Human-Understandable Communication for Industrial Multi-Robot Systems Using LLMs

The coordination of heterogeneous robotic systems presents significant challenges in industrial automation, particularly in enabling seamless communication, efficient task allocation, and adaptability across diverse robot types. This research proposes a framework that integrates Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) to enhance multi-robot collaboration. Unlike existing approaches, this framework enables natural language-based communication among robots and camera agents while leveraging MQTT-based messaging for scalable, event-driven communication. This approach reduces reliance on predefined control protocols and enhances real-time adaptability in multi-agent environments. The LLM dynamically generates low-level task sequences, assigns tasks in real time, and facilitates adaptive multi-agent communication. A prototype evaluation focuses on task delegation, analyzing how the LLM interprets user commands, generates task allocations, and assigns agents for execution. The results demonstrate that the proposed framework effectively automates task allocation and ensures precise delegation to the appropriate agents, supporting scalable and resilient automation in next-generation smart manufacturing.