REAL TIME INVENTORY OPTIMIZATION SYSTEM FOR STREAMLINED SUPPLY CHAIN MANAGEMENT

ABSTRACT The proposed study is the implementation of Real-Time Inventory Optimization System that would improve the efficiency of the supply chain by utilizing Actor-Critic Deep Reinforcement Learning. Through Ray RLlib, the devised system will be able to solve several typical problems, including unbalanced stock supply, slow reaction to the changes in demand, and the necessity to spend many resources. The framework allows dynamic decision-making using real-time streams of data by modelling nodes of the supply chains as intelligent agents. The actor-critic architecture enables the context of the continuous learning and policy improvement and the use of Ray RLlib allows conducting the training in multiple environments in a distributed and scalable manner. The experimental results have shown that the proposed system has an important impact on the reduction of stockout rates and holding costs and increment of overall service levels when compared to traditional rule-based and standard reinforcement learning based approaches. The system is responsive to the dynamic demand and supply conditions and this is a good solution to contemporary inventory management. This study highlights the possibility of intelligent real time optimization method in streamlining and responsive supply chain operations. Keywords: Real-time inventory, supply chain optimization, actor-critic reinforcement learning, Ray RLlib, intelligent inventory control Type of paper: Research Paper