LLM-Guided Multi-Agent System for Natural Language-Based Robot Navigation

Kurzfassung

Natural language-driven robot navigation has the potential to make human-robot interactions more intuitive. This paper presents an innovative approach that integrates a Large Language Model (LLM) with a Multi-Agent System (MAS) to enable autonomous robot navigation in response to verbal commands. We use GPT-4o for interpreting user commands, LangChain and LangGraph for MAS-based decision-making, and Rapidly-Exploring Random Tree (RRT) for path planning. The system is simulated in Webots, demonstrating its adaptability in various environments. Our results show that the integration of LLM and MAS enhances decision-making efficiency and enables flexible, real-time path adjustments.