Poster: Urban Traffic Monitoring via Machine Learning

Traffic monitoring systems are nowadays operated with traditionally wired systems which need expensive infrastructure. In this work, we propose a traffic monitoring method for urban environments. The proposed method is passive and exploits RF signals sent in a vehicular network. We utilize different machine learning algorithms to make inference on traffic conditions, directly based on signals observed at a receiver. To verify the feasibility of this approach, we created a database using a ray-tracing simulator and a traffic simulator. Our database stores wireless channel realizations under various traffic conditions created by the traffic simulator. The results show that our method is able to distinguish different traffic intensities with accuracy of %93.0. It also estimates the number of vehicles on the road with mean absolute error of 6.83 for the scenarios with the maximum number of vehicles of 92. The proposed method can be used alongside the current traffic monitoring systems to increase the accuracy of the systems.