Convolution-Based Data-Driven Simulation and Controller Design Method

Some of the challenges in data-driven control are the selection of the nominal model of the closed-loop system that gives the fastest response achievable and the validation of the tuning results. The virtual time-response-based iterative gain evaluation and redesign (V-Tiger) method is an approach to solve the problems in data-driven control, but it cannot account for the nonlinearity of the controller. The article proposes a data-driven simulation and controller design approach named the convolution-based data-driven simulation (CDDS) method. Based on time-domain convolution operations, the method enables closed-loop simulations without building a plant model. The method offers various approaches for controller design, such as directly specifying the overshoot and settling time and achieving the desired characteristics. Unlike the conventional methods, the CDDS method can explicitly handle the nonlinearity of the controller and is expected to be applicable to a wide range of control systems. The results of experiments conducted using a buck converter indicate that the CDDS method can reduce the estimation error by up to 95.0% compared with the conventional V-Tiger method. Furthermore, it can reduce the tuning error by more than 52.0% compared with the virtual reference feedback tuning and noniterative correlation-based tuning methods.