Optimal Iterative Learning Control for Mechatronic Systems

The relentless progress of technology promises improved performance, while requiring less resources, posing a persistent challenge for all fields of research. Particularly for model-based control design, this demand makes it necessary to take the mismatch between the considered system and an available model into account.

The goal of this dissertation is therefore to contribute to a reduction of discrepancy between theoretically expected and experimentally obtained results, where the applicability to and experimental validation on nonlinear mechatronic systems are of particular interest.

The problems caused by model-plant mismatch are approached by considering iterative learning control, which is an iterative procedure that updates the control signal to improve the performance of a system executing a repetitive task. The first part of the manuscript covers the extension of an existing theoretical framework to a generalised case, including a more versatile interpretation and novel results on stability and convergence. Following the theoretical analysis, an open-source software tool is introduced that implements the aforementioned theoretical framework and was developed to facilitate the modelling and control design. Using this software, two simulation cases are considered that study the characteristics and potential limitations of the approach when applied to nonlinear dynamics. Subsequently, an experimental validation is performed on an in-house developed educational platform, where - as an example of the versatility of the approach - the swinging up of a pendulum is iteratively learned. In the following, an adaptation of the theoretical framework is presented, which allows to simultaneously improve the performance and reduce the execution time of a task of a robotic manipulator. For this purpose, a number of modifications are made and an efficient solution strategy is implemented to handle the complex dynamics, while a final experimental validation shows the proclaimed properties and confirms the efficacy of the proposed approach.