Unexpected error with drag during one day numerical integration
In some specific conditions of propagation, the drag models raise an error of altitude too low which is not expected to happen during the one day propagation.
A file is attached to reproduce the problem. UnexpectedCrashWithDrag.java
We found some weird workarounds but that sounds like bad ideas :
- use a fixed step integrator (so far we use a variable step with max step of 100s)
- converting the equinoctial orbit into keplerian orbit before building the spacecraft state (looks like a side effect of the next point)
- update only the anomaly obtained by consecutive conversion of the orbit : equinoctial -> keplerian -> equinoctial