Risk Analysis in Road Safety: Probability and Vulnerability Models for Drivers and Pedestrians



Risk Analysis in Road Safety: Probability and Vulnerability Models for Drivers and Pedestrians

Authors

F A Marcianò, A Vitetta, University Mediterranea of Reggio Calabria, IT

Description

In this work a model for risk analysis of road accidents is presented.
The model permits to calculate the value of individual risk of drivers and pedestrians moving in given scenarios that can be adapted according with the analysis? purpose.

Abstract

In this work a model addressed to risk analysis of road accidents is presented.
The model permits to calculate the value of individual risk of drivers and pedestrians moving in given scenarios that can be adapted according with the analysis? purpose.
The models are specified regarding a nested structure (tree of the alternatives) representing the events that characterize the incident.
The first level of the tree contains the probability analysis that allows to calculate, through two models, the probability of the type of incident (incident of pedestrian or incident of vehicles) and the probability, conditioned by the happening of a pedestrian incident, that the pedestrian involved has age and sex assigned. The first probability is estimated with a disaggregate model of Logit type meant like a descriptive model of the phenomenon; the second probability is estimated with an aggregate model of regressive type regarding the age and the sex of the pedestrian.
The second level of the tree contains the vulnerability analysis that allows to calculate the probabilities, conditioned by the type of incident, that the pedestrian suffers a damage (hurt or died) or that the driver suffers a damage (uninjured, hurt or died). The probabilities are calculated through a system of disaggregate models of Logit type meant like a descriptive model of the phenomenon.
All the models have been calibrate regarding the data contained in the database of the incidents recorded in the 188 commons of the province of Milan for the biennium 2001/2002 realized by the National Institute of Statistics in collaboration with the Province of Milan; such database is constituted by a series of records everyone of which regards a single incident and of which are contained the more important informations (system driver, pedestrian, vehicle, road), for a total of 62359 records.
The method used for the calibration of the models is that of the Maximum Likelihood that provides the values of the unknown parameters that maximize the probability to observe the choices made by the customers.
At last on the risk model, calibrated on the data of the incidents of the province of Milan, has been executed a test on the common of Reggio Calabria with the aim to verify the transferability and the reliability of the results provided in the moment in which the model is applied on a different territorial truth from that one regarding which it has been calibrated; the results obtained, infact, were compared with the data contained in the database of the incidents happened in year 2005 in the common of Reggio Calabria provided by the Municipal Police of Reggio Calabria. The database is constituted from a series of records, in total 1146, everyone of which is referred to a single incident, of which are contained the informations relative to vehicle, driver, pedestrian and environmental conditions.

Publisher

Association for European Transport