Repetitive Control based on a Model Inversion: Application to a Lagrangian System

The aim of this thesis is to design and test a novel control scheme able to provide a very accurate tracking of a periodic reference trajectory. The main idea is to combine the Repetitive Control framework with a filter that reconstructs the control action that the Repetitive Control has to learn. Such filter, performing a plant model inversion, relies on the knowledge of the control action, the output measurements and the plant structure. This approach facilitates the stabilization of the resulting closed loop system that, for certain classes of systems, would be hard if not impossible. The performance of the proposed controller has been verified through simulation tests considering a Lagrangian system affected by a periodic disturbance. More than one configuration of the control scheme has been investigated and their performance compared. Very satisfying results have been obtained concerning the steady state tracking error. Moreover, the controller presents a certain robustness with respect to plant parameter uncertainties and to possible mismatches between the filter and the actual inverse dynamics.