Model Predictive Control Course
Teacher: prof. Alberto Bemporad
Period: November – December 2013
Location: IMT Institute for Advanced Studies Lucca, San Francesco (main building), Lucca, Italy
The course was intended for students and engineers interested in learning the theory and practice of Model Predictive Control (MPC) and optimization of constrained linear, stochastic, and hybrid dynamical systems. MPC has a well-established tradition in the process industries since a few decades for controlling large multivariable systems subject to various constraints on manipulated variables and controlled outputs. In recent years, MPC is also rapidly expanding in several other domains, such as in the automotive and aerospace industries, in smart energy grids, and in financial engineering. The theoretical foundations of MPC concern stability and robustness guarantees, and how to deal with hybrid and switched dynamics, fast-sampling processes, and stochastic uncertainty. The course made use of the MPC Toolbox for MATLAB developed by the speaker and co-workers (distributed by The MathWorks, Inc.) for basic linear MPC, and on the Hybrid Toolbox for explicit and hybrid MPC.
General concepts of Model Predictive Control (MPC). MPC based on quadratic programming. General stability properties. MPC based on linear programming. Models of hybrid systems: discrete hybrid automata, mixed logical dynamical systems, piecewise affine systems. MPC for hybrid systems based on on-line mixed-integer optimization. Multiparametric programming and explicit linear MPC, explicit solutions of hybrid MPC. Stochastic MPC: basic concepts, approaches based on scenario enumeration. Linear time-varying MPC and applications to nonlinear dynamics and obstacle avoidance problems. Selected applications of MPC in various domains, with practical demonstration of the MATLAB toolboxes.
Linear algebra and matrix computation, linear control systems, numerical optimization.
Full description available at: http://cse.lab.imtlucca.it/~bemporad/mpc_course.html