Periodic-Review Joint Replenishment Policy using Multi-Agent Reinforcement Learning

• Item type: SIG Technical Reports(1)
• 公開日: 2020-09-22
• タイトル: Periodic-Review Joint Replenishment Policy using Multi-Agent Reinforcement Learning
• タイトル / 言語: en
• タイトル: Periodic-Review Joint Replenishment Policy using Multi-Agent Reinforcement Learning
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Graduate School of Engineering, The University of Tokyo
• 著者所属: Graduate School of Engineering, The University of Tokyo
• 著者所属: Graduate School of Engineering, The University of Tokyo
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, The University of Tokyo
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, The University of Tokyo
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, The University of Tokyo
• 著者名: Hiroshi, Suetsugu; Yoshiaki, Narusue; Hiroyuki, Morikawa
• 著者名(英): Hiroshi, Suetsugu; Yoshiaki, Narusue; Hiroyuki, Morikawa
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: A periodic-review joint replenishment problem is considered. In literature, can-order and modified periodic-review policies have been proposed, and either of them cannot always outperform the other depending on the demand characteristics. In addition, whereas there are many types of joint-replenishment cost structures in practical settings, most studies have assumed the fixed joint-replenishment costs, and, for the periodic-review system, no study has been conducted to incorporate the practical cost structures into the existing heuristic approach. The purpose of this paper is to propose a multi-agent reinforcement learning-based solution for a joint replenishment problem, which can be used for problems with several demand settings, and be applied for various cost structures with minor modification. Our numerical experiments demonstrate that the performance of our proposed agent is equal or greater than that of the existing heuristic policies, that are can-order, and modified periodic policies.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: A periodic-review joint replenishment problem is considered. In literature, can-order and modified periodic-review policies have been proposed, and either of them cannot always outperform the other depending on the demand characteristics. In addition, whereas there are many types of joint-replenishment cost structures in practical settings, most studies have assumed the fixed joint-replenishment costs, and, for the periodic-review system, no study has been conducted to incorporate the practical cost structures into the existing heuristic approach. The purpose of this paper is to propose a multi-agent reinforcement learning-based solution for a joint replenishment problem, which can be used for problems with several demand settings, and be applied for various cost structures with minor modification. Our numerical experiments demonstrate that the performance of our proposed agent is equal or greater than that of the existing heuristic policies, that are can-order, and modified periodic policies.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10505667
• 書誌情報: 研究報告数理モデル化と問題解決（MPS）; 巻 2020-MPS-130, 号 1, p. 1-6, 発行日 2020-09-22
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8833
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会