Prediction of Congestion Patterns on Freeways
The project investigates prediction approaches for the occurrence of congestion patterns, mainly on southern Bavarian freeways, using machine learning methods (artificial intelligence). The aim is to make an occurring traffic congestion event predictable at an early stage in terms of time, space and classification on the basis of historical events along a larger section, in particular its end. The desired improvement over classical methods based on travel times is achieved on the one hand by working on congestion patterns, i.e. using the entire spatial and temporal information of congestion, and on the other hand by using state-of-the-art machine learning methods. In this, the Chair of Traffic Engineering applies previous research on congestion detection (definition, recognition, classification into different congestion patterns), traffic state estimation based on loop detector and vehicle trajectory data with machine learning methods, as well as traffic data analysis of large data sets with a wide variety of data science methods.