Data-driven Linear Parameter Varying Controller Synthesis using Iterative Feedback Tuning: An I/O approach

In this thesis the usage of the LPV-ARX representation of a LPV system for extending the applicability of the Iterative Feedback Tuning (IFT) framework from LTI control to LPV control was investigated. Two methods for extending the IFT framework were researched, both based on the LPV-ARX representation, namely; the extended regressor and the lifted LPV-ARX model. The main challenge for this extension is the scheduling sequence as it must be the same in both the reference and gradient experiments. To address this problem two compensation methods were devised, which allowed the algorithm to compensate for changes in the scheduling sequence. These compensation methods led to two distinct disadvantages namely; the need for a model of the plant (extended regressor) and the curse of dimensionality (lifted LPV-ARX). Although these disadvantages exist, both algorithms were shown to be capable of optimizing LPV controllers for SISO and MIMO LPV systems.Delft Center for Systems and ControlMechanical, Maritime and Materials Engineerin