A Velocity Algorithm for Nonlinear Model Predictive Control

Abstract

This brief presents a velocity-form nonlinear model predictive control (NMPC) scheme via velocity-based linearization. The main features of this approach are built-in offset-free control in the presence of disturbances, tracking of piecewise constant, possibly unreachable, reference signals, and simple implementation, as a parameterization of all equilibria is not necessary. Furthermore, the model in velocity form can be expressed as a quasi-linear parameter-varying (quasi-LPV) model, for which efficient online optimization algorithms exist. The proposed approach is experimentally validated on a 2-degree-of-freedom robotic manipulator, where its capabilities and efficiency are highlighted.