Interface and Optimizations for Crash Severity Estimation and Inevitability Modelling in Pre-Crash Safety Systems

Abstract

"In recent years, emergency braking systems were introduced to detect and prevent potentialaccidents. However, it is not always possible to avoid a crash. Hence, active safety sensors and passive safety systems are merged into an integrated safety system to reach a maximum safety level. For this, exteroceptive sensors such as radar, LiDAR, and camera monitor the vehicle’s surrounding and create a virtual map. Critical vehicle constellations are predicted and used to activate passive safety actuators milliseconds before the inevitable crash occurs. However, to securely activate these systems, it is necessary to predict the vehicle movement in critical situations that lead to an accident. Prediction methods must evaluate and interpret the exteroceptive sensors’ information to determine the inevitability of a crash and its upcoming crash severity. This paper presents an interface which links a method for the predictive determination of inevitable crash constellations with a crash severity estimation algorithm. In a conventional approach, the crash inevitability provides the trajectories, and the crash severity is calculated separately for the different possible crash constellations. Thus, the expected crash severity can be interpolated, triggering suitable vehicle restraint systems. A novel approach includes the sensor and system tolerances and their effect on crash severity estimation. The physically feasible ego and bullet vehicle trajectories are calculated, and all combinations are investigated for a possible collision. An adapted vehicle dynamics model was created to suit the needed accuracy and the time-critical condition of a pre-crash situation. This model calculates the time until collision as well as the parameters describing the collision. The parameters like relative speed between both cars ∆v, the impact angle α, and the expected impact location allow the crash severity estimation to compute the associated severity of each trajectory combination. The inherent variances of these parameters due to sensor and data tolerances are also considered. A quadruple mass-spring-dampermodel physically approximates the involved vehicles’ in-crash behavior and provides the resulting crash severity values, e.g., ASI. The resulting array of crash severity will be analyzed to identify trajectory combinations of similar severity and corresponding crash constellations. The influence of the tolerances on the described system can be assumed, and future simplifications of the crash calculations can be achieved. This paper introduces a methodology for a fast and robust combination of inevitability detectionand crash severity estimation for integrated safety systems, including the effects of sensor and datatolerances. The first results show that the model can reproduce various complex frontal crash constellations accurately. More research will be done to implement the algorithm in a prototype vehicle, which can be tested for validation in the research and test center CARISSMA at Technische Hochschule Ingolstadt."