Reinforcement Learning Approach in Supply Chain Management

Abstract

Supply Chain Management (SCM) is one of the emerging areas that involve complex and challenging systems with multiple stakeholders and activities, ranging from demand forecasting to logistics management. The occurrence of complexity, demand volatility, cost-effectiveness, and information-sharing issues poses significant obstacles to efficient SCM functioning. In addition to these, the unprecedented global issues have further disrupted global supply chains. As such, there is an immediate need for an effective solution than ever before to handle these problems. In this context, Machine Learning (ML) techniques, such as Artificial Neural Networks (ANNs), Time-Series Analysis, Deep Learning, and Reinforcement Learning (RL), have been researched and used to tackle these challenges in SCM. Among these, RL has emerged as a promising technology to handle SCM activities such as inventory management, production scheduling, transportation and logistics optimization, demand forecasting, supplier selection and management, and risk management. RL is a type of machine learning in which an agent learns to interact with its environment by taking actions and receiving feedback in the form of rewards or penalties. It was found that RL can adapt to changing environments, learn from experience, and optimize its actions to achieve long-term goals. These features make RL well-suited for handling the challenges faced by SCM, where demand and supply dynamics can change rapidly, and decisions must be made based on incomplete or uncertain information. Our review focuses on the challenges in SCM and the role of RL to address the same. Here, the adoption of RL in SCM is discussed with the help of a case study. The obtained outcomes and difficulties to the adoption of RL in SCM are investigated. Finally, a comparative study of various other ML techniques with RL is included to observe the efficacy and easy implementation.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 20 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.