Toyota Race Craft

Real-time strategy and decision-support engine for GR Cup racing teams

Comment

Inspiration

Motorsport teams make split-second decisions using massive amounts of telemetry. But most of that insight stays locked inside raw data streams. We wanted to build a system that transforms Toyota GR Cup telemetry into clear, actionable intelligence—helping engineers understand what really happened, why it happened, and how to improve. Toyota Race Craft was inspired by the desire to make professional-grade analytics accessible and intuitive.

What it does

Toyota Race Craft turns race telemetry into real insights and storytelling. It analyzes driver inputs, pace trends, battles, braking points, G-force patterns, and racing lines—then generates:

  • Real-time strategy guidance (pit windows, threats, pace shifts)
  • Post-event race narratives (turning points, mistakes, improvements)
  • Visual dashboards (timeline, heatmaps, track maps) It provides a complete data-driven view of every race.

How we built it

We ingested GR Cup telemetry from each track, normalized timestamps, reconstructed laps, and aligned multi-car data by lap distance. From there, we engineered features for braking, throttle, steering, G-forces, and degradation proxies, then applied lightweight ML models for pace forecasting and event detection. Finally, we built a structured event taxonomy and passed it into an LLM to generate human-readable narratives, wrapped in a modern Next.js dashboard.

Challenges we ran into

  • ECU timestamps were inconsistent, requiring rebuilding lap timing from lap-distance resets
  • Missing lap numbers forced us to derive laps using lapdist + meta_time
  • No tire temperature data—so we created a proxy degradation model
  • Multi-car alignment was difficult due to different sampling rates
  • Reducing noise in GPS and steering signals required multiple smoothing filters

Accomplishments that we’re proud of

  • Built a working end-to-end telemetry intelligence engine in hackathon time
  • Created clean visualizations that surfaced subtle race insights
  • Generated full, readable race stories directly from raw sensor data
  • Successfully inferred battles, pace shifts, and mistakes using only telemetry
  • Delivered a system that could realistically help GR Cup teams improve performance

What we learned

  • Motorsport telemetry is rich but extremely noisy—preprocessing is half the challenge
  • Lap-distance alignment is the key to all multi-car analysis
  • ML models don’t need to be complex; well-engineered features outperform heavy architectures
  • Narratives resonate with users far more than charts alone
  • Designing usable pit-wall tools requires balancing detail with clarity

What’s next for Toyota Race Craft

  • Add live data streaming for real-time race engineering
  • Integrate ElevenLabs voice narration for AI race commentary
  • Expand event taxonomy into full broadcast-grade motorsport analytics
  • Add driver coaching mode with corner-by-corner recommendations
  • Introduce multi-session comparison (practice → quali → race)
  • Deploy as a full cloud service for Toyota teams and GR Cup drivers

Built With

Share this project:

Updates