Numerical Methods for Nonlinear Model Predictive Control in Real-Time

Many control applications need to deal with nonlinear, unstable and constrained processes. Nonlinear Model Predictive Control (NMPC) can explicitly handle nonlinear dynamics while enforcing state and input constraints by solving on line an optimal control problem (OCP). In the past, the computational burden of solving an OCP has limited the application of NMPC to slow processes. Recent algorithmic advances allow to reduce the computational times to a few milliseconds, making it possible to apply NMPC also to fast dynamic systems. In this seminar, we will first introduce the NMPC algorithm and sketch the stability proof. Second, we will present the algorithms which allow for ultra-fast NMPC. Finally we will conclude by presenting some challenging automotive and aerospace applications