Model-Based Soil Trip Rollover Prediction Using Driving Dynamics

Abstract

This paper presents a new physical model-based approach for improved rollover detection. Using vehicle dynamics information allows the prediction of the future roll motion of a vehicle. The severity of a crash can be forecasted in a very early phase of the accident. Thus, a much faster rollover decision can be made, which is necessary in order to deploy nonreversible restraints, like pyrotechnical belt and curtain airbags in soil trip rollover scenarios in time. The performance of this new approach is validated by a comprehensive database recorded with a scaled rollover test vehicle. Furthermore, the performance and robustness is compared to state-of-the-art model-based rollover detection methods. It can be shown that in most soil trip rollover crashes the roll angle at airbag deployment times can be reduced by up to 52 % compared to standard rollover crash detection methods. In addition, the usage of a scaled test vehicle for the development and validation of soil trip rollover detection methods is discussed. On the one hand, the similarities between a scaled model and its original are described with the help of the dimensional analysis. On the other hand, recorded data from the scaled rollover test vehicle is compared to rollover data of a real vehicle.