Robust model predictive control using the unscented transformation

This paper presents a new approach for robust open-loop and closed-loop control of nonlinear processes with parameter uncertainties and a comparison with classical concepts. The approach leads to trajectories that show small variations if uncertain parameters and uncertain initial conditions are present. The algorithm utilizes the Unscented Transformation. It allows a 2nd order approximation of the first two statistical moments of the system´s output as a function of the stochastic system´s state and uncertain model parameters. Because the numerical burden is low, it can be used for optimization based online closed-loop process control as well