Scaling the Real-Time Traffic Sensing with GPS Equipped Probe Vehicles

In Intelligent Transportation System, GPS has become a major source of floating car data. GPS measurement from vehicles can be collected and be further analyzed for real-time urban traffic sensing or monitoring. However, the main challenge coming to the real-time traffic estimation system is how to scale up the system easily when the number of GPS vehicle probes grows dramatically. In this paper, a scalable real-time traffic estimation system based on distributed stream processing is developed. Differing from the traditional batch- style distributed computing techniques, e.g. MapReduce, the distributed stream processing focus on not only distributed computing but also real- time and in-memory computing such that latency is reduced. We show the system architecture, data flow and some implementation experiences for estimating urban traffic using Twitter Storm, which is the open source distributed stream processing framework. The experiment results illustrate that our system can scale well and scale up easily as the input GPS data increase. It is effective and efficient for applying stream computing in real- time traffic estimation system.