Re: [Orekit Users] DSST & Sequential Estimation

[MAISONOBE Luc <luc.maisonobe@c-s.fr>]

w.grossman@ieee.org a écrit :

> If I read the source material correctly, the advantage that DSST propagation
> has over Cowell is that large steps can be made while including special
> perturbations like drag, solar pressure, etc.

This is one of the advantages. Another one is that is can get both osculating
and mean elements, depending on what you need. Mean elements are very useful
for station keeping and orbit control as you do not want to waste fuel
compensating for harmonic terms that indeed have a zero mean (but not all over
the same period).

The two main use cases for DSST I know of are:

 - long term propagation (for example end of life analysis)
 - station keeping

There are other uses, of course, but probably more specialized (like for
example precisely analysing the behaviour of short periodic terms to set
up reference orbits for some specific missions).

> In performing sequential estimation where updates come every one or ten
> seconds, why is this large-step capability an advantage worth the complexity?
> Assuming that there are even occasional measurement outages of one or two
> minutes minutes, is this large-step capability worth the complexity?  What am
> I missing?  Is numerical roundoff error accumulation a driving force?

No. For estimation with very frequent measurement, DSST is not particularly
better than numerical propagator, except if you are really interested
in mean elements, where DSST is great!

> In a related question, the advantage equinoctal elements have over
> conventional elements isn't clear.  I understand that the conventional
> elements have singularities, but so do equinoctal--just in different places.
> For example, what happens when the eccentricity is near-zero and the periapse
> location is unstable?  Aren't we trading one set of singularities for another
> set?

No. Equinoctial elements are singular neither at zero eccentricity nor at
zero inclination. They have been used for years for geostationary
satellites for example.

Equinoctial elements are singular at inclination = 180° (i.e.
perfectly retrograde orbits), but the original DSST formulation in fact even
removes this singularity. We did not do it in the Orekit implementation
(i.e. our implementation *is* singular at i = 180°). The reason we did not
do it is that we have never seen a real satellite with i=180°. To be honest
I don't know what kind of mission would require that. So the reamining
singularity is acceptable.

best regards
Luc

> Thanks.
>
> Walter