Real time nonlinear model predictive control for fast systems

A practical nonlinear model predictive control (MPC) approach is presented for fast response systems. To this end, a nonlinear model of the system is first developed and linearised along a desired trajectory at each instant to obtain as many linear models as prediction horizon. The state variables are observed using an unscented Kalman filter (UKF). The MPC formulation is thus resulted in a linear quadratic optimisation problem. The proposed control methodology is implemented on a fast multi-input-multi-output nonlinear aerodynamic test-rig. The results show that the approach is feasible in real time situations.