Modelling pedestrian crossing behaviour using Bayesian networks

Abstract

The causes of pedestrian accidents are a major concern to transportation engineers and other road safety professionals. Although studies have aimed at modelling and analysing the causes, they are either too stochastically oriented or excessively macroscopic, and so fail to analyse the interactions between pedestrians and their immediate environment. This study applies a Bayesian network modelling approach to investigate factors influencing pedestrian crossing behaviour in spring and summer using variables known from the literature. The model analyses the problem on three levels: graphical level, information level and quantitative level. The results show that pedestrians often exhibit rational crossing behaviour influenced predominantly by personal motives rather than external factors, even though the roadway environment is not always favourable. A sensitivity analysis revealed that signal timing phase length is the most influential parameter that affects pedestrian crossing behaviour.