OCP QP Benchmark

Benchmarking framework for OCP QP solvers using acados.

Installation

pip install -e .

Usage

Run benchmark

ocp-benchmark

Add problems to dataset

add-problems /path/to/json/folder --name my_dataset

Python API

from acados_template import AcadosOcpQpOptions
from ocp_qp_benchmark.core import TestSet, SolverSet, Results, run, create_solver_options
from ocp_qp_benchmark.visualization import plot_metric, generate_labels

# Create solver configurations
solvers = [
    create_solver_options("FULL_CONDENSING_HPIPM"),
    create_solver_options("FULL_CONDENSING_HPIPM", tol_stat=1e-10),  # Custom tolerance
    create_solver_options("FULL_CONDENSING_DAQP"),
]

# Create test set from problem folders
test_set = TestSet(qp_folder_paths=["ocp_qp_dataset_collection/random_qp/prob_0"])

# Run benchmark
solver_set = SolverSet(solvers)
results = Results(file_path="results/results.csv", test_set=test_set)
run(test_set, solver_set, results)

# Generate labels (auto-detects differing options)
labels = generate_labels(solvers)

# Plot results
plot_metric(metric="runtime_fair", df=results.df, test_set=test_set, labels=labels)

Supported Solvers

• PARTIAL_CONDENSING_HPIPM
• FULL_CONDENSING_HPIPM
• FULL_CONDENSING_QPOASES
• FULL_CONDENSING_DAQP
• PARTIAL_CONDENSING_OSQP
• PARTIAL_CONDENSING_CLARABEL