Technical Note on Data Quality
This page complements the data quality page.
Last updated
This page complements the data quality page.
Last updated
The positioning system onboard our vehicles is designed to provide precise and reliable information by aggregating data from various positioning sensors aboard the vehicle.
For each recorded vehicle position, the system provides uncertainty estimates corresponding to the three axes of vehicle movement (x, y, z). These estimates are represented by standard deviations. There are statistical relationships between these standard deviations and metrics that reflect data accuracy.
We have established quality thresholds based on these parameters.
To evaluate the accuracy of our positioning, we use two metrics to qualify the corresponding 3D accuracy:
This metric refers to the Mean Radial Spherical Error (MRSE). This metric is calculated at each point of the trajectory based on standard deviations. Mathematically, this metric provides an estimate of the probability that the truck’s coordinate at time t is within a probability of 61% inside an error sphere.
Other metrics indicating 3D accuracy exist. We use one of them, the SAS99, also called the Spherical Accuracy Standard (SAS). Like the MRSE, this metric is calculated at each point of the trajectory. This metric provides an estimate of the probability that the truck’s coordinate at time t is within the error sphere. The radius of the error sphere is equal to the calculated value.
Our primary objective is to provide our users with a precise and reliable positioning experience. For this reason, we have chosen to use the SAS99, which indicates a probability sphere of 99%.
When Jakartowns is set to display the SAS99 value, it means that the origin point of the vehicle at time t is determined with a probability of 99% (i.e., with high certainty).
In the example above, the position is contained within an error sphere of no more than 4 cm.
For more details on the formula to calculate SAS99, please refer to the bottom of this page.
Our quality criterion is based on the analysis of SAS99 percentiles for a capture mission. The capture mission is a continuous measurement period that rarely exceeds 150 minutes.
We use the 80th percentile and 95th percentile of SAS99. The 80th percentile is a statistical measure indicating that 80% of SAS99 values are below this value for a given mission. The 95th percentile indicates that 95% of SAS99 values are below this value for the mission.
We will automatically publish a mission if one or both of the following conditions are met:
The 95th percentile of SAS99 is below 0.174 meters.
The 80th percentile of SAS99 is below 0.087 meters.
There are several environmental and technological factors that may justify publishing a mission even if it exceeds the above thresholds. The presence of vegetation or buildings that limit the portion of the sky visible to the GNSS antenna is an example of such factors.
In such a case, the mission publication will be manual, and a quality note will be associated with it. The user will be able to view this information to make an informed decision on using the mission based on their needs. If applicable, the information will be available in the Jakartowns web interface in an information banner.
where:
where:
In Jakarto's processing chain, we regularly refer to the 3D standard deviation or Mean Radial Spherical Error (MRSE). It is estimated that the SAS99 value corresponds to approximately 174% of the MRSE.
For example, on a trajectory map where:
the MRSE is 0.05 m, the SAS99 is approximately 0.08 m.
the MRSE is 0.10 m, the SAS99 is approximately 0.17 m.
For more information on these technical issues, we invite you to consult the following references:
standard deviation on the x-position standard deviation on the y-position standard deviation on the z-position
standard deviation on the x-position standard deviation on the y-position standard deviation on the z-position\