Ensuring continuity in multi arc-propagation

[mrevellino]

Hi everyone!
I’m currently trying to use Tudat to estimate and orbit starting from observations from the MPC. Given the availability of the observations, I am applying multi-arc propagation and integrating forward starting from the first available set of observations, and backwards starting from the last set, terminating each propagation at the middle epoch.
I was wondering if, in cases like this, it is possible to enforce continuity between the 2 arcs in Tudat, as I wasn’t able to find the relative documentation.
Thank you!
Margherita

[DominicDirkx]

To make sure I understand correctly, you have times t0<t1<t2, with observations around t0 and observations around t2. You want to propagate two arcs:

One propagating (forward) from initial state at t0 to t1
One propagating (backward) from initial state at t2 to t1

If this is the case, then the two states (one forward from t0, one backward from t2) will not match, but will be a result of your initial state and whatever your dynamics is. But, perhaps I am misunderstanding your setup.

A more general question: why use a multi-arc propagation, and not a single arc propagation centered at t1, where you propagate backwards to t0 and forward to t2 using this: https://py.api.tudat.space/en/latest/propagator.html#tudatpy.numerical_simulation.propagation_setup.propagator.non_sequential_termination termination setting

If it's an MPC object, I'd assume that you want to end up with a single continuous arc

[mrevellino]

Yes, exactly. The goal is in this case to estimate both the states at t0 and t2, and I was thus trying to see if a possible constraint could be set such that the states obtained at t1 from the 2 propagations had to be equal.
I tried what you suggested, but I saw no improvement in my estimation, so I was seeing if there were alternative roads to explore :)
Thank you!

[DominicDirkx]

If you do a multi-arc estimation, you allow the state to not be continuous at the arc splitting time, so for your application I think a single-arc propagation is the way to go.

If you're not getting a good fit, a first thing to try is to test the dynamical model. Rather than fitting from the MPC astrometry, try fitting to the MPC or Horizons ephemeris. That is, use Cartesian states from reference ephemerides (also called pseudo-observations) to fit the orbit of your body. That way, you can see -in an ideal situation- how well yoyr dynamical model can reproduce the reference ephemeris. Any mismatches you see from this will come entirely from the dynamics, which you can then modify in your models to improve the fit. Once you're confident in your dynamical model you can move to fitting based off of the astrometry instead, and any misfit you see can then be more robustly attributed to the data.

[mrevellino]

Ok, perfect! Thank you for the suggestions!