RoboMorph combines large language models with evolutionary algorithms to automate the design and optimization of modular robots, dramatically accelerating the traditionally time-consuming process.

• Represents robot designs as grammar that LLMs can understand and manipulate
• Employs innovative "best-shot prompting" to navigate complex design spaces
• Integrates reinforcement learning for control algorithm development
• Enables rapid optimization of robots for specific terrains and tasks

This research significantly advances engineering automation by reducing the computational demands of robot design while improving performance outcomes, potentially transforming manufacturing, exploration, and autonomous systems development.

RoboMorph: Evolving Robot Morphology using Large Language Models