Output-feedback data-driven model predictive control of nonlinear dynamical systems using a nonlocal model

In this paper, a new output-feedback data-driven model predictive controller (MPC) is proposed for single-input single-output nonlinear dynamical systems. For this purpose, a new method for the mathematical modelling of such systems using input/output data collected from the system was developed. The proposed model is in the form of a set of linear algebraic relationships between the input and output samples, which is added as an equality constraint to the convex quadratic optimization problem used to calculate the control. The introduced model is nonlocal, meaning that it only needs to be created once at the start of the control system, and there is no need to update it using the most recent data collected from the closed-loop control system while it is running. The proposed data-driven MPC is used to control a nonlinear dynamical system, and it is shown that the closed-loop control system can track different reference inputs even at the presence of measurement noise.