Re-computed estimated measurements

[MaximeJ]

Hi Romain, Bryan,

I’m not sure I understand your debate.

@bcazabonne

I think you are right. It looks like the method returns the estimated measurements based on the state optimized at the penultimate iteration.

If this is indeed how it works, it looks like a bug to me.

@bcazabonne

I see that the evaluations (i.e., the object returned by the getLastEstimation() method) is updated before the differential correction is applied to the state vector. In other words, the method gives the observed and estimated measurements used to compute the residual vector of the differential correction for the last estimation.

This is what Romain wants I think, the last estimated measurements when the convergence checker says it is ok to stop the iteration process. No ?
If you look up for example in the GaussNewtonOptimizer optimize method it does:

• Calls LeastSquareProblem.evaluate function (l. 163): current = lsp.evaluate(currentPoint);
  In Orekit, this calls AbstractBatchLSModel.value method which runs the propagation, compute the residuals and Jacobian and updates the last estimations (this is done at the end of the method: observer.modelCalled(orbits, evaluations);)
• Then checks convergence at line 169 and returns current point if the estimator has converged.

So, in my opinion, you should get the last estimations that lead to convergence. Is this what you are looking for Romain ?
And how did you recompute the theoretical measurements Romain ?

@Serrof

If that is confirmed, what would be the easiest way to retrieve the estimated measurements according to the final, converged (if applicable) iteration of the OD process (at least in the batch least square case )?

In this former post I uploaded some dedicated handlers to get the map of (observed, estimated) measurements from a propagator and a list of measurements in input. Maybe you could use something similar in Python ?

Maxime