Incident Detection by Spatiotemporal Analysis of GPS Data

We present a system to detect incidents causing traffic congestion in the road network by analyzing real-time GPS data. These data are collected from tracking devices installed in the vehicles or from drivers' smartphones. After positioning the GPS coordinates on the road map, the system assigns a traffic state to each road segment based on the velocities of vehicles, and generates alerts for incident based on a spatiotemporal analysis of these states. The system is validated by using GPS data simulated in typical traffic conditions in the city of Pisa, Italy. The results show an incident detection rate of 91.6% and an average detection time shorter than 7 minutes.