This research integrates Large Language Models (LLMs) with Reinforcement Learning (RL) to create adaptive controllers for Autonomous Underwater Vehicles (AUVs) in challenging ocean environments.

• Leverages LLMs to jointly optimize controller parameters and reward functions
• Develops adaptive S-surface controllers that handle unpredictable underwater disturbances
• Improves AUV stability by addressing complex coupling between different degrees of freedom
• Demonstrates enhanced maneuvering capabilities in extreme sea conditions

This innovation represents a significant advancement for marine engineering, offering more reliable autonomous operations in unpredictable ocean environments. The approach could potentially extend to other autonomous systems facing similar control challenges.

Never too Prim to Swim: An LLM-Enhanced RL-based Adaptive S-Surface Controller for AUVs under Extreme Sea Conditions