Reward Shaping to Improve the Performance of Deep Reinforcement Learning in Perishable Inventory Management

European Journal of Operational Research, 301(2), 535-545

27 Pages Posted: 2 Apr 2021 Last revised: 14 Apr 2022

See all articles by Bram J. De Moor

Bram J. De Moor

Research Center for Operations Management, KU Leuven

Joren Gijsbrechts

Catholic University of Portugal (UCP) - Catolica Lisbon School of Business and Economics

Robert N. Boute

KU Leuven - Faculty of Business and Economics (FEB); Vlerick Business School - Operations & Technology Management Center

Date Written: September 1, 2022

Abstract

Deep reinforcement learning (DRL) has proven to be an effective, general-purpose technology to
develop 'good' replenishment policies in inventory management. We show how transfer learning from existing, well-performing heuristics may stabilize the training process and improve the performance of DRL in inventory control. While the idea is general, we specifically implement potential-based reward shaping to a deep Q-network algorithm to manage inventory of perishable goods that, cursed by dimensionality, has proven to be notoriously complex. The application of our approach may not only improve inventory cost performance and reduce computational effort, the increased training stability may also help to gain trust in the policies obtained by black box DRL algorithms.

Keywords: Inventory, Perishable inventory management, Deep reinforcement learning, Reward shaping, Transfer learning

Suggested Citation

De Moor, Bram J. and Gijsbrechts, Joren and Boute, Robert N., Reward Shaping to Improve the Performance of Deep Reinforcement Learning in Perishable Inventory Management (September 1, 2022). European Journal of Operational Research, 301(2), 535-545, Available at SSRN: https://ssrn.com/abstract=3804655 or http://dx.doi.org/10.2139/ssrn.3804655

Bram J. De Moor (Contact Author)

Research Center for Operations Management, KU Leuven ( email )

Naamsestraat 69
Leuven, B-3000
Belgium

HOME PAGE: http://www.kuleuven.be/wieiswie/en/person/00139451

Joren Gijsbrechts

Catholic University of Portugal (UCP) - Catolica Lisbon School of Business and Economics ( email )

Palma de Cima
Lisbon, 1649-023
Portugal

Robert N. Boute

KU Leuven - Faculty of Business and Economics (FEB) ( email )

Naamsestraat 69
Leuven, B-3000
Belgium

Vlerick Business School - Operations & Technology Management Center ( email )

Belgium

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
375
Abstract Views
1,130
rank
110,736
PlumX Metrics