The development of methods for detection and assessment of safety critical events in car driving

Abstract: Improving our knowledge of drivers’ behaviour, especially in hazardous situations is a key to understanding why accidents occur and how to improve safety on our roads. Crash surrogate measures have proven to be very useful in traffic safety analysis. As a valid crash surrogate measure shares the same logical chain of events as actual crashes, studying those events increases the opportunities to analyze external circumstances as well as events and driver behaviour preceding the crash. A new definition of crash surrogate measures is proposed, i.e. safety critical braking events: Situations (including crashes) that require a sudden, evasive manoeuvre to avoid a crash or to correct for unsafe acts performed by the driver himself/herself or by other road users. This thesis develops a method for detecting Safety CRItical Braking Events (SCRIBE) and evaluates against other methods mostly used in large naturalistic driving studies. The evaluation shows promising results in the success rates of detecting safety critical events. Further, a Method for estimating the SEverity of safety Critical events (M-SEC) involving more than one road user, is developed and evaluated. The method combines a measure of closeness-to-collision, or safety margins, with a measure of possible consequences based on the speed and mass of the involved road users. Besides, a comparison is made with the Traffic Conflict Technique. Evaluation of the method shows that the estimated severity using M-SEC enables comparison of safety critical events involving different types of road users and is not limited to comparisons between similar types of events. In addition, the estimations when using M-SEC seem to reflect the seriousness of the safety critical events.