Home // ICDS 2022, The Sixteenth International Conference on Digital Society // View article

Machine Learning Method Within the Context of a Socially Aware Solution for Vehicle Routing Problems

Authors:
Robert Maurer
Andreas Johannsen

Keywords: Vehicle Routing Problem; Machine Learning; Reinforcement Learning; last mile logistics

Abstract:
The market for courier, express and parcel services has seen an immense increase in sales and relevance in times of the pandemic. Not only has the volume of shipments increased, but also the demand for social Vehicle Routing Problems (VRP) solution procedures based on modern IT solutions supporting the dispatching or routing process. This article provides an answer to social responsible and sustainable logistics services and a conceptual prototype for the practical implementation of a Machine Learning method to solve Vehicle Routing Problems (VRP) in the context of sustainable "last mile" logistics. Aspects of combinatorial optimization algorithms in the form of an ant algorithm were used to support the applied Machine Learning (ML) system. The prototype is based on the "Reinforcement Learning" system and uses "REINFORCE with baseline" as the algorithm for updating a parameterized policy. A benchmark analysis provides a comparison between the prototype and Google-OR, as a representative for combinatorial optimization algorithms, applied in two examples. The results show that Google-OR prevails over the prototype in terms of solution quality, but the prototype convinces in runtime and automatism. In addition, the applied Machine Learning context results only in minor advantages for small to medium sized logistic domains, as they do not generate enough data. Hence, using Machine Learning methods for Vehicle Routing problems is recommended for a larger stop volume in urban areas. Furthermore, the prototype represents an alternative solution to outsourcing to third party providers and provides an approach to gain a competitive advantage for solving Vehicle Routing Problems.

Pages: 1 to 5

Copyright: Copyright (c) IARIA, 2022

Publication date: June 26, 2022

Published in: conference

ISSN: 2308-3956

ISBN: 978-1-61208-981-2

Location: Porto, Portugal

Dates: from June 26, 2022 to June 30, 2022