Optimization of the Braking Strategy for an Emergency Braking System by the Application of Machine Learning

This paper explores a methodology whereby accident data is directly used to develop a braking strategy for an autonomous emergency braking system. Future vehicles will be equipped with additional technologies. Detailed information about accidents or critical situations can be recorded. If enough recorded data from critical situations are available, such data could be used to improve Active Safety Systems. In our approach, we do not model the behavior of pedestrians or drivers. The idea is to use the capability of machine learning to get the behaviors out of traffic data. Machine learning is used to derive the function design for an emergency braking system for pedestrians. Generated traffic scenarios are used to review the methodology. Random Forests and Neural Networks are used for the function designs and the learned function designs are compared with a reference implementation.