Rapid Assessment of Dynamic Friction Coefficient of Asphalt Pavement Using Reflectance Spectroscopy

Mapping road conditions is an important issue for city and state authorities worldwide. Today, pavement safety is assessed by specific assemblies based on a mechanical wheel device, which is a method that is limited in its potential product and operation. In this letter, we examined the possibility of harnessing remote reflectance spectroscopy to predict asphalt's dynamic friction coefficient, thereby enabling the identification and mapping of road conditions. We used a near-infrared analysis technique to evaluate an artificial neural network prediction model designed to assess the friction coefficient solely from spectral readings. This letter describes the method for extracting such a model and presents promising results with an accuracy of ${R} = 0.845$ and high significance of $P between actual and predicted friction values. This model was acquired using nine principal components and three neurons. The potential of this technology is also discussed.