To address the difficulty and high cost of inventory control in the environment of uncertain demand and supply delays, an inventory control method based on the improved twin delayed deep deterministic policy gradient (TD3) is proposed. Firstly, the inventory control is abstracted as a Markov decision process with dual objectives of service level maximization and cost minimization, which serves as the training environment for the TD3 algorithm. Secondly, the prioritized experience replay mechanism is adopted to improve the sampling efficiency of the TD3 algorithm, and the long short-term memory (LSTM) is integrated into the multi-layer perceptron of the TD3 algorithm to optimize the network structure. Finally, the TD3 algorithm is used to interact with the environment to optimize both cost and service level in inventory control. Experimental results demonstrate that the inventory control cost of the proposed method is 22.2% lower than that of the original TD3 algorithm when the service level threshold is reached.
PUTTEMANSI, CARISA, BRAEKERSK. The integration of location and inventory decisions in supply chain networks: a literature review and future prospects[J]. Transportation Research Part E: Logistics and Transportation Review, 2025,195:No.103970.
[2]
ZHANGS, HUANGK, YUANY F. Spare parts inventory management: a literature review[J]. Sustainability, 2021,13(5):No.2460.
[3]
WANGJ, ZHUX Y. Joint optimization of condition-based maintenance and inventory control for a k-out-of-n: F system of multi-state degrading components[J]. European Journal of Operational Research, 2021,290(2):514-529.
KHEDRA M, RANIS S. Enhancing supply chain management with deep learning and machine learning techniques: a review[J]. Journal of Open Innovation: Technology, Market, and Complexity, 2024,10(4):No.100379.
[6]
CUARTASC, AGUILARJ. Hybrid algorithm based on reinforcement learning for smart inventory management[J]. Journal of Intelligent Manufacturing, 2023,34(1):123-149.
[7]
TIANR, LUM, WANGH P, et al. IACPPO: a deep reinforcement learning-based model for warehouse inventory replenishment[J]. Computers & Industrial Engineering, 2024,187:No.109829.
[8]
ZHOUY F, GUOK, YUC, et al. Optimization of multi-echelon spare parts inventory systems using multi-agent deep reinforcement learning[J]. Applied Mathematical Modelling, 2024,125:827-844.
[9]
余诚.基于深度强化学习的多级备件库存策略研究[D].南京:东南大学,2022:1-71.
[10]
KEMMERL, VON KLEISTH, ROCHEBOUËTD D, et al. Reinforcement learning for supply chain optimization[C]//Proceedings of the 14th European Workshop on Reinforcement Learning. [S.l.]: [s.n.], 2018:1-9.
ZHANGF J, LIJ, LIZ. A TD3-based multi-agent deep reinforcement learning method in mixed cooperation-competition environment[J]. Neurocomputing, 2020,411:206- 215.
[13]
WUJ L, WUQ M J, CHENS Y, et al. A-TD3: an adaptive asynchronous twin delayed deep deterministic for continuous action spaces[J]. IEEE Access, 2022,10:128077-128089.
[14]
ALVESJ C, SILVAD M D, MATEUSG R. Applying and comparing policy gradient methods to multi-echelon supply chains with uncertain demands and lead times[C]//Proceedings of the 20th International Conference on Artificial Intelligence and Soft Computing. Cham, Switzerland: Springer, 2021:229-239.
[15]
STRANIERIF, FADDAE, STELLAF. Combining deep reinforcement learning and multi-stage stochastic programming to address the supply chain inventory management problem[J]. International Journal of Production Economics, 2024,268:No.109099.