Decomposition Strategies for Vehicle Routing Heuristics

This archive is distributed in association with the INFORMS Journal on Computing under the MIT License.

The software and data in this repository are a snapshot of the software and data used in the paper Decomposition Strategies for Vehicle Routing Heuristics by A. Santini, M. Schneider, T. Vidal, and D. Vigo. The snapshot is based on alberto-santini/cvrp-decomposition v0.1 in the development repository. Users are advised to use cvrp-decomposition, which might contain improvements and fixes compared to this repository.

This repository presents a proof-of-concept for including instance decomposition within two popular metaheuristics for the Capacitated Vehicle Routing Problem (CVRP): * The Adaptive Large Neighbourhood Search of Stefan Ropke * The Hybrid Genetic Algorithm of Thibaut Vidal

We propose several decomposition techniques, which we describe in detail in our paper:

Cite

bib @article{cvrp_decomposition, title={Decomposition strategies for vehicle routing heuristics}, author={Santini, Alberto and Schneider, Michael and Vidal, Thibaut and Vigo, Daniele}, year=2022, journal={{INFORMS Journal on Computing}}, pubstate={to appear} }

You can cite this repository via GitHub:

Cite

DOI: 10.1287/ijoc.2023.1288.cd bib @article{cvrp_decomposition_2022, title={{Repository cvrp-decomposition} Version v2022.0048}, author={Ropke, Stefan and Vidal, Thibaut and Santini, Alberto}, year=2022, doi={10.1287/ijoc.2023.1288.cd}, url={https://github.com/INFORMSJoC/2022.0048} }

Authors

The original ALNS code is by Stefan Ropke. Alberto Santini then edited Stefan's code and implemented instance decomposition. The original paper in which Stefan introduced ALNS is the following:

bib @article{ropke_2006, title={An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows}, author={Ropke, Stefan and Pisinger, David}, year=2006, journal={Transportation Science}, volume=40, number=4, pages={455--472}, doi={10.1287/trsc.1050.0135} }

The original HGS code is by Thibaut Vidal. Thibaut also released his code in repository vidalt/HGS-CVRP and through the below paper. Alberto Santini then edited Thibaut's code and implemented instance decomposition.

bib @article{vidal_2022, title={Hybrid genetic search for the CVRP: Open-source implementation and SWAP\textsupersctipt{*} neighborhood}, author={Vidal, Thibaut}, year=2022, journal={Computers \& Operations Research}, volume=140, doi={10.1016/j.cor.2021.105643} }

Usage

HGS

The preferred way to build the HGS programme is through cmake. We provide a CMakeList.txt file, which allows to build the executable as follows:

$ git clone https://github.com:alberto-santini/cvrp-decomposition.git $ cd cvrp-decomposition $ mkdir build $ cd build $ cmake -DCMAKE_BUILD_TYPE=Release .. $ cmake $ make

ALNS

The ALNS solver is distributed in binary form in folder bin. The executable (for Linux x86-64) is palns-cvrp. Running the executable file without any parameter displays the available parameters:

-f <arg> : Load problem given by <arg> -t <arg> : problem type to load: 1 = JFC CVRP (default) 2 = GWKC 3 = TSPLIB format (used by exact algs.) 4 = Tailard 5 = TSPLIB format, but #vehicles is free (for Uchoa et al. instances) -p <arg> : Use parameter file given by <arg> -P <arg> : override parameters from par file with <arg> -n <arg> : number of threads to use. -v <arg> : Number of vehicles (for type 3 instances) -r <arg> : Number of retries -h : This help -H : Add header line to output file -o <arg> : prefix of summary files

Instances

Instances contained in folder data are part of the so-called "Uchoa dataset". We refer to the following paper for further information on their generation and characteristics:

bib @article{uchoa_2017, title={New benchmark instances for the capacitated vehicle routing problem}, author={Uchoa, Eduardo and Pecin, Diego and Pessoa, Artur and Poggi, Marcus and Vidal, Thibaut and Subramanian, Anand}, year=2017, journal={European Journal of Operational Research}, volume=257, number=3, pages={845--858}, doi={10.1016/j.ejor.2016.08.012} }

Results

Folder results contains the figures used in the paper.

Acknowledgements

We are extremely grateful to Stefan Ropke for sharing with us his code for ALNS applied to the CVRP.

License

This software is released under the MIT license, which we report in file LICENSE.