Traffic volume information is critical for traffic management and control. Traditionally, traffic volumes are primarily measured by fixed location sensors. However, the high installation and maintenance cost of fixed location sensors often leads to the missing data problem and the low coverage problem. In recent years, researchers have proposed to solve the problems using probe vehicle data. Nevertheless, when only low market penetration probe vehicle data are available, it is difficult to estimate real time traffic volume information if we consider each time slot and each road separately Noticing that traffic volumes in a transportation network are correlated spatially and temporally, we try to capture the correlation by fusing probe vehicle data and partial fixed location sensor data, which are complementary to each other In this work, we propose low rank representation methods to estimate the unknown traffic volumes. The proposed methods take advantage of the correlation of traffic volumes in different locations and different time slots and thereby achieve good estimation accuracy even if the probe vehicle data are sparse. Validation results show that the proposed methods can solve the missing data problem and the low coverage problem at the same time, and they have great potential for real world implementation.

Yan Zhao is currently a PhD candidate in the Mechanical Engineering Department at the University of Michigan. Yan also works with Professor Henry Liu in the Department of Civil and Environmental Engineering on traffic state estimation and transportation network analysis using trajectory data.