Data-Driven Controller Design Based on Response Estimation for Multi-Performance Optimization

This study presents a novel data-driven controller design method based on response estimation. We estimate the response during the controller adjustment process using only the initial input/output data from the closed-loop experiment. Then, we tune the controller parameters by optimizing an objective function based on the estimated response data. Our proposed tuning method simultaneously improves multiple performances: tracking performance, response speed, and signal smoothness. The proposed method can predict the input/output response of the closed-loop system before applying the tuned controller to the actual system, thus avoiding damage to the machine and reducing the cost of repeated experiments. Furthermore, the total variation denoising method is introduced to handle the initial input/output data that contains noise. Finally, the effectiveness of the proposed method is verified by numerical examples.