SC42125 Model predictive control

Model Predictive Control (MPC) is perhaps the most effective optimal control strategy for constrained dynamical systems. The basic concept of MPC is to exploit a dynamic model to forecast the system behavior, and optimize the forecast to determine the control input as the best decision at the current time. With emphasis on constrained linear systems, the course will present the theoretical fundamentals of MPC, such as Lyapunov stability, optimality and robustness, and its computational methods, such as quadratic programming.

Study Goals

  • Derive state prediction matrices from discrete-time linear models
  • Design the cost function, state and input constraints
  • Design MPC controllers with guaranteed recursive feasibility and asymptotic stability via appropriate terminal cost and constraint set
  • Formulate and solve constrained-linear-quadratic MPC problems via quadratic programming
  • Implement and simulate closed-loop systems controlled by MPC on Matlab
  • Design MPC controllers with integral action for reference tracking


S. Grammatico

Last modified: 2023-11-03


Credits: 4 EC
Period: 0/0/4/0