Resumen
This paper describes the main findings of a study performed by INSIA-UPM about the improvement of the reconstruction process of real world vehicle-pedestrian accidents using PC-Crash® software, aimed to develop a software tool for the estimation of the variability of the collision speed due to the lack of real values of some parameters required during the reconstruction task. The methodology has been based on a sensibility analysis of the factors variation. A total of 9 factors have been analyzed with the objective of identifying which ones were significant. Four of them (pedestrian height, collision angle, hood height and pedestrian-road friction coefficient) were significant and were included in a full factorial experiment with the collision speed as an additional factor in order to obtain a regression model with up to third level interactions. Two different factorial experiments with the same structure have been performed because of pedestrian gender differences. The tool has been created as a collision speed predictor based on the regression models obtained, using the 4 significant factors and the projection distance measured or estimated in the accident site. The tool has been used on the analysis of real-world reconstructed accidents occurred in the city of Madrid (Spain). The results have been adequate in most cases with less than 10% of deviation between the predicted speed and the one estimated in the reconstructions.