Estimation of Vehicle Side-Slip Angle at Varying Road Friction Coefficients Using a Recurrent Artificial Neural Network

The side-slip angle is one of several crucial states in vehicle dynamics allowing to judge the current status of stability and the ride comfort for the subsequent usage in active assistance systems. Unfortunately, the measurement of side-slip angle is very costly and therefore it is usually not provided in production vehicles. As an alternative, the side-slip angle can be estimated based on the available sensors. The most common approach is model-based state estimation using different forms of Kalman filters (KF). However, this method comprises the complex steps of system identification and robust filter design. In recent years, data-based approaches gain popularity among researchers throughout different fields of research including the state estimation in vehicle dynamics. These allow direct extraction of an estimation algorithm from recorded data. This contribution presents the utilization of recurrent artificial neural networks (RANN) to side-slip angle estimation in a Volkswagen Golf GTE Plug-In Hybrid on varying road surfaces. The inputs to the RANN are signals from sensors in serial production configuration. In comparison to the model-based approach of a sensitivity-based unscented Kalman filter (sUKF) from previous works, the data-based approach shows competitive experimental results.