GR Cup Cognitive Load Index (CLI) Monitor

AI-powered tool that estimates driver cognitive load in real time using telemetry, helping teams predict focus drops, prevent mistakes, and optimize race strategy

Comment

Inspiration

Motorsport teams understand everything about the car, but almost nothing about the driver’s mental state during a race. A single moment of cognitive overload can cause missed apexes, inconsistent braking, or costly mistakes—yet telemetry never shows it. I wanted to bridge that gap by creating an AI system that transforms raw telemetry into real-time insights about a driver’s focus, fatigue, and mental load.

What it does

The GR Cup Cognitive Load Index (CLI) Monitor estimates a driver’s cognitive load using telemetry alone. It provides:

  • A real-time Focus Score (0–100)
  • Predictions of upcoming focus drops
  • Alerts for unusual or risky driving patterns
  • Driver-specific baselines and mental profiles
  • A full interactive dashboard with replay, analytics, and comparisons

It gives engineers a new layer of intelligence: how the driver is performing mentally, not just mechanically.

How I built it

I processed over 640 laps from Sonoma and Road America, engineered 95 detailed features, and trained multiple ML models including:

  • An XGBoost focus-drop classifier (96.6% AUC)
  • An XGBoost focus score regressor
  • An Isolation Forest anomaly detector

I built a FastAPI backend for real-time inference and a Streamlit dashboard to visualize focus, alerts, driver baselines, and session timelines—all with <100ms latency.

Challenges I ran into

  • Defining a reliable proxy for cognitive load using telemetry alone
  • Handling differences across drivers, tracks, and driving styles
  • Avoiding false positives in focus alerts
  • Achieving real-time performance while computing dozens of features
  • Designing a dashboard simple enough for rapid race-day decisions

Accomplishments that I'm proud of

  • Creating the first AI cognitive-load estimator built specifically for motorsport
  • Achieving 96.6% accuracy in predicting upcoming focus drops
  • Building personalized mental baselines for 16 drivers
  • Designing a full real-time monitoring dashboard from scratch
  • Transforming telemetry into psychological insights—something not currently done in racing

What I learned

I learned that telemetry reveals far more about human behavior than expected. Small variations in steering entropy, throttle jerk, and braking smoothness can predict mental fatigue. I also learned the importance of explainability (via SHAP) when building tools that race engineers must trust in real time.

What's next for GR Cup Cognitive Load Index (CLI) Monitor

  • Integrate radio audio + video for multimodal cognitive analysis
  • Expand to full-race real-time deployment
  • Add team-wide strategy tools based on driver mental trends
  • Test and fine-tune the system with more tracks and drivers
  • Build a mobile engineer interface for pit-wall use

Ultimately, the goal is to make CLI a standard tool in motorsport, turning invisible human performance into data teams can use to win races.

Share this project:

Updates