Leveraging Floating Car Data (FCD) for Traffic Forecasting on Sensorless Roads

Urban sprawl, with a consequent increase in the number of circulating vehicles, has accelerated the emergence of issues such as congestion, pollution, and road safety in cities. Although Intelligent Transportation Systems (ITS) provide efficient tools for traffic control, their widespread use is often limited by the prohibitive cost of deploying and maintaining fixed sensors. This work proposes a two-phase approach to estimate and predict traffic volumes on sensorless roads using Floating Car Data (FCD). In the first phase, FCD data are scaled based on real traffic flows obtained from a subset of sensor-monitored road sections, using machine learning techniques. In the second phase, the scaled volumes are used to make traffic flow predictions up to four hours into the future. Results demonstrate the potential of the methodology to expand the scope of urban traffic estimation and prediction and reduce the dependency on physical sensors, enabling more efficient and economically viable management of road networks.