This research introduces a novel approach where GPT-4o automatically creates agent networks that help robots adapt to changing environments.

• Collects environmental status data to generate specialized agents
• Creates interconnected agent networks based on specific conditions
• Enables robots to dynamically adjust plans as situations change
• Demonstrates practical applications in engineering and robotics

This innovation matters for engineering because it addresses a fundamental challenge in robotics: creating systems that can autonomously respond to unpredictable real-world scenarios without pre-programmed instructions.

LLM-mediated Dynamic Plan Generation with a Multi-Agent Approach