Hierarchical Graph Structures for Congestion and ETA Prediction
This work addresses traffic forecasting for urban planning and navigation, presenting an incremental improvement through hierarchical graph structures.
The authors tackled traffic prediction by developing a hierarchical graph neural network that uses road topology from OpenStreetMap to improve information flow between intersections and paths, resulting in a multi-task model for congestion and ETA prediction with released code and models.
Traffic4cast is an annual competition to predict spatio temporal traffic based on real world data. We propose an approach using Graph Neural Networks that directly works on the road graph topology which was extracted from OpenStreetMap data. Our architecture can incorporate a hierarchical graph representation to improve the information flow between key intersections of the graph and the shortest paths connecting them. Furthermore, we investigate how the road graph can be compacted to ease the flow of information and make use of a multi-task approach to predict congestion classes and ETA simultaneously. Our code and models are released here: https://github.com/floriangroetschla/NeurIPS2022-traffic4cast