Nonlinear system identification of a furuta pendulum using machine learning techniques

Abstract

Usually, dynamical systems can be described by differential equations. An accurate model is essential when designing and optimizing a controller. However, not every system can be modeled easily by physical models due to highly nonlinear behavior, such as friction or backlash. Then, a data based approach, such as machine learning, might be helpful. The focus in this work is set on modeling dynamical systems using neural networks and deep learning, which are growing subjects in research and industry to identify nonlinear dynamics.