Preisträger Best Contribution MIT 2014
Methods for predicting crash severity prior to vehicle head-on collisions
Upcoming safety systems like accident-adaptive restraint systems may help to improve vehicle safety. One major challenge for the realization of these systems is that they may require a fast and interpretable function which predicts the severity of an accident prior to collision. Therefore, only with accident parameters estimated by precrash car sensors the severity of the upcoming collision has to be predicted. In this work, we give an overview of data-driven methods to find classification and regression models for this problem automatically. For that, we preprocess crash simulation data and train different models. We also evaluate their performance and discuss the results. Finally, we finish with a conclusion and research questions, which may lead to an application of these models for future, safer vehicles.