Too stringent convergence threshold in Eckstein-Hechler mean parameters computation
In some cases, using the new computeMeanOrbit
method from EcksteinHechlerPropagator
fails to converge.
In all cases encountered up to now, this is due to the algorithm getting stuck in a loop with a cycle containing two or three very close orbits. The cycle therefore generates an infinite loop and the method failed after 100 iterations.
The cases encountered showed that the two or three orbits differences were a few nanometers in semi major axis and less than one pico radian in latitude argument. This was the latitude argument that was too stringent.
Having a user-definable threshold would imply changing a lot of constructors in an incompatible way (or duplicating the constructors, and there are already too many in the Eackstein-Hechler propagator). A workaround is to slightly raise the hardcoded threshold.