Real-time urban traffic information extraction from GPS tracking of a bus fleet

We present a novel system named Speed-O to estimate in real time the average speed of traffic on every section of urban roads in the city of Thessaloniki, Greece. Speed-O processes the telemetry data reported by the fleet of more than 600 public transportation buses. The system is solving for the mean speed at different temporal and spatial scales much finer than the raw telemetry data. Speed-O is comprised of a large and sparse model of the road system, the bus routes, and the fleet dynamics together with a fast solver that renders the model solution. The solver is multi-layer and iterative. It exploits the physical properties of the problem to operate efficiently in realtime on a multicore processor. We demonstrate the Speed-O system using the Google Maps API as an interactive map that refreshes every minute with the average speed displayed as color-coded road sections.