, I'm in.
Goals
  • Push the car to its real limit on closed tracks
  • Earn more performance through smoothness
  • Get coaching that actually improves lap times
  • Compare laps over time and see progress
Frustrations
  • Nanny systems that intervene mid-corner
  • ESC that triggers when he doesn't want it
  • Apps that show data but don't help him improve
  • Coaches that lecture instead of point to the gap
During the drive
  • · Dynamic racing line on the HUD
  • · Carrot progression bar for clean sectors
  • · Short technical coaching at corner exits
Post-drive insights
  • · Lap comparison + sector breakdown
  • · Pinpoint the time gained vs the Golden Model
  • · Next training goal automatically scoped
Design note:Tal wants partnership, not protection. Design with technical respect — earn his trust by giving him real telemetry and real reasons.
Learner mode · cognitive load

Four principles for a calmer cabin

Reduce stress, increase focus, build driver confidence — through empathy and intelligent adaptation.

01

Cognitive Load Management

Entertainment audio is lowered during complex maneuvers. Secondary feedback (Eco scores, gamified counters) is hidden until basic competence is achieved.

02

Calming Visual Cues

Soft illumination highlights potential hazards. No aggressive colors, no complex racing lines in Learner Mode. Gentle cues guide — they never alarm.

03

Empathetic Voice Assistant

Calm, encouraging voice. Actionable timely guidance — "You have space, Maya. Smoothly accelerate to match traffic speed." — instead of harsh alerts.

04

Progressive Complexity

The experience evolves as the driver improves. The system gradually introduces more information and rewards growth toward mastery.

User journey

Noam's first solo drive

From the moment he gets in to the moment he's home — no nagging, just a calm presence.

Beat 1

The car recognises Noam

Seat moves into his saved position. Mirrors adjust. The dashboard simplifies — only speed, route, safe distance, parking assist. A friendly voice: 'Hi Noam. New Driver Mode is active.'

Open in simulation
Beat 2

Leaving the parking space

He shifts into reverse. 360° view appears. Instead of beeping, the car speaks calmly: 'You're doing well. Turn slightly left… great. Now pause for a moment.' Nice job.

Open in simulation
Beat 3

The first merge

Cars are faster than he expected. His foot hovers. The car notices the hesitation: 'There's a car on the left — let's wait. Now you can merge gently. Light pressure on the accelerator.' Smile.

Open in simulation
Beat 4

An interruption

A car behind honks. Heart races. The car responds before pressure takes over: 'You're driving correctly. There's a crosswalk ahead. Don't let the honking rush you.' Quiet support.

Open in simulation
Beat 5

Eco coaching moment

He braked a little late. The car waits, then softly: 'Next time, when we see a red light early, release the accelerator sooner. It saves fuel and the stop feels smoother.' Learning.

Open in simulation
Beat 6

Parking at the mall

The system finds a spot at his current level. Step-by-step guidance. 'You were a little close on the right, but you noticed and corrected. That's real progress.' The car learned his approach angle for next time.

Open in simulation
User journey · sport

Tal's Track Day

From mode activation to a new personal best — through consistency, not risk.

Beat 1

Track Mentor Mode loads

Tal enters the cockpit. The car switches to Track Mentor Mode — loads his racing profile, previous lap history, weak corners, today's goals. The voice: 'Hi Tal. Today's focus: braking consistency, cleaner corner exits, smoother throttle control.' Dynamic Unlock is active. Better driving will unlock more performance.

Open in simulation
Beat 2

Controlled track entry

Tal leaves the pit lane and starts his warm-up lap. The system monitors tyre and brake temperature, track grip, conditions. Instead of pushing him immediately: 'Warm-up lap. Build tyre temperature gradually.' Live data quietly collected for the first fast lap.

Open in simulation
Beat 3

Performance diagnosis · Turn 4

Tal begins his first flying lap. HUD shows the racing line and tracks performance. In the first corners, small steering corrections noted, short technical feedback after each. At Turn 4 — his known weak corner — he brakes too late again. The system highlights it clearly: 'Turn 4 entry was too late. Exit speed dropped. Brake earlier next lap and use a wider entry.' Tal immediately understands where he's losing time.

