Using GTSAM for tracking an ECEF state

[Thomas Eberhard]

Hi I would like to use GTSAM iSAM for a state that is represented in ECEF coordinates, which is quite big in comparison to an indoor example state. Is there something I need to consider or could read up on?

Best,
Thomas

[Dellaert, Frank]

The optimization always happens in the tangent space, so in theory or should not be a problem. There is a possibility that there are numerical issues that we don’t know about in computing the derivatives, so I would look out for that. For example, you could add some tests for the factors that you’re most interested in and use examples with realistic ECEF coordinates. You can always use some anchos pose to establish a local frame (that’s basically what UTM does) but *in theory* this is not needed. Come to think of, UTM has fairly big numbers as well and we did not run into problems then, I believe (this was for a DARPA project a while back).

 

FD

--
You received this message because you are subscribed to the Google Groups "gtsam users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gtsam-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gtsam-users/831f34fe-3b0c-4725-9004-7d9408589735n%40googlegroups.com.

[Mike Sheffler]

FYI, I use iSAM2 in a problem involving satellites in (mostly low earth) orbit and observers on or near the surface of the Earth. Most of the state stuff is represented in ECEF (some satellite stuff is in an ECI frame), and I don't have any special issues. I am reasonably cautious about numerics (ECEF coordinates are large and satellites are fast), but it's almost never a problem.

If you are working with _really_ small quantities, like integrating measurements from an IMU, it might be trickier, numerically. However, you're already going to be having a bad time, on account of Z not being 'down' (or 'up').

Mike Sheffler

To view this discussion on the web visit https://groups.google.com/d/msgid/gtsam-users/BN7PR07MB483518AB7B5414156D7FE494CF139%40BN7PR07MB4835.namprd07.prod.outlook.com.