Some theoretical backgrounds for reinforcement learning model of supply chain management under stochastic demand

摘要

Integrated and collaborative decision-making systems have been increasingly employed by all companies involved in the supply chain. However, steadfast fact-based decision-making tools are still difficult to get as far as these require adapted methodology. This paper studies the sustainable management problem of a two-echelon supply chain in which a supplier has a limited production capacity to satisfy the quality and quantity requirements of a retailer under stochastic demand. We motivate the model selection, which can be used in reinforcement learning, as well as sustainability constraints for a one-product manufacturing system taking into account production, maintenance, quality, and inventory operations and explore the supplier’s production behavior with a required quality level. Pursuing the profit maximization, formed as the difference between the sales revenue and the expenditures on the abovementioned operations, we determine and characterize the optimal conditions for the supply chain management in the presence of the stochastic demand and quality requirements. Our study shows that the stochastic formulation of the optimal control problem and its solution allow taking into account the short-range and long-range dependences exhibited by consumers’ behavior. In this context, we characterize the optimal policy for the supplier and the retailer. Finally, we provide managerial insights concerning the sustainable coordination conditions of the supply chain under and quality requirements.