Dp 583 en
A framework for nonlinear model predictive control
Author: Mikuláš Ondřej
Main objective of this thesis is development and implementation of a modular computer framework for nonlinear model predictive control (NMPC). Modularity of the framework is achieved by dividing NMPC algorithm into several logical blocks that can be implemented independently. The NMPC algorithm is described from the theoretical point of view. Commonly used approaches and individual computation steps are discussed. The control problem formulation is studied and practical guidelines are given to demonstrate how the control objectives can be transformed into the optimization problem cost function and constraints. The NMPC framework is implemented in Matlab and can be used as a tool for NMPC control development and prototyping. As the framework is modular, different optimization solvers, numerical integration routines and NMPC approaches can be evaluated for any particular application. Functionality of the developed NMPC framework is demonstrated on illustrative control problem examples. The first example is a simplified vehicle steering control and the second one is a diesel engine air-path control.