Nonlinear model predictive control from data: a set membership approach

SUMMARY A new approach to design a Nonlinear Model Predictive Control law that employs an approximate model, derived directly from data, is introduced. The main advantage of using such models lies in the possibility to obtain a finite computable bound on the worst‐case model error. Such a bound can be exploited to analyze the robust convergence of the system trajectories to a neighborhood of the origin. The effectiveness of the proposed approach, named Set Membership Predictive Control, is shown in a vehicle lateral stability control problem, through numerical simulations of harsh maneuvers.