A Novel Approach To Traffic Flow Estimation based on Floating Car Data and Road Topography: Experimental Validation in Lyon, France

Today, Floating Car Data (FCD), which are obtained from the GPS sensor measurements of vehicles, represent a particularly interesting source of data for road traffic managers. Indeed, they give information on the traffic flow and the average speed on a whole territory, enabling many applications such as congestion or air quality prediction. However, this information is partial in the sense that only the vehicles equipped with GPS are counted. Models are therefore necessary to convert the partial FCD count into the total traffic volume. In this study, we propose to cross the FCD with loop detectors data, which capture the real traffic flows but only at certain locations, in order to train a flow prediction model. The proposed approach is based on multiple linear regressions, the weight of each regression being determined by a Random Forest (RF) algorithm. The methodology is calibrated and tested on data from Lyon, France, collected in 2021.