Monitoring Road Surface Conditions for Bicycles – Using Mobile Device Sensor Data from Crowd Sourcing

Abstract

This paper introduces an approach for monitoring cycleway conditions by collecting crowdsourced data from mobile devices. To collect the data, an application was developed and optimized to be used by many cyclists. The application uses acceleration and gyroscopic sensors to collect and upload road roughness data into a classification platform. A classification model classifies the monitored routes into three quality classes and synchronizes the results with the application. The methodology shows how to collect and classify road surface conditions of cycleways. By using the K-Nearest Neighbor machine learning algorithm as a classifier, we were able to achieve a forecast accuracy above 90% on average. We report on our experiences with classification accuracy of four different classifiers as well as the experimental evaluations of the system. The results support the potential development of a community portal that provides detected cycleway conditions from the up-to-date mobile crowdsensing application.