A smartphone-based system for automated congestion prediction

Abstract

Accurate collection of traffic data is essential, tactically for efficient highway operations and strategically for capacity planning purposes. Currently, this traffic data is collected by physical sensors, which suffers from many limitations. These sensors gather limited measurements of vehicle speeds and times from their fixed locations and cannot systematically acquire mobility dynamics of individual vehicles and the interactions among them. Also, because deploying and maintaining physical sensors is expensive and time consuming, typically they are installed to measure traffic only on the busier road segments within a transportation network. To overcome these issues, this paper reports on a general-purpose, location-aware, smartphone-based traffic monitoring system as a simple and an inexpensive alternative to collect dynamic vehicle data extensively through a transportation network. A privacy-preserving smartphone application was developed, deployed and tested on the network surrounding the University of Connecticut (UConn) in Storrs. It securely transmits individual trip data over the Internet to the servers hosted at UConn. Preliminary experimentation suggests that crowdsourcing the collection of traffic data with smartphones can be cost effective and can lead to richer data sets spanning the entire web of roadways.