This research presents a flexible framework that enables robots to adapt their movement trajectories based on natural language instructions.

• Leverages pre-trained LLMs to modify trajectory waypoints through code generation
• Works with existing motion planners (RRT, A-star) and human demonstrations
• Successfully tested across manipulators, aerial vehicles, and ground robots
• Enables more intuitive human-robot interaction without specialized training

This breakthrough matters because it creates a more accessible interface between humans and robots, allowing non-technical users to guide robotic behavior using natural language rather than specialized programming knowledge.

Trajectory Adaptation using Large Language Models