ADRC Parameters Tuning using Traditional and AI-based Methods: A Survey

Active Disturbance Rejection Control (ADRC) has emerged as a prominent strategy for mitigating uncertainties and disturbances in complex systems. However, the performance of ADRC heavily depends on how well its parameters are tuned, which can be quite challenging in dynamic and nonlinear environments. This paper reviews both traditional and AI-based methods for tuning ADRC parameters, covering developments from 1990 to 2024. Traditional methods like trial-and-error, mathematical optimization, and frequency domain analysis are appreciated for their simplicity and stability guarantees. However, they often struggle with complex, nonlinear, or time-varying systems. On the flip side, AI-based methods—including machine learning, evolutionary algorithms, and neural networks—offer greater adaptability and efficiency, effectively addressing many limitations of traditional techniques. However, they also introduce new challenges related to data quality, computational complexity, and how interpretable they are. We highlight the potential of hybrid approaches that blend the strengths of both traditional and AI-based methods. By providing a thorough analysis of various tuning techniques, we aim to help design and implement more effective ADRC systems. Our findings are particularly relevant for advancing control engineering and automation in applications where high performance and robustness are crucial. We also point out future research directions, such as developing hybrid methodologies, creating interpretable AI models, and designing real-time adaptive tuning algorithms to meet the evolving demands of sophisticated control systems.