Research question
How should production adapt when downstream retail capabilities are uncertain and distribution decisions shape what can actually be sold?
Approach and decision mechanism
The paper connects adaptive production control with distribution planning in a hierarchical reinforcement learning framework. Transaction feedback helps the production controller learn about retail capabilities. The lower-level distribution policy forms part of the environment in which production decisions are evaluated.
- Production decisionChoose output using the inventory state and learned value of production.
- Distribution decisionPlan downstream distribution using the inventory made available.
- Feedback & adaptationUse realized transactions and costs to update production control.
Inventory links the two decision levels. Holding and opportunity costs shape the production objective; an exponential inventory penalty represents pressure from limited storage capacity.
Evidence and scope
A steel-industry case study uses real-world transaction data to evaluate production control under different retail strategies.
The evidence is specific to the studied steel setting and modeled retail strategies. Transfer to another operation requires checking its data, constraints, and downstream behavior.
Read the publication recordCite this paper
Deng, Y., Chow, A. H. F., Yan, Y., Su, Z., Zhou, Z., & Kuo, Y.-H. (2025). Hierarchical production control and distribution planning under retail uncertainty with reinforcement learning. International Journal of Production Research, 63(12), 4504–4522. https://doi.org/10.1080/00207543.2025.2452386
View BibTeX
@article{deng2025production,
title = {Hierarchical production control and distribution planning under retail uncertainty with reinforcement learning},
author = {Yang Deng and Andy H. F. Chow and Yimo Yan and Zicheng Su and Zhili Zhou and Yong-Hong Kuo},
journal = {International Journal of Production Research},
year = {2025},
volume = {63},
number = {12},
pages = {4504--4522},
doi = {10.1080/00207543.2025.2452386},
url = {https://doi.org/10.1080/00207543.2025.2452386}
}