Application of Reinforcement Learning to Improve Finished Goods Inventory Management: A Systematic Review

Efficient inventory management is crucial for ensuring financial stability and operational efficiency, especially in dynamic supply chains characterized by demand variability and logistical complexity. Reinforcement Learning (RL), a branch of Artificial Intelligence, has emerged as a promising approach to optimize inventory control in uncertain environments. Unlike traditional inventory management techniques, RL dynamically learns decision-making strategies through interaction with the environment, adjusting stock levels based on real-time feedback. This systematic review examines the application of RL in finished goods inventory management by analyzing 57 relevant studies. The findings highlight RL’s potential to enhance demand forecasting, reduce stockouts and excess inventory, and improve overall supply chain resilience. However, practical challenges such as computational complexity, data-sharing limitations, and integration with legacy systems hinder widespread adoption. The review identifies key research opportunities, including the integration of RL with emerging technologies such as IoT, blockchain, and digital twins, as well as the development of hybrid models that combine RL with traditional improvement methods. These insights contribute to advancing the field, offering a roadmap for future research and industrial applications.