How to fix a tumbling Kerbal

[Loren Pechtel]

I bumped the target rocket at about 1 m/s and was set tumbling.  The Kerbal is *not* ragdolled but I can't find any solution to get him under control.

Normally you can kill rotation with a quick bit of timewarp but it does no good in this case.  He keeps trying to point in the right direction, oversteering and burning the jets in the opposite direction, again oversteering.  If I try to timewarp without turning off RCS the jets continue to burn (but not jump around) during timewarp.  (Yes, timewarp, not physics warp.  How can this even happen?)

It looks like the game is unable to cancel rotation that is on multiple axes at once.

Short of caging him in some sort of cargo bay I see no way to solve this mess.  (Beyond the obvious of HyperEditing him back to Kerbin, which worked.)

[bewing]

Pick a new direction for him to face, and when he's halfway through an oscillation, hit the spacebar. It sounds like he happened to get stuck in a magic orientation that happens to oscillate -- so just breaking the pattern might be enough. Or try to hit that timewarp halfway through an oscillation. Or both.

[Loren Pechtel]

21 minutes ago, bewing said:

Pick a new direction for him to face, and when he's halfway through an oscillation, hit the spacebar. It sounds like he happened to get stuck in a magic orientation that happens to oscillate -- so just breaking the pattern might be enough. Or try to hit that timewarp halfway through an oscillation. Or both.

I've already tried both of those.

And after recovering him I had it happen again, this time the velocity was unknown but probably under 1/2 m/s.

[Loren Pechtel]

I believe I have solved this.

Leaving it in precision mode virtually guarantees an uncontrollable tumble, although it generally recovers if you take it out of precision mode.

I watched a Kerbal go into oscillations while simply drifting in space, no impulse applied.

[Loren Pechtel]

And now I found an even nastier version, although I'm not sure who is to blame, KSP or MechJeb.

I was trying to land a large rocket on Minmus and of course it doesn't turn well--which caused the same problem.  Fire the engines, overshoot, fire them the other way, overshoot, back and forth, back and forth.

I think both cases fundamentally come down to the same problem:  I think KSP does not lead it's target when it attempts to stop rotation.  So long as there is plenty of turn authority this isn't a problem, but when it's lacking...

Since it was behaving just like the EVA Kerbal I think it's unlikely that MechJeb is the culprit, I only mention it at all because I was using it's landing guidance at the time.

[bewing]

It's actually more complicated than that. The SAS algorithm is not what is used for Kerbals on EVA. They have their own control algorithm.

And the SAS algorithm does, in fact, lead the target for stopping rotation most of the time. However, you are correct that there is a known bug -- when control authority is very limited, you do tend to get oscillations.

http://bugs.kerbalspaceprogram.com/issues/13099

 

[Loren Pechtel]

17 minutes ago, bewing said:

It's actually more complicated than that. The SAS algorithm is not what is used for Kerbals on EVA. They have their own control algorithm.

And the SAS algorithm does, in fact, lead the target for stopping rotation most of the time. However, you are correct that there is a known bug -- when control authority is very limited, you do tend to get oscillations.

http://bugs.kerbalspaceprogram.com/issues/13099

 

Maybe the real problem is a bug in the SAS prediction.  With good authority it gets damped out and goes unnoticed.  I have repeatedly noticed small overshoots even with good authority.

Why wouldn't Kerbals and SAS use the same pointing logic?