PitWindow AI

Predict. Decide. Win. With PitWindow AI.

Comment

📘 About the Project — PitWindow AI

Inspiration

In real racing, pit strategy wins or loses races. A single lap can completely change the outcome. When Toyota GR released real GR Cup datasets—telemetry, lap times, braking pressure, throttle, and more—I immediately saw an opportunity: What if we built an AI that could think like a race engineer and calculate the perfect pit window in real time?

That question became the inspiration behind PitWindow AI — a digital race strategist designed to simulate, predict, and optimize decisions under pressure.

What it does

PitWindow AI transforms raw racing telemetry into live strategic insights:

  • Analyzes lap-by-lap telemetry in real time
  • Simulates race outcomes using a predictive model
  • Calculates the optimal pit-stop lap dynamically
  • Visualizes performance, degradation, and strategy deltas
  • Replays a race using actual GR Cup telemetry
  • Supports multi-driver and multi-car selection
  • Provides a clean Toyota GR–style dashboard for engineers

In short: It’s the race engineer’s co-pilot — powered by AI.

How we built it

🧠 Backend (Express.js + Node)

  • CSV ingestion for full GR Cup datasets
  • A custom Pit Strategy Engine that simulates future laps
  • Real-time telemetry stream using a replay loop
  • Multi-driver, multi-dataset support
  • Endpoints for strategy, live state, datasets, and drivers
  • Predictive modeling for pit windows
  • Degradation model: [ \text{lap_time}(age) = \mu + 0.15\cdot\text{age} + C_{\text{compound}} ]

🖥️ Frontend (Next.js + Recharts)

  • Toyota GR–inspired UI
  • Live replay dashboard (speed, gear, throttle, brakes, laps)
  • Interactive strategy visuals (line charts, timeline logs)
  • Optimal pit window card + estimated time gain
  • Race timeline generated on the fly

🔄 Datasets

We integrated actual GR Cup CSVs:

  • Lap times
  • Lap start/end intervals
  • Full telemetry (Speed, Gear, APS, Brake pressures, etc.)
  • Weather & sector data (optional)

Everything is processed live in a simulated race replay.

Challenges we ran into

  • Understanding GR Cup data structure Telemetry fields like aps, pbrake_f, and timestamp offsets required careful mapping.

  • Synchronizing multi-driver telemetry Cars did not always share identical sample frequencies or lap alignment.

  • Real-time simulation design Creating a realistic “live feed” from static telemetry required custom replay logic.

  • Strategy modeling Lap degradation, compound effects, and pit loss modeling had to feel believable while running fast enough for real-time updates.

  • UI clarity under information density Motorsports telemetry is noisy; designing a dashboard that remains readable was non-trivial.

Accomplishments that we're proud of

  • Built a fully working end-to-end racing strategy system in under the hackathon timeframe
  • Achieved real-time lap simulation and pit window calculation
  • Designed a clean, authentic Toyota GR–style UI
  • Added multi-dataset, multi-car support for deeper comparisons
  • Created a tool that could genuinely be used by race engineers

PitWindow AI isn’t just a demo — it’s a practical instrument.

What we learned

  • The complexity and richness of real motorsport telemetry
  • How critical lap alignment is when computing strategic decisions
  • The importance of UX clarity in high-pressure decision environments
  • How AI-based simulation can assist engineering decisions
  • How Toyota GR uses data to optimize performance in real-world conditions

What's next for PitWindow AI

  • Reinforcement-learning strategy engine Using RL agents to predict best pit laps under varying race conditions.

  • Safety car + yellow flag modeling Simulating unexpected events to adjust the pit window dynamically.

  • Driver Coaching Mode Break down anomalies in throttle, braking, or lines for training.

  • Weather integration Adjust degradation and strategy based on humidity/temperature data.

  • Full multi-car race replay mode Compare strategies side-by-side in a visual timeline.

  • Exportable strategy reports Generate PDF insights for post-race analysis.

PitWindow AI is just the beginning — the vision is to build the AI race strategist of the future.

Built With

Share this project:

Updates