Model Predictive Control (MPC), also known as moving horizon control or receding horizon
control, refers to a class of algorithms which make explicit use of a model to optimize the future predicted behavior of a system. During the past 40 years, MPC has proved enormously successful in industry mainly due to the ease with which input and state constraints can be included in the controller formulation. Originally developed to cope with the control needs of power plants and petroleum refineries, it is currently successfully used in a wide range of applications, ranging from the process industry to the automotive and the biomedical sectors. This tremendous success is in part motivated by the fact that MPC can be also applied to hybrid systems and to fast processes. For these latter, explicit MPC for example allows performing most of the computation off-line thus reducing the control law to a simple look-up table. On the other side, online computation can benefit of the use of hardware like FPGA and GPU.

The aim of the course is to provide students with a general overview about this emerging control technique. The objectives of this course are:

• gaining understanding of optimization
• understanding the basic concepts of MPC
• implementing programs to solve simple optimization and control problems

Basic notions of Control Theory and a good knowledge of Matlab programming are expected.

The course will start with basics on optimization. We will focus on convex optimization problems and in particular on linear and quadratic problems. Basics of mixed integer optimization will also be provided. Then we will introduce Model Predictive Control, an optimization based control technique that allows minimizing a given performance index while dealing with state and input constraints. We will first focus on linear systems and then show how this technique can also be applied to hybrid systems. Finally we will consider Explicit MPC as a possible way to alleviate online complexity. Exercise sessions throughout the course will help the students to gain confidence with the implementation of MPC in Matlab.

Equipment

Since the lessons will be held in the library on the 3rd floor of Treitlstrasse, in order to follow the lab experience it is suggested (but not required) to bring your own laptop (or to share a laptop with a collegue) equipped with the following software:

• MATLAB MathWorks
• The Multiparametric Toolbox for Matlab
• IBM CPLEX

This is a visiting professor course of Vienna PhD School of informatics for Computer Engineering. Lecturer: Dr. Davide Raimondo (University of Pavia, Italy)

Registration takes place in TISS

http://www.nobhillpublishing.com/mpc/electronic-book.pdf