Reinforcement learning based dispatching solutions in semiconductor manufacturing: a literature review on validation and deployment

AI-based scheduling methods have been shown to be highly effective in increasing the overall efficiency of production facilities such as semiconductor wafer fabrication in academic works. Especially RL has been demonstrated to be a powerful tool for global optimization. However, the public body of knowledge lacks validation and deployment approaches as well as guidelines for implementation in real fabrication facilities. We conduct a structured literature review to identify the key challenges regarding validation and deployment that are mentioned in literature. Next, we provide a summary of these challenges and introduce solution methods, as well as guidelines for implementation. As a practical example, we propose a Machine Learning pipeline for continuous deployment, as well as an architecture that combines classical dispatching and scheduling methods with RL.