Open in simulation
Beat 4

Earned performance · Dynamic Unlock

Over the next section Tal drives more smoothly. Braking stabilizes, line improves. The system rewards him: 'Dynamic Unlock increased.' The car becomes more responsive, more performance available. A few corners later, Tal pushes too hard into Turn 8. The system detects low grip margin: 'Grip margin low. Reduce throttle.' Tal corrects. The system slightly *reduces* the unlock — performance is earned, and can be pulled back when risk rises.

Open in simulation
Beat 5

Pit-to-track improvement loop

After the first session Tal returns to the pit lane. A short session summary: strongest section, weakest corner, main improvement area. Turn 4 is clearly the biggest issue. In the next session Tal applies the advice — brakes earlier into Turn 4, enters wider, gets back on throttle sooner. The system confirms: 'Excellent. Exit speed improved. That was your cleanest Turn 4 today.' The lap isn't just faster — it's smoother.

Open in simulation
Beat 6

Eco-performance cooldown

On the cooldown lap the system switches to EcoDrive performance mode. It helps Tal preserve tyres, cool the brakes, carry momentum efficiently. Feedback stays short: 'Lift earlier. Use momentum. Keep airflow through the brakes.' For Tal, EcoDrive becomes part of *race strategy*, not just efficiency.

Open in simulation
Beat 7

Precision final lap · new PB

Last session, one clean lap. Tal follows the racing line, brakes earlier into Turn 4, keeps the car smooth through the key sections. The system says very little — it lets him focus. As he crosses the line: 'New personal best.' The car explains the improvement came from smoother, more precise driving — not from taking more risk.

Open in simulation
Beat 8

Driver reflection

Later, Tal reviews the session. Lap comparisons, corner analysis, key progress points. One message stands out: 'Your fastest lap was also your smoothest lap.' That captures the whole experience. The car didn't simply help Tal go faster — it helped him become more precise, more consistent, more aware as a race driver.

Open in simulation
HUD · track mode

Coaching on the windshield, not alarms

Dynamic Racing Line

The HUD projects the optimal racing line onto the windshield. Color shifts dynamically — blue for acceleration, red for braking — based on the Golden Model for that circuit.

The "Carrot" Progression Bar

Subtle gamified progress indicator. When the driver strings together smooth corners, the bar fills toward the next Performance Unlock. No flashing, no shame.

Non-Intrusive Coaching

Instead of red warning signs for mistakes: a ghost car showing the ideal line, or a calm arrow pointing to a slightly wider turn entry. Visual coaching, never visual punishment.

Track mode

Dynamic Unlock · Golden Model

The engine doesn't open up until you've earned it. The car compares every brake, line, and pedal input to a Golden Model of the circuit. Three clean sectors and more performance unlocks.

Push too hard? ESC threshold drops back, unlock retracts. Real speed comes from precision, not risk.

Stage 1
Warm-up
30%
Baseline throttle map
Stage 2
Three clean sectors
70%
Throttle sharpens · ESC threshold raised
Stage 3
Precision lap
90%
+10% torque · personal best mode
Engineering parameters

Same architecture, two modes

Learner Mode looks for stress to increase support. Track Mode looks for consistency to unlock power. Same loop — sensors, decision, output.

Safety & Stress Telemetry

Learner Mode looks for stress indicators to *increase* active support.

Inputs · sensor layer
  • Driver physical stress: facial expressions, grip pressure, blink rate
  • Environment: speed limit, dangerous traffic proximity
  • DMS cameras + external road cameras cross-referenced
Evaluation logic

The AI analyzes cognitive load relative to the environment. High stress + complex road conditions → it *increases* active intervention rather than issuing yet another alert.

Output · microcontrollers act
  • Power Limiting · caps acceleration & top speed for a smooth, predictable ride
  • Steering Assist · slightly stiffens the wheel so the driver stays centered without overcorrecting
  • The Parental Loop · if a young driver tries to override safety, the system halts the request and pings the parent/guardian for real-time authorization

Golden Model Telemetry

Track Mode looks for performance *consistency* to unlock more power.

