<![CDATA[Modern production systems are increasingly complex, requiring scheduling methods that can handle dynamic environments, diverse constraints, and large-scale operations. Traditional approaches often lack flexibility, while machine learning (ML)–based methods, despite their potential, still face limitations related to scalability, generalizability, interpretability, and computational efficiency. This study presents a systematic literature review of 77 primary studies published between 2014 and 2024, conducted in accordance with the Kitchenham and Charters framework. The review analyzes major research outlets, commonly applied ML techniques, reported performance, and proposed enhancements. Reinforcement learning, particularly deep reinforcement learning, dominates the literature, with methods such as Q-Learning, Deep Q-Networks, and Proximal Policy Optimization showing promise for dynamic scheduling. However, challenges remain regarding convergence speed, data requirements, reward design, and real-time adaptability. Future research should focus on scalable, adaptive, interpretable models and tighter integration with real-time data and Industry 4.0 systems. ]]>
Copyrights © 2026
