Validation of Google floating car data for applications in traffic management
conference paper
Accurate and reliable traffic state estimations are extremely important for road operators and
service providers, because they are the basis for decision making. Traffic movements are
generally measured by different types of sensors that are placed in or along the road
infrastructure. In this paper, aggregated and anonymized data from Google, originating from
mobile devices and apps, is analyzed for its potential to be used for traffic management. These
floating car data are speed time series at measurement locations. The traffic state estimations
from Google’s data are validated by comparing them with data from over 2200 sensor locations
on Dutch motorways for a period of 4 months. This dataset contains over 58 Million data points.
On Dutch motorways congestion and incidents are recurrent at a daily basis and traffic
management is essential. The coverage and accuracy is analyzed on link and route level by
computing the MAPE, a reliability indicator and the Jaccard similarity index to detect the tails of
jams. This paper contributes to existing literature by providing insight in the quality of the speed
data of a FCD provider with one of the highest penetration rates worldwide and by not only
showing the impact of replacing or fusing loop detector data with FCD on the quality of speed
indicators, but by also providing insights in the potential costs savings. It is concluded that
replacing some road sensors with Google data has limited and acceptable impact on quality and
can lead to substantial cost reductions.
service providers, because they are the basis for decision making. Traffic movements are
generally measured by different types of sensors that are placed in or along the road
infrastructure. In this paper, aggregated and anonymized data from Google, originating from
mobile devices and apps, is analyzed for its potential to be used for traffic management. These
floating car data are speed time series at measurement locations. The traffic state estimations
from Google’s data are validated by comparing them with data from over 2200 sensor locations
on Dutch motorways for a period of 4 months. This dataset contains over 58 Million data points.
On Dutch motorways congestion and incidents are recurrent at a daily basis and traffic
management is essential. The coverage and accuracy is analyzed on link and route level by
computing the MAPE, a reliability indicator and the Jaccard similarity index to detect the tails of
jams. This paper contributes to existing literature by providing insight in the quality of the speed
data of a FCD provider with one of the highest penetration rates worldwide and by not only
showing the impact of replacing or fusing loop detector data with FCD on the quality of speed
indicators, but by also providing insights in the potential costs savings. It is concluded that
replacing some road sensors with Google data has limited and acceptable impact on quality and
can lead to substantial cost reductions.
TNO Identifier
981662
Source title
TRB 97th Annual Meeting Compendium of Papers. Washington DC
Pages
1-15
Files
To receive the publication files, please send an e-mail request to TNO Repository.