03 02 tire degradation modeling - VforVitorio/F1_Strat_Manager GitHub Wiki

Tire Degradation Modeling

Relevant source files

Purpose and Scope

This document details the tire degradation modeling subsystem of the F1 Strategy Manager. The system uses machine learning to predict how tire performance deteriorates over successive laps in a Formula 1 race, which is critical for optimal pit stop strategy planning. This module provides accurate predictions of future tire performance that feed directly into the expert system's rule engine. For information about how these predictions are used in strategy decisions, see Degradation Rules.

Core Concepts

Tire Degradation in Formula 1

Tire degradation in Formula 1 racing refers to the progressive loss of tire performance over the course of a stint. This degradation is reflected in increasing lap times as tires wear out. The degradation pattern is not strictly linear and depends on multiple factors:

  • Tire compound (Soft, Medium, Hard)
  • Track conditions (temperature, surface)
  • Driving style
  • Fuel load Our system models this complex relationship using sequence-based machine learning techniques that can capture non-linear patterns in the data. The tire life cycle follows three distinct phases:
  1. Initial Phase: First few laps where performance is stable
  2. Critical Monitoring Phase: When meaningful degradation patterns emerge
    • Soft Compounds: ~6 laps
    • Medium Compounds: ~12 laps
    • Hard Compounds: ~25 laps
  3. End-of-Life Phase: Severe performance drop-off ("cliff") Our modeling focuses particularly on the Critical Monitoring Phase, as this is when strategic decisions become most relevant.

Fuel Effect Adjustment

A key aspect of accurate tire degradation modeling is isolating the tire effect from the fuel effect. As cars burn fuel during a race, they become lighter and therefore faster, which can mask the true tire degradation. Our system uses a constant value of 0.055 seconds per lap as the empirical improvement due to fuel burn. This value is added back to the lap times to create "fuel-adjusted" metrics that more accurately reflect the pure tire degradation.

FuelAdjustedLapTime = ActualLapTime + (LapsFromBaseline * 0.055)

Data Processing Pipeline

The tire degradation modeling system follows a sophisticated data processing pipeline that transforms raw telemetry data into prediction-ready inputs.

Data Pipeline Diagram

Required Input Data

The tire degradation model requires specific telemetry data for each lap:

Column Name Description Purpose
LapTime Time taken to complete the lap Primary performance metric
CompoundID Tire compound (1=Soft, 2=Medium, 3=Hard) Determines degradation pattern
TyreAge Number of laps tire has completed Primary predictor of degradation
Stint Current stint number Tracks tire changes
SpeedI1, SpeedI2, SpeedFL Speed at various track points Additional performance indicators
FuelLoad Estimated fuel load in kg Used in fuel adjustment
Position Current race position Race context
DriverNumber Driver's race number Identifies the driver

Degradation Metrics Calculation

The system calculates several key metrics to quantify tire degradation:

  1. FuelAdjustedLapTime: Actual lap time with fuel effect added back
  2. FuelAdjustedDegPercent: Percentage degradation compared to baseline (first lap on a new set)
  3. DegradationRate: Rate of change in lap time from one lap to the next

Sequence Creation

The system treats tire degradation as a sequential problem, using a window of consecutive laps to predict future degradation. Sequences are created based on compound-specific starting laps:

These thresholds focus the model on the most strategically relevant phases of tire life.

Model Architecture

Temporal Convolutional Network (TCN)

The tire degradation prediction system uses an enhanced Temporal Convolutional Network (TCN) architecture, which is well-suited for capturing both short and long-term patterns in sequential data.

Key Model Components

  1. Multi-scale Dilated Convolutions: Captures patterns at different time scales
    • Dilation rates: 1, 2, 4, 8
    • Each block includes BatchNorm and ReLU activation
    • Residual connections to help with gradient flow
  2. Temporal Attention Mechanism: Focuses the model on the most relevant time steps
    • Implemented as a learnable weighted sum over the time dimension
    • Helps extract key information from the input sequence
  3. Fully Connected Layers: Transform convolutional features into predictions
    • Dropout (0.3) for regularization
    • Output is the predicted degradation for the next 3 laps

Ensemble Prediction Approach

The system employs an ensemble approach that combines:

  1. Global Model: Trained on data from all tire compounds
  2. Specialized Models: Separate models trained specifically for each compound Predictions are combined using a weighted average based on model performance:
ensemble_prediction = (global_weight * global_prediction + 
                      specialized_weight * specialized_prediction)

Weights are derived from the inverse of each model's RMSE (Root Mean Square Error):

Model RMSE Use Case
Global 0.355017 General predictions across all compounds
Soft (ID: 1) 0.334325 Specialized predictions for soft tires
Medium (ID: 2) 0.392661 Specialized predictions for medium tires
Hard (ID: 3) 0.295417 Specialized predictions for hard tires

Prediction Process

Model Loading and Preparation

The system loads pre-trained models from disk, including both the global model and any available compound-specific models:

Making Predictions

The prediction process involves several steps:

  1. Sequence Preparation: Convert sequences of laps into tensors
  2. Global Model Prediction: Get predictions from the global model
  3. Specialized Model Prediction: For applicable compounds, get predictions from specialized models
  4. Ensemble Weighting: Combine predictions with appropriate weighting
  5. Formatting Results: Structure predictions in an easy-to-use format

Output Format

The final output of the tire degradation prediction system is a structured DataFrame containing:

Column Description
DriverNumber Driver identifier
Stint Current stint number
CompoundID Numeric compound ID
CompoundName Human-readable compound name (Soft, Medium, Hard)
CurrentTyreAge Current age of tires in laps
CurrentLapTime Most recent lap time
FutureLap Future lap number (TyreAge)
LapsAheadPred How many laps ahead of current (1-3)
PredictedDegradationRate Predicted degradation rate in seconds

Real-time Prediction for Strategy Decisions

The system supports real-time prediction during a race, with monitoring starting at compound-specific lap thresholds:

This approach focuses computational resources on strategically relevant predictions while aligning with how F1 teams monitor tires during races.

Conclusion

The tire degradation modeling subsystem provides crucial strategic intelligence for F1 race strategy. By accurately predicting how tire performance will evolve over future laps, it enables more informed pit stop decisions and compound selections. Key strengths of the system include:

  1. Fuel-adjusted metrics that isolate true tire degradation
  2. Compound-specific modeling that accounts for different degradation patterns
  3. Ensemble approach that combines general and specialized knowledge
  4. Integration with expert system for actionable strategy recommendations This module represents a critical component in the F1 Strategy Manager's ability to provide data-driven strategy recommendations.

On this page