Application of multi-agent Reinforcement Learning to supply chain ordering management

Coordination in the Supply Chain Management (SCM) plays a major role in competitive advantages in today's uncertain business environments. There is strong evidence of success in the supply chain performance in cases with high coordination among echelons. The bullwhip effect is an important phenomenon of amplification of demand in supply chain. By eliminating the bullwhip effect, it is possible to increase product profitability. Reinforcement Learning (RL) is successfully applied to some dynamical and unpredictable issues. By surveying some efficient multi-agent RL mechanisms, this paper presents a multi-layer ordering control strategy by a multi-agent coordination model with RL mechanism for alleviating, effectively and efficiently, the bullwhip effect of a supply chain with multiple members and multi commodities. As an improvement to previous works using RL in supply chain ordering management domain, the method presented in this paper can be used to deal with multiple members with multi commodities in each echelon and each player. As a result, the methodology presented in this study is expected to help address issues regarding the uncertainty and complexity of deriving the maximal profit using RL technique in the stochastic supply chain with multiple echelons and multiple members with multiple commodities among supply chain members for efficient supply chain coordination.