Deep reinforcement learning for inventory control: a roadmap

Robert N. Boute; Joren Gijsbrechts; Willem van Jaarsveld; Nathalie Vanvuchelen

doi:10.1016/j.ejor.2021.07.016

Deep reinforcement learning for inventory control: a roadmap

Robert N. Boute^*
, Joren Gijsbrechts
, Willem van Jaarsveld
, Nathalie Vanvuchelen

^*Autor correspondente para este trabalho

Resultado de pesquisa › revisão de pares

167 Citações (Scopus)

36 Transferências (Pure)

Resumo

Deep reinforcement learning (DRL) has shown great potential for sequential decision-making, including early developments in inventory control. Yet, the abundance of choices that come with designing a DRL algorithm, combined with the intense computational effort to tune and evaluate each choice, may hamper their application in practice. This paper describes the key design choices of DRL algorithms to facilitate their implementation in inventory control. We also shed light on possible future research avenues that may elevate the current state-of-the-art of DRL applications for inventory control and broaden their scope by leveraging and improving on the structural policy insights within inventory research. Our discussion and roadmap may also spur future research in other domains within operations management.

Idioma original	English
Páginas (de-até)	401-412
Número de páginas	12
Revista	European Journal of Operational Research
Volume	298
Número de emissão	2
DOIs	https://doi.org/10.1016/j.ejor.2021.07.016
Estado da publicação	Publicado - 16 abr. 2022

ODS da ONU

Este resultado contribui para o(s) seguinte(s) Objetivo(s) de Desenvolvimento Sustentável

ODS 9 Indústria, inovação e infraestrutura

Acesso ao documento

10.1016/j.ejor.2021.07.016Licença:

37747702Final published version, 761 KBLicença:

http://hdl.handle.net/10400.14/34576Licença:

Outros arquivos e links

Link to publication in Scopus

CUBE: Católica Lisbon Research Unit in Business and Economics
Machado, F. (PI), Aranda, A. M. (Investigador), Costa, A. I. D. A. (Investigador), Martins, A. P. (Investigador), Bernard, A. (Investigador), Fidalgo, A. (Investigador), Reis, C. (Investigador), Abecassis-Moedas, C. (Investigador), Santos, C. (Investigador), Tran, D. (Investigador), Fernandes, D. (Investigador), Bonfim, D. (Investigador), Schliephake, E. (Investigador), Kalogirou, F. (Investigador), Reis, F. (Investigador), Santos, F. (Investigador), Machado, F. (Investigador), Sguera, F. (Investigador), Cerqueiro, G. (Investigador), Reis, H. (Investigador), Barreto, I. (Investigador), Silva, J. (Investigador), Salvado, J. C. (Investigador), Neves, J. C. D. (Investigador), Gijsbrechts, J. (Investigador), Stahl, J. (Investigador), Guedes, J. C. (Investigador), Faias, J. (Investigador), Wagner, L. (Investigador), Correia, I. (Investigador), Saldanha, M. F. (Investigador), Fontes, J. C. (Investigador), Modesto, L. (Investigador), Meira, M. (Investigador), Matos, M. G. D. (Investigador), Gouveia, M. (Investigador), Kozeniauskas, N. (Investigador), Bertani, N. (Investigador), Couzoff, P. (Investigador), Barroso, P. (Investigador), Encarnação, P. (Investigador), Raposo, P. (Investigador), Teles, P. (Investigador), Bastos, W. (Investigador), Bohnsack, R. (Investigador), Vale, R. C. D. (Investigador), Ferreira, R. (Investigador), Lloyd-Braga, T. (Investigador), Ramus, T. (Investigador), Colaço, V. H. (Investigador), Cavicchini, A. (Investigador), Parada, P. (Investigador), Pieters, R. (Investigador), Pinto, F. S. (Investigador), Colaço, V. H. (Investigador) & Venter, Z. (Investigador)
1/01/20 → 31/12/25
Projeto

Citação

@article{235e687c8cfd4479a87e1b10e6ee391f,

title = "Deep reinforcement learning for inventory control: a roadmap",

abstract = "Deep reinforcement learning (DRL) has shown great potential for sequential decision-making, including early developments in inventory control. Yet, the abundance of choices that come with designing a DRL algorithm, combined with the intense computational effort to tune and evaluate each choice, may hamper their application in practice. This paper describes the key design choices of DRL algorithms to facilitate their implementation in inventory control. We also shed light on possible future research avenues that may elevate the current state-of-the-art of DRL applications for inventory control and broaden their scope by leveraging and improving on the structural policy insights within inventory research. Our discussion and roadmap may also spur future research in other domains within operations management.",

keywords = "Inventory management, Machine learning, Neural networks, Reinforcement learning",

author = "Boute, \{Robert N.\} and Joren Gijsbrechts and Jaarsveld, \{Willem van\} and Nathalie Vanvuchelen",

note = "Publisher Copyright: {\textcopyright} 2021 The Author(s)",

year = "2022",

month = apr,

day = "16",

doi = "10.1016/j.ejor.2021.07.016",

language = "English",

volume = "298",

pages = "401--412",

journal = "European Journal of Operational Research",

issn = "0377-2217",

publisher = "Elsevier B.V.",

number = "2",

}

TY - JOUR

T1 - Deep reinforcement learning for inventory control

T2 - a roadmap

AU - Boute, Robert N.

AU - Gijsbrechts, Joren

AU - Jaarsveld, Willem van

AU - Vanvuchelen, Nathalie

PY - 2022/4/16

Y1 - 2022/4/16

N2 - Deep reinforcement learning (DRL) has shown great potential for sequential decision-making, including early developments in inventory control. Yet, the abundance of choices that come with designing a DRL algorithm, combined with the intense computational effort to tune and evaluate each choice, may hamper their application in practice. This paper describes the key design choices of DRL algorithms to facilitate their implementation in inventory control. We also shed light on possible future research avenues that may elevate the current state-of-the-art of DRL applications for inventory control and broaden their scope by leveraging and improving on the structural policy insights within inventory research. Our discussion and roadmap may also spur future research in other domains within operations management.

AB - Deep reinforcement learning (DRL) has shown great potential for sequential decision-making, including early developments in inventory control. Yet, the abundance of choices that come with designing a DRL algorithm, combined with the intense computational effort to tune and evaluate each choice, may hamper their application in practice. This paper describes the key design choices of DRL algorithms to facilitate their implementation in inventory control. We also shed light on possible future research avenues that may elevate the current state-of-the-art of DRL applications for inventory control and broaden their scope by leveraging and improving on the structural policy insights within inventory research. Our discussion and roadmap may also spur future research in other domains within operations management.

KW - Inventory management

KW - Machine learning

KW - Neural networks

KW - Reinforcement learning

UR - https://www.scopus.com/pages/publications/85111846139

U2 - 10.1016/j.ejor.2021.07.016

DO - 10.1016/j.ejor.2021.07.016

M3 - Review article

AN - SCOPUS:85111846139

SN - 0377-2217

VL - 298

SP - 401

EP - 412

JO - European Journal of Operational Research

JF - European Journal of Operational Research

IS - 2

ER -

Deep reinforcement learning for inventory control: a roadmap

Resumo

ODS da ONU

Acesso ao documento

Outros arquivos e links

Impressão digital

Projetos

CUBE: Católica Lisbon Research Unit in Business and Economics

Citação