LLM-based PID controller optimization

This study explores the application of Large Language Models (LLMs) to optimize Proportional-Integral-Derivative (PID) controller parameters, aiming to automate a process traditionally reliant on manual, iterative tuning. Conventional methods often depend on heuristics and trial-and-error, requiring significant time and expertise. To address this, an LLM-based approach was developed that iteratively adjusts controller gains using real-time system feedback, substantially reducing the need for manual input. An adaptive tuning strategy, featuring predefined modes and levels of aggressiveness, was designed to balance response speed, minimize overshoot, and ensure system stability. The method was validated through computer simulations across a range of carefully selected process models and benchmarked against state-of-the-art expert-driven tuning techniques, demonstrating its effectiveness in reducing tuning time according to the selected mode while maintaining or improving system robustness. Results showed that LLMs, without additional model fine-tuning and guided solely by prompt engineering, can effectively optimize PID parameters. These findings demonstrate the promise of LLM-driven methods for advancing automation in PID controller tuning.