Reinforcement Learning with Large Language Model for Hybrid Disassembly Lines in Remanufacturing Contexts

Large language models (LLM), ChatGPT is making substantial impact across various fields. This study for the first time presents a novel approach to the hybrid disassembly line balancing problem using LLM and reinforcement learning algorithms in remanufacturing contexts. The problem is divided into two sub-stages. LLM is innovatively used to capture a disassembly sequence well in the first stage, while reinforcement learning is utilized to address the problem in the second stage. Upon comparing the performance with and without LLM, the proposed approach significantly reduces the trial-and-error space and achieves faster convergence to achieve the desired solution. Future work of this study is also discussed.