evovrp

Solving multi-depot vehicle routing problem (MDVRP) using evolutionary algorithms.

Requirements

• Python 3.6+
• Pip

Dependencies in order for the project to run are:

numpy, imageio, matplotlib and NiaPy==2.0.0rc4

You can install them using the setup.py script:

$ git clone https://github.com/mpinta/evovrp
$ cd evovrp
$ python setup.py install

Usage

In main.py import any evolutionary algorithm from the NiaPy library and set the input parameters of the main function:

• file_name - location of a dataset file,
• algorithm - imported evolutionary algorithm from the NiaPy library,
• iterations - number of iterations,
• population_size - number of instances inside one generation,
• phenotype_coding - genotype-to-phenotype coding method; method.Method.FIRST for the first and method.Method.SECOND for the second method.

Example

The following code solves p01 MDVRP case from the C-mdvrp dataset using a genetic algorithm of 10 generations, population size of 5 and first genotype-to-phenotype coding method:

from NiaPy.algorithms.basic.ga import GeneticAlgorithm

if __name__ == '__main__':
    main('C-mdvrp/p01', GeneticAlgorithm, 50, 5, method.Method.FIRST)

You can run the program with following commands:

$ cd evovrp/evovrp
$ python main.py

Output

Once the evaluation finishes, program returns information about the instance with the best fitness value into the console. It creates .png format images and associated .gif animations for each instance, solving the MDVRP case. Program also creates gif animation of best instances from each generation and a bar graph, showing all fitness values through generations.

Datasets

Used datasets are taken from the University of Málaga - Networking and Emerging Optimization Groups website about vehicle routing problem:

• Multi-depot VRP Instances
• Multiple Depot VRP with Time Windows Instances

Both used datasets are designed by Cordeau.

Publications

The code was originally used in the following publications:

Pintarič Matic, (2019).
Reševanje problema usmerjanja vozil s pomočjo evolucijskih algoritmov.
Maribor: University of Maribor, Faculty of Electrical Engineering and Computer Science.

Pintarič Matic, Karakatič Sašo, (2019).
Solving multi-depot vehicle routing problem with particle swarm optimization.
In: Iztok Fister Jr., Andrej Brodnik, Matjaž Krnc and Iztok Fister (eds.). Proceedings of the 2019 6th Student Computer Science Research Conference - StuCoSReC, (pp. 53-56).
Koper: University of Primorska Press.

Disclaimer

The goal of the project is not optimization of evolutionary algorithms, but the use of different algorithms to solve the multi-depot vehicle routing problem.