CrowdAbout: Using Vehicles as Sensors to Improve Map Data for ITS

Crowdsourcing can be seen as an opportunity to provide important information for Intelligent Transportation Systems to improve the service quality of various applications in this domain. Autonomous or assisted vehicles need the most accurate map data possible to adjust the respective assistants to it. In this work, we present CrowdAbout, a system that uses the crowd as mobile sensors to collect data from smartphone sensors during trips. The system recognizes special traffic events like roundabouts with the help of machine learning. These findings are used to automatically correct OpenStreetMap data and adapt them to a changing road network. An evaluation of different machine learning algorithms using self-collected realworld data of over 200 roundabouts shows that the recognition of roundabouts including exit and radius is possible with high accuracy.