GEO (SBAS) satellites handling

[gwbres]

Hello @ChristopherRabotin,

In RINEX, the ephemeris for GEO and Glonass vehicles describe a state vector, not keplerian parameters, and we are expected to propagate this state in time (either backwards or forward) to get the state at the desired instant.
The typical interpolation formula, that we are expected to use is the follwoing:

r(t) = r(t0) + v(t0) *dt + 1/2 * a(t0) * dt^2 

for example, this is how RTKLIB does it.

Currently I do not support that interpolation, so I cannot consider any GEO or Glonass vehicles for my navigation processes, and I'm investigating what needs to be done to change that.

If I were to instead deploy a Nyx propagator from that reference state, can I expect significant improvements compared to the basic interpolation method ? thanks

[ChristopherRabotin]

Hi Guillaume,

I assume that a(t0) is the acceleration at the initial time, and that it is provided in the RINEX file. Nyx only requires a position and velocity, and the accelerations are computed for the user given the dynamics, typically a minimum of Earth, Sun, Moon gravities, and possibly a solar radiation pressure if the reflectivity coefficient and area is known.

Given the r(t0) and v(t0), Nyx can propagate the state of the vehicle into the future (or in the past if that's what you're seeking). You should expect a much higher fidelity with this approach compared to a simple interpolation, but it is significantly more computationally involved under the hood.

[gwbres]

yes, that is what I meant for both.

That's very interesting, so it's good I asked, it makes sense to eventually propose it then.
So my plan is to conclude the basic interpolation method, and I will provide de Nyx propagator under a dedicated feature (similarly to the pending SP3 PR).

Although that might be off topic a little bit, what is the order of performance gain we can expect, compared to the basic interpolation ?
Also, it is not clear to me how the performance of the propagation decays the further we propagate, do you have some reference for these aspects ? thanks again 👍

[ChristopherRabotin]

I expect the interpolation as it's currently done to be pretty low fidelity and work only for a few hours at most (I've seen interpolation methods work for longer, but they're usually much more details than a quadratic interpolation). In contrast, with Nyx, you can expect the propagation to be valid for many decades.