Hybrid Model Based Pre-Crash Severity Estimation for Automated Driving

In recent years emergency braking systems became a standard in modern vehicles. However, these systems can not prevent every collision. Integrated safety systems allow bringing vehicle safety to the next level. This paper introduces a crash severity estimation algorithm based only on information received from environmental sensors like radar, camera, and LiDAR. Using a quadruple Kelvin model, the physical behavior of the ego vehicle during the crash is approximated, and thus, the crash severity parameters are derived. This paper focuses on the headon collisions with different relative velocities and approach angles. More than 50 finite element method simulations (FEM) with the same crash scenarios were performed to compare and validate the model’s results. The results prove that the presented methodology can reproduce the crash behavior and reliably approximates the crash severity parameters with-in the desired range.