A methodology for probabilistic pavement condition forecast based on Bayesian filters

Abstract

Decision making in pavement management relies on current road condition and the condition forecast. In this study, it is shown that both the condition forecast as well as the condition measurements are affected by uncertainties as demonstrated in a literature review and in preliminary studies of road condition data. If these uncertainties are to be considered in forecast models, the need for a probabilistic approach is evident. In this study a methodology based on an Extended Kalman filter (EKF) was developed and tested, which allows combining both empirical models and collected condition data for the development of section-based pavement forecast models. The model has been validated to predict the condition state effectively for all selected condition indicators. All relevant steps for the condition forecast have been implemented into a prototype to evaluate the applicability of the methodology using collected data on road networks from Germany, Austria, and Switzerland.