Treffer: Dynamic scheduling of wafer batch processing machines via reinforcement learning enhanced by expert-guided lightweight LLM.

Wafer batch processing machines scheduling is critical to the efficiency of semiconductor manufacturing, where highly dynamic task arrivals, complex constraints, and reentrant processing pose significant challenges. To tackle these challenges, this study introduces a novel multi-agent collaborative reinforcement learning (RL) framework enhanced by a lightweight large language model (LLM). The proposed framework incorporates two dedicated agents–a batch formation agent and a batch assignment agent–specifically designed to optimise scheduling decisions in dynamic and constraint-rich production environments through collaborative interaction. A lightweight LLM is integrated as an auxiliary module to provide semantic action guidance through a two-stage fine-tuning process that combines expert knowledge and RL experience, enabling the agents to generate more effective and context-aware policies. Furthermore, a Transformer-based architecture is employed to fuse dynamic information across agents, facilitating coordination and joint decision-making. Experimental results demonstrate that the proposed framework significantly improves scheduling performance, reducing average task flow time by over 20% on benchmark cases and by more than 25% compared to rule-based and heuristic methods in real-world scenarios, while also enhancing equipment utilisation. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of Production Research is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)