Retrieval-Augmented LLM-Driven Multi-Agent Optimization Framework for Intelligent Manufacturing Scheduling

作者：Xiang Li, Xiaolong Zhou, Jinbo Li, Qingyue Di, Lei Jing, Bin Fan · 年份：2025 · DOI：10.1109/hpcc67675.2025.00214 · 被引用次数：2 · 研究领域：Scheduling and Optimization Algorithms、Digital Transformation in Industry、Metaheuristic Optimization Algorithms Research

Optimizing production scheduling in complex manufacturing environments is essential for improving efficiency and maximizing resource utilization. However, in some situations, traditional optimization methods and current large language models (LLMs) have limitations. We propose a Retrieval-Augmented LLM-driven Multi-Agent Optimization (RALMAO) framework to overcome these shortcomings. This novel architecture integrates Retrieval-Augmented Generation with LLM-driven multi-agent systems to incorporate domainspecific knowledge, which enables inter-agent collaboration and cross-system optimization in manufacturing ecosystems. Validation using real-world industrial datasets-including eye drop and apparel manufacturing-demonstrates the framework's superior performance compared to current enterprise solutions. Our study facilitates the practical application of LLMs in complex scheduling problems by providing actionable insights, which is conducive to improving the intelligent level of enterprises.