An estimation of road surface conditions using participatory sensing

Abstract

When natural disasters occur, some roads could be blocked and cannot be used. Road surface conditions also deteriorate. Thus, collecting and providing the information on usable roads and road surface conditions can allow people to be evacuated safely. In this study, we proposed an estimation system of the road surface conditions by collecting accelerometer data from pedestrians' smartphones. The method estimates whether the road surface condition is a flat pavement road, a rough road, a slope or a stair by using supervised machine learning method. From the results of experiment, we found that the system can estimate six types of road surface conditions with a high accuracy when training the model with the data from the users.