PyVRP is a Python package implementing a state-of-the-art Vehicle Routing Problem (VRP) solver, achieving top results in competitions and offering flexibility for researchers and practitioners.
Abstract
Introduction:
PyVRP introduced as a high-performance VRP solver package.
Authors from various departments in the Netherlands.
PyVRP Features:
Implements Hybrid Genetic Search algorithm for VRPTW.
Combines Python's flexibility with C++'s performance.
Achieved 1st place in 2021 DIMACS VRPTW Challenge.
Usage and Installation:
Available under MIT license, easily installable via PyPI using pip install pyvrp.
Allows customization of algorithms using Python, supports additional VRP variants.
Technical Implementation:
Core components include genetic algorithm, population management, local search educator.
Detailed explanation of Hybrid Genetic Search algorithm implementation.
Benchmark Results:
Achieved state-of-the-art results on CVRP and VRPTW benchmarks.
Conclusion:
PyVRP aims to improve the state-of-the-art in VRP solving by providing an easy-to-use, extensible solver.