Re: [Orekit Users] Kalman Filter

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

w.grossman@ieee.org a écrit :

> I am trying to "reverse engineer" the OreKit Kalman filter orbit  
> estimator.  I
> am very familiar with Kalman Filters but a newbie to OreKit.  Specifically, I
> am trying to understand what parameters are being estimated beyond the PV or
> orbit elements.  I find it a bit difficult to work through the  
> implementation.

The implementation was based on the existing implementation for the  
batch least
squares. The user is responsible for deciding what to estimate and what will
remain fixed. This is done by listing all available parameters  
(orbital parameters,
force parameters, measurements parameters) and calling  
setSelected(boolean), so
when the boolean is true the parameter will be estimated and when it is false
the parameter will keep its initial value throughout the run. By default, only
orbital parameters are estimated (these can be position-velocity if propagator
uses CartesianOrbit, but it could also be Keplerian, circular or equinoctial
parameters). Depending on the measurements and force models, other parameters
could be drag coefficient, SRP coefficients, gravitational constant, ground
station positions, measurements biases... It is also possible for expert users
to add their own parameters to be estimated (say if they create their own
measurements or measurements modifiers or force models). Everything that is
wrapped in a ParameterDriver can be estimated by either batch least squares
or Kalman filter.

> Do you have any papers or literature you can point me to that served as the
> basis for the design?  What was the performance of the design?

We relied only on classical general references for these packages. The
ParameterDriver mechanism is an Orekit invention. It was also adapted to
the Rugged geometry correction in remote sensing images for the spacecraft
viewing model refining. It works well and most importantly allows users
to add their own parameters in an existing orbit determination. I have
presented a paper about this at SpaceOps 2018 last month (see  
<https://www.researchgate.net/publication/325387275_Multi-satellites_Precise_Orbit_Determination_an_adaptable_open-source_implementation>).

best regards,
Luc

> Thanks.