This work addresses and evaluates the likelihood of human casualty in highway crashes, projected on the basis of field crash data that may become available electronically by sensors at crash time, and/or observed at the crash scene by emergency attendants. Termed collectively as a "crash signature", such data are treated as predictors and are selected from: crash severity, general area of damage, direction of force, occurrence of rollover, intrusion, vehicle crush and its specific horizontal location, collision partner, vehicle class and size, occupant age, gender, restraint use and type, seating position, and other. Crash signatures are converted into responses such as: (a) the likelihood of the most severe outcome, fatality or survived injury, by severity AIS per occupant; and (b) the same per vehicle. Cars are the vehicles selected for this investigation. A likelihood is quantified by a probability of occurrence, as a function of a string of predictors selected for maximum resolution and sensitivity, and minimum contribution to error. (A)
Abstract