Paper: https://arxiv.org/abs/2307.12136
Solving the 3L-CVRP (as proposed by Gendreau et al.) with reinforcement learning (6-month MSc Thesis)
We develop a reinforcement learning model to solve the three-dimensional loading capacitated vehicle routing problem in linear time. Current operations research methods suffer from non-linear scaling with increasing problem size and are therefore bound to limited geographic areas in order to compute results in time for day to day operations. This only allows for local optima in routing and leaves global optimization potential untouched. While the three-dimensional loading capacitated vehicle routing problem has been studied extensively in operations research, no publications on solving the problem with reinforcement learning exist. We demonstrate the linear time scaling of our reinforcement learning model with computational experiments and benchmark our routing performance against state-of-the-art methods. The model performs within an average gap of 3.83% to 7.65% of the established methods. Our model, therefore, not only represents a promising first step towards large scale logistics optimization with reinforcement learning but also lays the foundation for this stream of research.
@misc{schoepf20233LCVRP,
title={Unlocking Carbon Reduction Potential with Reinforcement Learning for the Three-Dimensional Loading Capacitated Vehicle Routing Problem},
author={Stefan Schoepf and Stephen Mak and Julian Senoner and Liming Xu and Netland Torbjörn and Alexandra Brintrup},
year={2023},
eprint={2307.12136},
archivePrefix={arXiv},
primaryClass={cs.LG}
}