Large-Scale Estimation of Arterial Traffic and Structural Analysis of Traffic Patterns from Probe Vehicles

Estimating and analyzing traffic conditions on large arterial networks is an inherently difficult task. The first goal of this article is to demonstrate how arterial traffic conditions can be estimated using sparsely sampled GPS probe vehicle data provided by a small percentage of vehicles. Traffic signals, stop signs, and other flow inhibitors make estimating arterial traffic conditions significantly more difficult than estimating highway traffic conditions. To address these challenges, a statistical modeling framework is proposed that leverages a large historical database and relies on the fact that traffic conditions tend to follow distinct patterns over the course of a week. This model is operational in North California, as part of the Mobile Millennium traffic estimation platform. The second goal of the article is to provide a global network-level analysis of traffic patterns using matrix factorization and clustering methods. These techniques allow the characterization of important spatial configurations in the network and the analysis of traffic dynamics at a network scale. Traffic patterns are identified that indicate intrinsic spatio-temporal characteristics over the entire network and give insight into the traffic dynamics of an entire city. By integrating our estimation technique with our analysis method, a general framework is achieved for extracting, processing and interpreting traffic information using GPS probe vehicle data.