AI-Aided Prediction of Traffic Flow Using Real-World Urban Traffic Data

The increasing traffic load in cities around the world demands solutions for intelligent urban traffic management. A key component of such systems is monitoring and understanding traffic conditions in real-time to dynamically adapt and optimize its flow. Therefore, it is essential to retrieve accurate and extensive realtime information on urban traffic. While floating car data provides traffic measure-
ments across most of the road network, it only captures a subset of all vehicles on the road. In contrast, stationary traffic sensors provide accurate data, but are spatially sparse. This study seeks to generate accurate traffic flow predictions from floating car data by training a model on stationary sensor data. The goal is to gain a comprehensive understanding of the current traffic conditions, while also paving
the way for improved predictions of future traffic flow. Leveraging data from the city of Herne, Germany, we construct different machine learning models aimed at predicting actual traffic flow using incomplete floating car data.

DOI 10.3233/ATDE83