MetaGA Route Inspection Solver

The research source code for a Metaheuristic Genetic Algorithm (MetaGA) that solves route inspection arc routing problems min max k chinese postman problem (mmkcpp) and multi depot variants.

This repository uses PyGAD for easy to adopt source code and genetic algorithm parameters for further experimentation and Weights & Biases for experiment tracking and dataset versioning.

Experiments

Runs / Random Seeds

Here is a list of 30 random seeds used in my experiments.

8115,3520,8647,9420,3116,6377,6207,4187,3641,8591,3580,8524,2650,2811,9963,7537,3472,3714,8158,7284,6948,6119,5253,5134,7350,2652,9968,3914,6899,4715

Getting Started

Add bridge graph instances

git submodule add https://github.com/BryanDedeur/bridge-graph-instances

Running

Example

python main.py -i benchmarks/bmcv/C05.dat -k 0,0 -s 8115,3520,8647,9420,3116,6377 -j MMMR

Windows & VSCode Environment

It is recommended to use Anaconda to create a isolated environment with the necessary packages versions that support PyGAD since the various dependent libraries (numpy, scipy) .

1. Install Anaconda
2. Run Anaconda
3. Add Python environment version [3.X.X] that supports PyGAD (see pygad latest docs)
4. Open Terminal through Anaconda for the Environment
5. Add pip packages
   • pip install Numpy - for dependencies in pygad
   • pip install PyGAD - for genetic algorithm
   • pip install wandb - for database
6. Edit vscode python interpreter
   • CTRL + SHIFT + P
   • Type python interpreter
   • Select the conda environment
7. [Optional] Edit environment variables to support wandb