| OS | C++ | Python | Dotnet | |:-------|-----|--------|--------| | Unix (linux + macos) | | | | | Windows | | | |

cspy

A collection of algorithms for the (resource) Constrained Shortest Path (CSP) problem.

Documentation here.

The CSP problem was popularised by Inrich and Desaulniers (2005). It was initially introduced as a subproblem for the bus driver scheduling problem, and has since then widely studied in a variety of different settings including: the vehicle routing problem with time windows (VRPTW), the technician routing and scheduling problem, the capacitated arc-routing problem, on-demand transportation systems, and, airport ground movement; among others.

More generally, in the applied column generation framework, particularly in the scheduling related literature, the CSP problem is commonly employed to generate columns.

Therefore, this library is of interest to the operational research community, students and academics alike, that wish to solve an instance of the CSP problem.

Algorithms

Currently, the exact and metaheuristic algorithms implemented include:

• [x] Bidirectional labeling algorithm with dynamic halfway point (exact) (also monodirectional) Tilk et al. (2017);
• [x] Heuristic Tabu search (metaheuristic);
• [x] Greedy elimination procedure (metaheuristic);
• [x] Greedy Randomised Adaptive Search Procedure (GRASP) (metaheuristic). Adapted from Ferone et al. (2019);
• [x] Particle Swarm Optimization with combined Local and Global Expanding Neighborhood Topology (PSOLGENT) (metaheuristic) Marinakis et al. (2017).

Please see the docs for individual algorithms Python or C++ API documentation, as well as some toy examples and further details.

• Bidirectional and monodirectional algorithms
• Heuristic Tabu Search
• Greedy Elimination Procedure
• GRASP
• PSOLGENT

Getting Started

Prerequisites

Conceptual background and input formatting is discussed in the docs.

Module dependencies are:

• NetworkX
• NumPy

Note that requirements.txt contains modules for development purposes.

Installing

Installing the cspy package with pip should also install all the required packages. You can do this by running the following command in your terminal

none pip install cspy

or

none python3 -m pip install cspy

Quick start

Python

```python

Imports

from cspy import BiDirectional from networkx import DiGraph from numpy import array

maxres, minres = [4, 20], [1, 0]

Create a DiGraph

G = DiGraph(directed=True, nres=2) G.addedge("Source", "A", rescost=[1, 2], weight=0) G.addedge("A", "B", rescost=[1, 0.3], weight=0) G.addedge("A", "C", rescost=[1, 0.1], weight=0) G.addedge("B", "C", rescost=[1, 3], weight=-10) G.addedge("B", "Sink", rescost=[1, 2], weight=10) G.addedge("C", "Sink", res_cost=[1, 10], weight=0)

init algorithm

bidirec = BiDirectional(G, maxres, minres)

Call and query attributes

bidirec.run() print(bidirec.path) print(bidirec.totalcost) print(bidirec.consumedresources) ```

For more details see the Python API

Cpp

```cpp

include "bidirectional.h"

namespace bidirectional {

void wrap() { // Init const std::vector maxres = {4.0, 20.0}; const std::vector minres = {1.0, 0.0}; const int numbervertices = 5; const int numberedges = 5; auto bidirectional = std::makeunique( numbervertices, numberedges, 0, 4, maxres, min_res);

// Populate graph bidirectional->addNodes({0, 1, 2, 3, 4}); bidirectional->addEdge(0, 1, 0.0, {1, 2}); bidirectional->addEdge(1, 2, 0.0, {1, 0.3}); bidirectional->addEdge(2, 3, -10.0, {1, 3}); bidirectional->addEdge(2, 4, 10.0, {1, 2}); bidirectional->addEdge(3, 4, 0.0, {1, 10});

// Run and query attributes bidirectional->run();

auto path = bidirectional->getPath(); auto res = bidirectional->getConsumedResources(); auto cost = bidirectional->getTotalCost(); }

} // namespace bidirectional ```

C

```csharp DoubleVector maxres = new DoubleVector(new List() {4.0, 20.0}); DoubleVector minres = new DoubleVector(new List() {0.0, 0.0}); int numbervertices = 5; int numberedges = 5; BiDirectionalCpp alg = new BiDirectionalCpp(numbervertices, numberedges, 0, 4, maxres, minres);

// Populate graph alg.addNodes(new IntVector(new List() {0, 1, 2, 3, 4})); alg.addEdge(0, 1, -1.0, new DoubleVector(new List() {1, 2})); alg.addEdge(1, 2, -1.0, new DoubleVector(new List() {1, 0.3})); alg.addEdge(2, 3, -10.0, new DoubleVector(new List() {1, 3})); alg.addEdge(2, 4, 10.0, new DoubleVector(new List() {1, 2})); alg.addEdge(3, 4, -1.0, new DoubleVector(new List() {1, 10})); alg.setDirection("forward");

// Run and query attributes alg.run();

IntVector path = alg.getPath(); DoubleVector res = alg.getConsumedResources(); double cost = alg.getTotalCost(); ```

Examples

• vrpy : External vehicle routing framework which uses cspy to solve different variants of the vehicle routing problem using column generation. Particulatly, see subproblem_cspy.py.
• jpath : Simple example showing the necessary graph adptations and the use of custom resource extension functions.

Building

Docker

Using docker, docker-compose is the easiest way.

To run the tests first, clone the repository into a path in your machine ~/path/newfolder by running

none git clone https://github.com/torressa/cspy.git ~/path/newfolder

Running the Cpp tests

cd ~/path/newfolder/tools/dev ./build

Running the Python tests

cd ~/path/newfolder/tools/dev ./build -c -p

Locally

Requirements:

• CMake (>=v3.14)
• Standard C++ toolchain
• Python (>=3.6)

Then use the wrapper Makefile e.g. make in the root dir runs the unit tests

License

This project is licensed under the MIT License - see the LICENSE.txt file for details.

Contributing

Issues

If you find a bug or there are some improvements you'd like to see (e.g. more algorithms), please raise a new issue with a clear explanation.

Contributing to the Software

When contributing to this repository, please first discuss the change you wish to make via an issue or email. After that feel free to send a pull request.

Pull Request Process

• If necessary, please perform documentation updates where appropriate (e.g. README.md, docs and CHANGELOG.md).
• Increase the version numbers and reference the changes appropriately. Note that the versioning scheme used is based on Semantic Versioning.
• Wait for approval for merging.

Seeking Support

If you have a question or need help, feel free to raise an issue explaining it.

Alternatively, email me at torressa at tutanota.com.

Citing

If you'd like to cite this package, please use the following bib format:

none @article{torressa2020, doi = {10.21105/joss.01655}, url = {https://doi.org/10.21105/joss.01655}, year = {2020}, publisher = {The Open Journal}, volume = {5}, number = {49}, pages = {1655}, author = {{Torres Sanchez}, David}, title = {cspy: A Python package with a collection of algorithms for the (Resource) Constrained Shortest Path problem}, journal = {Journal of Open Source Software} }