One-Shot Data-Driven Design for Feedback Controller and Reference Model With BIBO Stability

This paper presents a one-shot data-driven design method for obtaining parameters of both a feedback controller and a reference model, which achieve a fast response while satisfying the user-defined model-matching error, using only single-experiment data without requiring knowledge of a plant model. Conventional data-driven tuning methods face two challenges, as reference model tuning has not been established and closed-loop stability cannot be guaranteed. The proposed method employs a data-driven predictive approach that can predict the closed-loop output. By leveraging this predictive framework, a new optimization problem is formulated to obtain the optimum parameters of both the controller and the reference model. The effectiveness of the proposed method is verified through numerical simulations for a plant with a dead time and experimental validation on a real application. The results demonstrate that the proposed method optimizes the feedback controller and the reference model parameters with bound-input bound-output (BIBO) stability from single-experiment data, which realizes a fast response while achieving the user-defined model-matching error.