Bin Packing Optimization

A research-oriented Python package for benchmarking exact, heuristic, and hybrid algorithms on the 1D Bin Packing Problem.

Project overview

The package under bin_packing_optimization provides solver implementations, dataset parsers, and a reusable benchmarking harness. The repository is intended for reproducible experimentation rather than production deployment.

Repository layout

.
├── bin_packing_optimization/
│   ├── datasets/
│   ├── exact_methods/
│   ├── specific_heuristics/
│   ├── trajectory_based_metaheuristics/
│   ├── population_based_metaheuristics/
│   ├── learning_guided_metaheuristics/
│   │   └── hybrid_alns/
│   └── utilities/
├── pyproject.toml
├── requirements.txt
└── results/

Solver families

Family	Description
Exact methods	Optimal search strategies such as branch-and-bound and dynamic programming.
Specific heuristics	Fast constructive heuristics and local improvement procedures.
Trajectory-based metaheuristics	Single-solution methods such as simulated annealing and tabu search.
Population-based metaheuristics	Population-driven methods such as genetic algorithms and ACO.
Learning-guided metaheuristics	Hybrid ALNS pipelines that combine search heuristics with learned repair models.

Installation

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pip install -e .

Quick start

import importlib

from bin_packing_optimization.utilities.benchmarking import create_benchmark

solver_module = importlib.import_module("bin_packing_optimization.exact_methods.solver")

benchmark = create_benchmark(
    dataset_key="falkenauer-t",
    solver_module=solver_module,
    time_limit=None,
)
benchmark.run(method="branch and bound")
benchmark.save_results_to_csv()

Benchmark outputs are written under the repository-wide results/ tree, with CSV files grouped by dataset key and timestamp. Graphs are emitted alongside the CSV output when graph generation utilities are used.