Inputs · sensor layer
  • Steering angle rates, lateral/longitudinal G-forces
  • Pedal application smoothness
  • Tyre & brake temperatures, grip margin
Evaluation logic

Compares real-time telemetry against the optimal Golden Model for that circuit. Looks for *consistency*, not raw speed. Erratic steering or abrupt braking prevents progression.

Output · microcontrollers act
  • Throttle Mapping · sharpens pedal response after three consecutive clean laps
  • Motor Torque · releases an additional ~10% of available power
  • ESC Intervention · raises the Electronic Stability Control threshold so the driver feels real control before the system steps in
Special mechanism

The Parental Loop

If a young driver tries to override a safety restriction — disable a speed cap, raise power, mute a warning — the system doesn't just refuse. It halts the request and sends a real-time authorization prompt to the registered parent on their phone.

The parent approves or declines. No live tracking, no shadowing every small mistake — just the moments that matter.

01
Young driver
Tries to override a safety limit
02
DriveMentor
Halts the request
03
Parent app
Real-time authorization request
04
Approve / deny
Vehicle acts on the answer
User research

Survey of 111 drivers

111
Total respondents
31%
New drivers
84%
Want active intervention

Users value safety technology — but despise nagging interruptions.

Our team

Who built Drive Mentor+

One of 30 teams out of hundreds of universities to reach the finals of the Volkswagen student competition.

A
Head of Prototyping
Adir Ben Ya'aqov
B.Sc. Mechanical Engineering
Head of Prototyping · concept, code, deployment
Project team
R
Rim Gahnim
M.Sc. Energy Engineering
Project Manager
D
Doron Farhi
B.Sc. Electrical Engineering
Computer Vision · Project & Product Mgmt · Data Analysis
A
Arad Rotem
B.Sc. Computer Science
Machine Learning · Data Analysis · Algorithmic Thinking
A
Amit Mendelbaum
B.Sc. Electrical Engineering
Computer Vision · Project & Product Mgmt
E
Emil Virnik
B.Sc. Mechanical Engineering
Vehicle Dynamics & Technical Architecture
Z
Ziv Shamli
M.Sc. Intelligent Systems
Computer Vision · Project & Product Mgmt · Data Analysis
K
Katya Reznik
B.Sc. Electrical Engineering
AI
I
Izhak Genish
B.Sc. Electrical Engineering
HW-SW Integration · Computer Vision · Sensor Data Analysis
A
Alon Dori
B.Sc. Electrical Engineering
AI
A
Adi Harel
B.Sc. Software Engineering
Python · C++ · Java · UX
M
Michal Lev
B.Sc. Software Engineering
DBA · System Automation · Fullstack · UX
D
Daniel Nitbach
B.Sc. Software Engineering
Computer Vision · Machine Learning · Data Analysis · Fullstack
Faculty & mentors
M
Michal Hochman, PhD
Autonomous Vehicles · Pedestrian Interaction · UX Research · PM
P
Prof. Erel Avineri
Transportation Sciences · Technion
N
Nir Kahn
Vehicle Designer
E
Eyal Katz, PhD
Electrical & Computer Engineering · Computer Vision · Multi-Modal LLMs
S
Sharon Yalov-Handzel, PhD
Computer Science · AI & Robotics
Y
Yair Even-Zohar, PhD
Machine Learning · LLM · NLP
D
Dan Hermann, PhD
Vehicles and Engines
I
Ido Israel Antebi, PhD
Mechanical Engineering · Systems Engineering
A
Adi Fox, PhD
Industrial Engineering · DSS & Process Engineering
K
Keren Ben Haim
Attorney · Director of OFEK Skills Development Center
T
Tair Kowalski
Head of Innovation & Entrepreneurship Center
Program sponsored by Volkswagen CarBrain. The simulation here is the team's prototype — not an official Volkswagen vehicle.

Ready for the drive?

Four personas, 21 scripted beats, real Gemini voice, and more.

Try the simulation