We put our virtual sensors to the test by working with the Utah Department of Transportation (UDOT) to conduct a study of our virtual sensors when compared to their detectors in the field.
The results proved that we’re generating measurements closer to the actual counts than the measurements gathered from any singular source. Our virtual sensors were 94.4% accurate.
They outperformed detectors everywhere they were placed,
generating more accurate results in areas with high detector volumes, low detector volumes, and, of course, in areas with no detectors at all. Additionally, our sensors had a GEH score, another commonly used metric to determine the rate of error, of 1.6. Any number under 5 is considered acceptable. The physical detectors scored a 3.5.
Most impressively, we achieved these results in areas with only 0 and 3.2%
of the cars registering as connected vehicles. As millions of CVs join our roadways each year, we’ll be able to gather even more GIS data, and our measurements will become more reliable.
We’ll also continue to add new datasets, such as additional forms of sensors.
You can read our full case study here