I’m working now on Batch Least Squares after my first steps with orekit and IOD.
I succedded to run the code correcly on my environement, and now I’m trying to use this code from my own data.
Here the steps I followed :
- Simulation of an a trajectory from Keplerian parameters & Keplerian propagator (got 720 *
- Exctration of some points of this trajectory to fit with the field of view of my Radar (got 131 *
- Noizing of those points from Sigmas of my Radar passing threw azimuth/elevation (got 131 *
- IOD (Gibbs, Herrick-Gibbs and others) from 3 points (first, middle and last) of those Noised measurements (got
Now I joining the code of GorgiAstro, just before the definition of the propagator. I use the DormandPrince853IntegratorBuilder and I have some troubles with the convergence of the BatchLS.
First of all, I have sometime an error message saying that it’s impossible to compute the excentric anonaly from the true anomaly (50 interations). Sometimes I have another message saying that the Jaccobian matrix is singular for my orbit. Sometimes the BatchLS is not able to converge after 1000 iterations, and finally sometimes a result is computed, but I’m not really satified with them (less good than the IOD estimation).
I was wondering if the problem is not comming from the fact that my estimator should use a KeplerianPropagator in order to fit with the initial Simulated orbit, But the object BatchLSEstimator is not able to take as an input a KeplerianPropagatorBuilder, it just needs a NumericalPropagatorBuilder.
I’m just a begginer with OD problems and orekit so it’s a complete mess right now, I don’t really know what to do. Is the Levenberg-Marquart method is not supposed to converge all the time ?
Can’t I just put a KeplerianPropagatorBuilder in input of the BatchLSEstimator ?
Thank you in advance for your help !