Approximating open-loop and closed-loop optimal control by model predictive control

Abstract

We consider a finite-horizon continuous-time optimal control problem with nonlinear dynamics, an integral cost, control constraints and a time-varying parameter which represents perturbations or uncertainty. After time-discretization of the problem we employ a Model Predictive Control (MPC) algorithm with a shrinking horizon, which uses a prediction/forecast for the uncertain parameter and (possibly inexact) measurements of the state vector, and generates a piecewise constant control signal by solving auxiliary open-loop control problems. In our main result we derive an estimate of the difference between the MPC-generated control and the optimal feedback control, both obtained for the same value of the perturbation parameter, in terms of the step-size of the discretization and the measurement error. We also give an estimate for a norm of the difference between the MPC-generated control and the optimal open-loop control in the problem with the true value of the uncertain parameter, depending on the prediction error.