2021 journal article
Dynamic matching with deep reinforcement learning for a two-sided Manufacturing-as-a-Service (MaaS) marketplace
MANUFACTURING LETTERS, 29, 11–14.
author keywords: Cloud manufacturing; Cyber-enabled manufacturing; Resource allocation; Two-sided matching; Dynamic and stochastic knapsack problem (DSKP); Cloud based design and manufacturing (CBDM)
TL;DR:
Empirical simulations demonstrate that DRL has considerably better performance compared to four baselines and demonstrates a learning approach for near real-time decision making for suppliers participating in a MaaS marketplace.
(via Semantic Scholar)
UN Sustainable Development Goal Categories
9. Industry, Innovation and Infrastructure
(Web of Science)
12. Responsible Consumption and Production
(Web of Science)
16. Peace, Justice and Strong Institutions
(OpenAlex)
Source: Web Of Science
Added: October 4, 2021
Suppliers registered within a manufacturing-as-a-service (MaaS) marketplace require near real time decision making to accept or reject orders received on the platform. Myopic decision-making such as a first come, first serve method in this dynamic and stochastic environment can lead to suboptimal revenue generation. In this paper, this sequential decision making problem is formulated as a Markov Decision Process and solved using deep reinforcement learning (DRL). Empirical simulations demonstrate that DRL has considerably better performance compared to four baselines. This early work demonstrates a learning approach for near real-time decision making for suppliers participating in a MaaS marketplace.