Calibration and validation of a simulation model for predicting pedestrian fatalities at unsignalized crosswalks by means of statistical traffic data

This work presents a simulation model for unsignalized crosswalks which takes into account collisions between vehicles and pedestrians, thus allowing to assess the estimated yearly pedestrian fatality. In particular, we focus on a method to calibrate such a model combining measurable crosswalk characteristics, such as maximum speed limit or drivers' compliance, with statistical data for past accidents obtained from local municipality. In order to perform simulations under realistic conditions, we constructed a one-week scenario where pedestrian and vehicle traffic vary using specific patterns each hour of the week. The constructed traffic profile is based on openly available data and the suitability for the scenario considered (a crosswalk in Milan, Italy) is investigated showing that cultural/lifestyle elements determine the variation of weekly traffic. Simulations using the constructed one-week scenario were used to obtain the only non-measurable parameter which account for pedestrians' and drivers' distraction. In addition, we also focused on the presence of elderly pedestrians which have different physiological characteristics compared to adults or children and are becoming an important part of the population in several countries around the globe. The simulation model presented here and the method suggested for calibration may be employed in different contexts, thus allowing to build an important tool to be used not only for transportation efficiency/optimization but also for safety analysis.