sr-tire

Utilities for:

• loading and processing tire test data
• plotting measured tire data together with the force model
• fitting a symbolic mu(v) expression with Flex symbolic regression

Layout

• src/sr_tire/process_data.py: load and split the tire dataset
• src/sr_tire/force.py: force-model implementation
• src/sr_tire/dynamic_pde/simulate_pde.py: JAX translation of the dynamic PDE simulation
• src/sr_tire/dynamic_pde/compare_pacejka.py: compare PDE forces against Pacejka data
• src/sr_tire/plot.py: plotting utilities
• src/sr_tire/steady_stribeck/flex_sr_run.py: Flex symbolic regression entry point
• src/sr_tire/steady_stribeck/config.yaml: Flex configuration
• data/lateral_tire_test.csv: tire dataset

Install

From the repo root:

python -m pip install -e .

Make Plots

To plot the default force model against the processed tire data:

python -m sr_tire.plot

Run Dynamic PDE Simulation

To run the dynamic PDE tire simulation:

python -m sr_tire.dynamic_pde.simulate_pde

The module saves the bristle deformation, force, and slip plots under the repository-root results/pde/. To open interactive plot windows as well:

python -m sr_tire.dynamic_pde.simulate_pde --show

To save those figures programmatically:

from sr_tire.dynamic_pde.simulate_pde import plot_simulation, simulate_pde

result = simulate_pde()
plot_simulation(result, output_dir="results/pde", show=False)

To run the equivalent dctkit cochain-based spatial discretization with Diffrax time integration:

python -m sr_tire.dynamic_pde.simulate_pde_dctkit

To compare the dynamic PDE forces with the comparable Pacejka force trajectory in data/Fx_dataset.csv, using parameterization=1, excitation=var2, run=5, sigma_0=0.12, amplitude ratio 0.4, and omega=10*pi:

python -m sr_tire.dynamic_pde.compare_pacejka

This writes comparison plots and per-trajectory error metrics to the repository-root results/pde/.

By default, the comparison plot shows that same var2, run 5 pair. To choose another pair within parameterization=1:

python -m sr_tire.dynamic_pde.compare_pacejka --excitation var2 --run 1

To run Flex SR for an additive term in the dynamic PDE right-hand side:

python -m sr_tire.dynamic_pde.flex_sr_run

The run is configured by src/sr_tire/dynamic_pde/config.yaml and scores each candidate by solving the PDE and comparing the resulting force trajectory with the selected trajectory from data/Fx_dataset.csv. The default configuration is intentionally small so it can complete on modest machines; increase num_individuals, generations, n_x, max_time_points, and the PSO settings only after the smoke-sized run finishes.

Run Flex SR

To run a single training run without hyperparameter optimization:

python -m sr_tire.steady_stribeck.flex_sr_run

To run multiple independent training runs:

python -m sr_tire.steady_stribeck.flex_sr_run --num-runs 5

To control the base random seed used for repeated runs:

python -m sr_tire.steady_stribeck.flex_sr_run --num-runs 5 --seed 123

To run with Optuna HPO:

python -m sr_tire.steady_stribeck.flex_sr_run --hpo

Run Random Search Baseline

To run random expression search with the same primitive set as Flex SR and a 7-minute wall-clock budget per run:

python -m sr_tire.steady_stribeck.random_search_run

To run multiple independent random-search runs:

python -m sr_tire.steady_stribeck.random_search_run --num-runs 5

To change the per-run time budget:

python -m sr_tire.steady_stribeck.random_search_run --time-budget-seconds 420

Results

After running Flex SR, the script:

• prints dataset summary information
• fits the symbolic mu(v) model
• saves all steady Stribeck artifacts under:

results/steady-stribeck/

Flex SR writes:

results/steady-stribeck/best_model_results.txt
results/steady-stribeck/best_model_plot.png
results/steady-stribeck/best_mu_plot.png
results/steady-stribeck/flex_runs/

The random-search baseline writes analogous results to:

results/steady-stribeck/random_search_best_model_results.txt
results/steady-stribeck/random_search_best_model_plot.png
results/steady-stribeck/random_search_best_mu_plot.png
results/steady-stribeck/random_search_runs/

Notes

• The current dataset split is:
  • bins 1-3: training
  • bin 4: validation
  • bin 5: test
• The CSV is loaded from data/ at the repo root.