A computational model of pedestrian road safety: The long way round is the safe way home.

We propose a novel linear model of pedestrian safety in urban areas with respect to road traffic crashes that considers a single independent variable of pedestrian path safety. This variable is estimated for a given urban area by sampling pedestrian paths from the population of such paths in that area and in turn estimating the mean safety of these paths. We argue that this independent variable directly models the factors contributing to pedestrian safety. This contrasts previous approaches, which, by considering multiple independent variables describing the environment, traffic and pedestrians themselves, indirectly model these factors. Using data about 15 UK cities, we demonstrate that the proposed model accurately estimates numbers of pedestrian casualties.