Sigmoid-Based Method for Longitudinal Crash Pulse Prediction in Intelligent Vehicles

Autonomous vehicles require simple, real-time capable, robust, and accurate methods for predicting crash severity to ensure optimum occupant safety. Current restraint systems which are designed for specific scenarios and seating positions, need an upgrade (deploy in pre-crash) to cater to the challenges from new occupant postures such as sleeping, facing each other, etc. The deployment strategy during a collision depends mainly on the crash pulse and the occupant positions. This research paper investigates an inverse sigmoid approximation method for crash pulse prediction. The results highlight the potential for use in crash severity prediction with a mean absolute error of 0.17m/s and 3 g for the velocity and acceleration pulses, respectively. Moreover, it requires only three parameters for the prediction, illustrating the feasibility of real-time application in intelligent vehicles.