Inventory Optimization of Multi-Echelon Supply Chain Under Market Uncertainties

Efficient inventory management is crucial for businesses operating within complex supply chain networks, especially when facing unpredictable and fluctuating customer demands. Maintaining meticulous control over inventory levels is paramount for ensuring the sustained success of the enter-prise. The article presents Deep Reinforcement Learning (DRL) algorithm to optimize inventory management within complex supply chain networks. The focus is on utilizing policy based DRL algorithm, namely, Proximal Policy Optimization (PPO) models to handle uncertain demands and unpredicted lead times. Experimental results demonstrate the superiority of the PPO model through consistent profit escalation considering both deterministic and stochastic demand scenarios as compared to classical inventory policies. Findings indicate that the average profit of the PPO model increases by 14% and 63% to DQN and classical (s, S) inventory policy. The comprehensive approach outlined in the article will contribute to a deeper understanding of how these methods perform under different conditions and provide valuable insights for efficient inventory management strategies in market uncertainties.