Feedback Controller Optimization for Mechatronic Systems with Unexpected Plant Perturbations Using Support Vector Machine

This study aimed to introduce a simple optimization method that can enhance the feedback control system of mechatronic systems. This method can cope with unexpected plant perturbations without needing any dynamic models or frequency response knowledge of the system. The method uses support vector machine (SVM) techniques to fine-tune feedback controller parameters according to the experimental classification results. The method separates "success" and "failure" classes using hyperplanes and optimizes the parameters to achieve the desired performance. It enables control engineers to attain the desired robust performance without relying on any dynamic models, even for unstable and multi-output systems. To demonstrate the effectiveness of this method, we apply it to an inverted pendulum robot and show that it can achieve the desired performance against individual variations. This method can be applied to various mechatronic systems that face uncertainties and disturbances in their operating environments and has the potential to be widely used in the field of mechatronics.