Predicting Pavement Marking Retroreflectivity Using Artificial Neural Networks: Exploratory Analysis

Abstract

Providing adequate nighttime visibility to roadway users is an important consideration for state and local transportation agencies. Driving at night is less dangerous when pavement markings are easily discernable. Retroreflectivity is a measure of nighttime visibility. Transportation agencies could use estimates of the expected service life of pavement markings to plan restriping operations at a time when markings are near a minimum threshold level of retroreflectivity. The present study proposes the use of an artificial neural network to predict pavement marking retroreflectivity as a function of initial retroreflectivity, the age of the markings, traffic flow, pavement marking type, and route location information using data from North Carolina. The results show that many of the input variables have a nonlinear association with pavement marking retroreflectivity. Surface plots of the degradation pattern are provided to illustrate the relationship between input and output variables. Estimates of service life are provided to show how the output can be used to manage pavement marking systems.