A data-driven model predictive control approach toward feedback linearization of nonlinear mechanical systems

This paper presents a novel approach to linearize the input-output (IO) response of nonlinear mechanical systems by using model predictive control (MPC) with integral action and solving a one-step-ahead reference tracking problem. The discrete-time MPC controller, which builds on a nonlinear data-driven state-space model, controls a continuous-time plant. State estimation is performed by means of the unscented Kalman filter (UKF). The overall effectiveness of this MPC-based approach is validated in the time and frequency domains by conducting simulations on a mechanical system with output nonlinearities of polynomial type. The obtained results show that the proposed method is superior in terms of performance and robustness when compared to the classical feedback linearization techniques based on Lie algebra.