Reliability analysis of roadway departure risk using stochastic processes

The work presented here aims to develop a warning device to prevent roadway departure while cornering. Given the random variability arising from the driver, the vehicle and the infrastructure at the entrance of the curve, a probabilistic strategy is adopted to assess the roadway departure risk. Random variables and processes are introduced in a specifically developed vehicle dynamics model. The driver's behaviours are deduced from real traffic measurements. Structural reliability methods are employed to compute a roadway departure risk index, used to take the decision of an alarm triggering. Particular care is brought to the reduction of the computational cost. Refinements made on the standard reliability methods to handle with the model non-linearities and the stochastic dimension are presented.