GR Race Strategist AI

Real-time pit window, tire wear, pace prediction, and undercut strategy engine built from TRD race data to help engineers make faster, smarter race-day decisions.

Comment

🚦 Overview

GR Race Strategist AI is a real-time decision-making system built on TRD’s race telemetry to help engineers identify optimal pit windows, predict tire degradation, simulate undercut/overcut scenarios, and forecast pace drops before they happen.

Modern motorsports is won through strategy, not just speed. A wrong pit stop timing can cost 15–20 seconds instantly. This project aims to give race engineers a practical tool that uses real data to make smarter, faster, more confident decisions during the heat of competition.

🎯 What the Project Does

GR Race Strategist AI provides four core features:

  1. Real-Time Tire Wear Prediction A model that estimates Tire Wear Index (TWI) per lap using braking heat, lateral load, acceleration, and pace data.

  2. Pace Projection Engine Forecasts the next 1–3 laps using tire wear, stint history, and track evolution.

  3. Optimal Pit Window Detection Alerts race engineers when tire degradation or pace drop reaches the point where pitting yields better long-run pace.

  4. Undercut/Overcut Simulator Compares predicted pace curves between cars to estimate the probability of gaining position via pit strategy.

🛠️ How I Built It Data Engineering

Processed TRD telemetry: lap times, sector deltas, throttle %, brake pressure, lateral G, tire temperature & load, and stint history.

Generated engineered features such as:

Tire Wear Index (TWI) Driver Consistency Score Traffic Loss Factor Track Evolution Rate Pace Deviation Model Machine Learning Models Random Forest → Tire Wear prediction LightGBM → Lap pace forecasting Rule-Based Strategy Engine → Pit window + undercut logic Simulation Engine → 3-lap forward prediction

Frontend

A race-engineer console displaying: Tire wear curve Pit window indicator Pace projection graph Undercut meter Alert panel Designed using Toyota GR colors: black, red, white, asphalt grey. Backend FastAPI endpoints for predictions, simulation, and telemetry ingestion.

💡 What Inspired Me

In motorsports today, strategy tools are often proprietary and unavailable to the public. This hackathon provided a rare chance to build something real using genuine Toyota race telemetry — something close to what actual TRD engineers might use on the pit wall.

🔥 What I Learned

How to transform raw motorsport telemetry into usable insights Designing predictive models around tire physics Building real-time simulation logic Structuring a project around actual race-day workflows Creating a UI for high-pressure environments (clarity, speed, contrast)

🚧 Challenges

Telemetry noise & missing data Synchronizing features with lap numbers Predicting tire degradation with limited direct tire metrics Balancing ML accuracy vs. real-time performance Designing an interface that works in a stressful race setting

🏁 Impact & What’s Next

This tool could support: Race engineers during live sessions Drivers in simulators Coaching programs for junior drivers Broad telemetry education for fans

Next steps:

Add caution-flag prediction Add multi-car strategy comparison Integrate weather-based strategy AI Deploy as a full pit-wall web app

Built With

Share this project:

Updates