Discrete Optimization

Discrete Optimization is a python library to ease the definition and re-use of discrete optimization problems and solvers. It has been initially developed in the frame of scikit-decide for scheduling. The code base starting to be big, the repository has now been splitted in two separate ones.

The library contains a range of existing solvers already implemented such as:

• greedy methods
• local search (Hill Climber, Simulated Annealing)
• metaheuristics (Genetic Algorithms, NSGA)
• linear programming
• constraint programming
• hybrid methods (LNS)

The library also contains implementation of several classic discrete optimization problems:

• Travelling Salesman Problem (TSP)
• Knapsack Problem (KP)
• Vehicle Routing Problem (VRP)
• Facility Location Problem (FLP)
• Resource Constrained Project Scheduling Problem (RCPSP). Several variants of RCPSP are available
• Graph Colouring Problem (GCP)

In addition, the library contains functionalities to enable robust optimization through different scenario handling mechanisms) and multi-objective optimization (aggregation of objectives, Pareto optimization, MO post-processing).

Installation

Prerequisites

• Install minizinc.

• Optionally, install gurobi with its python binding (gurobipy) and an appropriate license, if you want to try solvers that make use of gurobi.

  NB: If you just do pip install gurobipy, you get a minimal license which does not allow to use it on "real" models.

Normal install

Install discrete-optimization from pip:

pip install discrete-optimization

Install in developer mode

You can also install the library directly from the repository in developer mode:

git clone https://github.com/airbus/discrete-optimization.git
cd discrete-optimization
pip install --editable .

If you encounter any problem during installation, please fill an issue on the repository.

Examples

Notebooks

In the notebooks directory of the repository, you will find several jupyter notebooks demonstrating how the library can be used

• on a knapsack problem,
• on a scheduling problem (RCPSP).

Scripts

The examples directory of the repository gather several scripts using the different features of the library and how to instantiate different problem instances and solvers.

Unit tests

Unit tests are available in tests/ directory of the repository. To test the library, you can install the library with the "test" extra dependencies by typing

git clone https://github.com/airbus/discrete-optimization.git
cd discrete-optimization
pip install --editable .[test]

Then run pytest on tests folder:

pytest -v tests

License

This software is under the MIT License that can be found in the LICENSE file at the root of the repository.

Some minzinc models have been adapted from files coming from

• https://github.com/MiniZinc/minizinc-benchmarks under the same license,
• https://github.com/youngkd/MSPSP-InstLib for which we have the written authorization of the author.