F1 Aerodynamic Configuration Optimization

A comprehensive machine learning system for optimizing Formula 1 car aerodynamic configurations using CFD data, telemetry, and driver feedback. The system combines deep learning techniques with gradient boosting to provide optimal setup recommendations for different tracks and conditions.

Overview

This project implements a comprehensive pipeline for F1 car setup optimization, focusing on aerodynamic configuration. It processes multiple data sources:

• CFD (Computational Fluid Dynamics) simulations
• Real-time telemetry data
• Track-specific characteristics
• Weather conditions
• Structured driver feedback

Architecture

Deep Learning Pipeline

• CNN Module: Processes spatial CFD data to extract flow patterns
• LSTM Module: Handles temporal telemetry data sequences
• Fusion Layer: Combines features from multiple data sources
• XGBoost Optimizer: Final setup parameter optimization

Data Processing

• CFD Data: 32x32 spatial grid with 16 channels (pressure and velocity fields)
• Telemetry: 32 sensors including speed, temperatures, and aerodynamic measurements
• Weather: 8 parameters including temperature, humidity, wind conditions
• Driver Feedback: 12 structured metrics for car handling characteristics

Installation

# Clone the repository
git clone https://github.com/a-romero/f1-aero-optimization.git
cd f1-aero-optimization

# Create and activate virtual environment (optional but recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Requirements

• Python 3.8+
• PyTorch 2.0+
• NumPy
• Pandas
• SciPy
• XGBoost
• PyTorch Lightning

Usage

Basic Usage

from aero_optimization import optimize_aero_setup
from data_generation import generate_sample_data

# Generate sample data for a specific track
track_data, weather_forecast, initial_setup = generate_sample_data("Silverstone")

# Run optimization
optimal_setup = optimize_aero_setup(track_data, weather_forecast, initial_setup)

# Print optimized setup
for param, value in optimal_setup.items():
    print(f"{param}: {value:.2f}")

Track-Specific Configuration

The system supports different F1 circuits, each with its unique characteristics:

tracks = ["Monaco", "Monza", "Silverstone", "Spa"]
for track in tracks:
    track_data, weather, setup = generate_sample_data(track)
    optimal_setup = optimize_aero_setup(track_data, weather, setup)

Data Structure

CFD Data

cfd_data.shape = (n_samples, 32, 32, 16)
# - n_samples: Number of measurements
# - 32x32: Spatial grid points
# - 16 channels: Pressure (0-7) and velocity (8-15) fields

Telemetry Data

telemetry_data.shape = (n_samples, 100, 32)
# - n_samples: Number of laps
# - 100: Time steps per lap
# - 32: Sensor measurements

Driver Feedback Categories

• Overall balance
• Front/rear grip
• Straight-line stability
• Corner entry/exit performance
• Brake stability
• Traction
• Kerb riding
• Turn-in response
• Mid-corner balance
• Tire degradation

Model Parameters

Aerodynamic Setup Parameters

• Front/rear wing angles (degrees)
• Brake balance (% forward)
• Differential settings (entry/mid/exit %)
• Ride heights (mm)
• Anti-roll bar settings (1-11 scale)

Future Improvements

• Real-time optimization during race weekends
• Integration with wind tunnel data
• Enhanced weather impact modeling
• Driver preference learning
• Multi-objective optimization for different racing scenarios

License

This project is licensed under the MIT License - see the LICENSE file for details.

Authors

• Alberto Romero - @a-romero