Towards a probabilistic alert procedure of roadway departure

In this paper, roadway departure risk while cornering is assessed using probabilistic methods. Sensitivity analysis conducted through Sobol' indices lead to the selection of the parameters that are modeled by random variables. Then structural reliability methods are applied to compute a risk indicator. Importance measure of several parameters are studied to define the quantities that must be statistically determined for an accurate risk assessment at the entrance of the curve. This study is a part of the french national road safety project SARI, and involves several partners for the implementation of a roadway departure warning system.