The quality of road pavements influences noise and vibration emission caused by tire-road interactions. This affects the drivers, passengers and load but also health and well-being of residents near these roads. Road pavement quality degrades over time due to wear, accidents, and infrastructure works. Monitoring road pavement state can rely on dedicated vehicles equipped with a CPX-trailer (Close-Proximity method) or with laser texture scanning. However, using this approach, it remains difficult to cover the whole road infrastructure network at regular intervals. In this paper, an opportunistic approach is proposed: equipping cars that are on the road for other purposes with noise and vibration sensors. This way, personnel costs are avoided, and timeliness of the information could be increased. The proposed method collects spectral sound and vibration data from a sensor box placed near the rear wheel of the car. These data are transmitted over 3G to a central server. The box is also equipped with a GPS tracker that allows locating the vehicles on a road map and deriving their driving speed. Data analytics accounts for modifiers such as the transfer function between the tire and the microphone and driving speed. Features related to the roughness of the pavement are extracted and the abundance of data is used to eliminate confounders such as engine noise and vibrations, other cars and trucks driving near the sensor box, music and voices. The resulting roughness indicators for each 20-meter road segment correlate very well to CPX and laser roughness measurements (r2 = 0.6). The difference between worn-out roads (> 15 years) and new pavements (
