MyTraffic has developed a unique methodology. We couple our connected vehicle data with automatic and manual counts from open and qualitative data from our providers to ensure that the adjustment of our raw data is as close as possible to the field experience.
More than 10k field counting points are used for calibration purposes. We have an automated process to collect data and assess the quality of a counting point. Furthermore, we only keep the most recent ones to guarantee the freshness of our predictions and ensure that our Machine Learning algorithms do not become disconnected from the reality on the ground.
We then associate these field measurements with our massive data sets to scale our algorithms to Europe, moving from the discrete nature of field measurements to continuous estimates covering the entire territory.
First, daily, we couple massive connected vehicle GPS data and road network data to associate vehicle position signals to road segments. A road segment is defined as a portion of a road between two intersections.
However, at this stage the data is still noisy; vehicles may not emit position signals or pings at a continuous frequency for various physical and environmental reasons. To clean and enrich the data, we use our path reconstruction algorithm to come up with a preliminary estimate of daily traffic.
Then through several computation steps, including adjusting based on field counting data and validation checks, we can extrapolate the best estimates for any road segment.
We provide more than 80% coverage of the territory in the European countries (France, United Kingdom, Germany, Spain, and Italy) where we are currently deployed.
Comments
0 comments
Article is closed for comments.