NavSat / GPS sensor noise has wrong description? Or gz-sensors wrong implementation?

[peci1]

navsat/horizontal noise says the noise is in meters:

sdformat/sdf/1.12/navsat.sdf

Line 10 in ed018d8

Noise parameters for horizontal position measurement, in units of meters.

NavSat sensor in gz-sensors adds noise in degrees of lat/lon:

https://github.com/gazebosim/gz-sensors/blob/d266990ace6bb0d4da1ac79e55731d3b47890616/src/NavSatSensor.cc#L180

this->SetLatitude(GZ_DTOR(iter->second->Apply(this->Latitude().Degree())));

There is a related issue in gazebosim/gz-sensors#325 .

We need to decide what is correct and what is not.

I'd vote for meters because that is how the error really manifests. Error in degrees of longitude has a different metric meaning for each latitude and is thus much less useful.

The migration consequences of choosing meters wouldn't be that bad. If somebody had specified noise of 1e-5 to get a +- meter precision, he'd end up with a super precise sensor. If we chose that degrees are correct, the user would end up with a GPS that jumps kilometers.

---

To be precise, this is what gz-sim currently does:

• position:
  • horizontal: degrees
  • vertical: meters
• velocity:
  • horizontal: meters/s
  • vertical: meters/s