This repository contains the code used to benchmark for a preliminary paper on heuristic algorithms for the multiagent MWLP. graph.py defines a custom graph class and algos.py defines the core algorithms in the paper. There are two test files that can be run with pytest and 100% code coverage of graph.py and algos.py can be verified with the Coverage.py module. benchmark.py contains helper code for running various benchmarks and creating some plots used in the paper. mass_benchmark.py and champaign.py have code used to create the actual benchmarks presented in the paper. alpha.py contains some testing code to test parameters for the transfer-and-swaps algorithms.
The code has been linted for readablity using flake8 and pylint and the black autoformatter was also used to aid in readability. Type hints, checked with mypy, were used in the majority of the code to also aid with readability.
The code was developed using Python 3.9.5 so as of now that is the recommended version that should be used. MatPlotLib 3.5.1, NetworkX 2.6.3, and OSMNX 1.1.2 are also needed.
We thank Pranay Thangeda for his help in identifying bugs in early versions of the code for this paper as well as providing baseline code for the Champaign simulations. We also thank Diego Cerrai for clarifying previous work on this topic and kindly sharing his recent results.