Continuing efforts in the field of traffic accident research has led to the development of various active and passive safety systems. They act and influence an incident/accident at different points in time in order to mitigate or avoid a collision. In the event of a collision, the decision to deploy passive safety systems must be made quickly, as the typi- cal activation time is between 10 to 40 ms after the initial contact. However, for future interior/seating concepts and earlier deployment times of restraint systems it is necessary to predict an unavoidable collision much earlier. As the knowledge about this so-called Point Of No Return (PONR) is crucial, this paper introduces an approach to calculate it through using numerical simulation.

This paper uses real world accidents out of the GIDAS (German In-Depth Accident Study) project. The database contains more than 33.000 accidents with personal damage. The reconstruction of the several phases (normal driv- ing, critical situation, pre-crash, in-crash, post-crash) is the basis for the estimation. Therefore, an imminent collision is predicted by simulating the vehicle’s possible behavior using a multi body system. If any physically possible vehi- cle reaction exists that leads to an avoidance of the collision, the PONR has not yet been reached. If all simulation solutions would lead to a collision, the calculations must go one time step backwards. Through an efficient iterative approximation procedure the PONR can be found in a reasonable number of iterations. The approach focuses on the maxima in longitudinal direction (full acceleration or deceleration), in lateral direction (full steering right or left) and in all four combinations of steering and accelerating/decelerating.

The approach can be generally used for all collision types. Here, it is applied to rear-end collisions between two ve- hicles, highlighting the potential of different avoidance strategies like full deceleration or full steering as a function of time. The distribution of time across all PONRs shows that passive safety measures can be activated prior to the collision in the vast majority of cases. Therefore, occupant protection can be further improved and accidents conse- quences could be mitigated to a higher degree.

The suggested approach can estimate the PONR for real accidents. An adaption to naturalistic driving data as well as real time estimation is conceivable. This would signify a crucial contribution to the current research on the distinc- tion between accidents and incidents. However, some adaptations would be necessary to enable such calculations. The current simulations are based on idealized acceleration/deceleration and steering behavior, while traffic flow is neglected. Both simplifications lead to an underestimation of the PONR. As the approach is modularized, it can be further developed towards other vehicle behavior maneuvers, specific driver models, or interactions with Advanced Driver Assistance Systems (ADAS).

The PONR is an important value for improved vehicle safety. The developed approach allows to estimate the further potential of passive safety systems with regard to earlier activation times. Furthermore, it can be used to evaluate collision avoidance strategies and to parameterize ADAS.