A Graph-based U-Net Model for Predicting Traffic in unseen Cities

TL;DR

This paper introduces a novel graph-based U-Net model that enhances traffic prediction accuracy and generalization to unseen cities by integrating graph layers sensitive to geographical topology.

Contribution

It combines U-Net with specialized graph layers to improve spatial generalization in traffic forecasting across different city road networks.

Findings

Improved accuracy in traffic prediction for unseen cities.

Enhanced spatial generalization over traditional U-Net models.

Effective adaptation of pooling and upsampling for graph-structured data.

Abstract

Accurate traffic prediction is a key ingredient to enable traffic management like rerouting cars to reduce road congestion or regulating traffic via dynamic speed limits to maintain a steady flow. A way to represent traffic data is in the form of temporally changing heatmaps visualizing attributes of traffic, such as speed and volume. In recent works, U-Net models have shown SOTA performance on traffic forecasting from heatmaps. We propose to combine the U-Net architecture with graph layers which improves spatial generalization to unseen road networks compared to a Vanilla U-Net. In particular, we specialize existing graph operations to be sensitive to geographical topology and generalize pooling and upsampling operations to be applicable to graphs.