Traffic Congestion Prediction Using Vehicle Trajectories from Drones and Combined Graph Attention and Temporal Convolutional Network

Congestion prediction is an important and challenging task in intelligent transportation systems. This task should effectively capture the spatio-temporal dependencies in road networks because traffic congestion can propagate spatially over time. With recent advancements in computer vision, the use of Unmanned Aerial Vehicles (UAVs) has gained increasing attention as a tool for traffic data collection. In this work, we evaluate the performance of a Temporal Convolutional Network (TCN) per edge versus a Temporal Graph Neural Network for predicting the congestion ratio of road segments in a road network. The modeling of the graph structure was extracted from open vehicle trajectories data recorded using a swarm of drones. Two different models were implemented and compared for the task of congestion prediction per edge aiming to capture structural correlations between edges, spatial effects, and temporal traffic patterns.