Low-cost smartphone-based sensing approaches have recently emerged as an alternative to the traditional and oftentimes expensive or laborious road surface monitoring technologies. We develop an approach, resting on a rigorously derived mechanistic and stochastic model in spectral domain, that relates the vertical acceleration signal measured by a smartphone positioned in a moving car to road surface roughness metrics. We illustrate that the inferred roughness metrics from the acceleration data sets, acquired with different smartphones and vehicle types, are in good agreement with the ones obtained from the expensive laser measurements. The data set we used for our analysis is hence comprised of (I) acceleration signals recorded by smartphones placed at different locations of the vehicle, and (II) laser-measured road longitudinal profiles. The validation analysis is performed for two test tracks:  (1) 13.8 km of a thoroughfare with a dominating inner-city component with poor ride quality, and (2) 11.6 km of a highway with good ride quality. The coordinates (latitude and longitude) of the two test tracks are also available in the provided data set.

The procedure for processing is described in the manuscript. The code for identification of parameters is also provided.