What happens if unsupervised craft enters atmosphere?

[THX1138]

In the screenshot you can see the situation. I have two craft that I'm controlling on similar timescales. One is returning to Kerbin using air-braking to expend excess energy (periapsis is 42 km) and the other is placing a satellite before heading to Mun. My question is, why KSP just declare the unattended craft a wreck if it enters the atmosphere without me controlling it? Or will it do the calculations properly and realise that the unattended craft should actually be able to orbit a few times before landing happens?

(click to enlarge)

Edited December 23, 2014 by THX1138

[Anquietas314]

Err, what screenshot? You didn't post one. Never mind, either forums derped or you edited it and it's not showing the edit. However, to answer the title: it gets deleted if it goes below about 30km. No refunds, any Kerbals on board are killed. If periapsis is above that, it just stays in orbit. Atmospheric drag does not apply unless it's within 2.5km of the active ship (this also will prevent deletion).

You should be fine to leave that one though until you match the target orbit. It's so far out it probably won't enter the atmosphere any time soon.

Edited December 23, 2014 by armagheddonsgw

[THX1138]

I edited the post because I forgot the screenshot. I did it immediately so I guess it doesn't count edits that are made within a certain time limit.

Does that mean that the orbit wouldn't change at all on the unsupervised craft? TBH that's still not ideal, I don't think. It's probably saved some people some heartache but I still wanted the air-braking to occur while I did my other stuff.

[Anquietas314]

I edited the post because I forgot the screenshot. I did it immediately so I guess it doesn't count edits that are made within a certain time limit.

Does that mean that the orbit wouldn't change at all on the unsupervised craft? TBH that's still not ideal, I don't think. It's probably saved some people some heartache but I still wanted the air-braking to occur while I did my other stuff.

I believe there's a mod somewhere that makes physics work on ships that enter the atmosphere outside of your control, but I have no idea if it's still maintained or what it's called. The problem is you can time-warp when you're in another ship. How should the game handle your trajectory through the atmosphere? If the ship has wings things get really complicated really, really fast. If the aerodynamics model was any good at all, it would get really really complicated, really, really, really fast.

[lirg03]

I believe when a ship 'on rail' enters the atmosphere, either nothing happens, or it gets deleted. Depends on how deep the PE is. For Kerbin, the threshold is ~23km, any ships on rail enters atmosphere below that height will be deleted. (ref: http://wiki.kerbalspaceprogram.com/wiki/Atmosphere)

[THX1138]

Well the game could handle the situation in different ways. For a start it could warn the player that they have another ship entering the atmosphere of planet (that seems a pretty user-friendly thing to do). Then, if the player doesn't want to switch to the ship in question (madness!), the game can just put the calculations "on the back burner" and perform them when the spare GPU cycles are available. If the player tries to change to the ship in question before the calculations are complete, they can watch the simulation as a recording of what happened. Not all calculations should need to be performed in real-time if the user is off lollygagging elsewhere.

[Anquietas314]

THX1138: The last time I checked (admittedly, a while ago), KSP's physics (mostly managed by unity afaik) is done on the CPU. The amount of lag I get for even relatively small ships (say, 100-150 parts) on my old hardware would certainly suggest it. Most of the time the GPU's doing diddly squat the entire time, while the CPU's pegged at about 50% (quad core), half of which is KSP; which means it's CPU-bound and mostly serialized (i.e. more or less single-threaded).

As to your proposed solution: what happens if the ship is about to enter the atmosphere at the same time you're in the middle of a long, possibly complicated burn that you absolutely cannot stop in the middle of? Or here's a really nasty one: suppose you have not one but two or more ships entering the atmosphere (possibly different atmospheres!) at the same time?

[Claw]

I believe when a ship 'on rail' enters the atmosphere, either nothing happens, or it gets deleted. Depends on how deep the PE is. For Kerbin, the threshold is ~23km, any ships on rail enters atmosphere below that height will be deleted. (ref: http://wiki.kerbalspaceprogram.com/wiki/Atmosphere)

Yes.

which means it's CPU-bound and mostly serialized (i.e. more or less single-threaded).

And yes. KSP is not multi-threaded. Also, the Unity engine does an amazing amount of part-to-part comparison each frame. That's why lag increases exponentially with part count.

Cheers,

~Claw

[Anquietas314]

And yes. KSP is not multi-threaded. Also, the Unity engine does an amazing amount of part-to-part comparison each frame. That's why lag increases exponentially with part count.

Cheers,

~Claw

There are pretty simple ways that could be brought down to something closer to polynomial complexity without adversely affecting physics accuracy to a noticeable degree, at least in theory; for example processing part physics (and anything else) in stages in a manner similar a parallel prefix sum algorithm. How practical that is in Unity I don't know though, and of course there's always the possibility of complications.

[WuphonsReach]

I believe there's a mod somewhere that makes physics work on ships that enter the atmosphere outside of your control, but I have no idea if it's still maintained or what it's called.

Stage Recovery (and yes it works with 0.90).

Put parachutes on your booster stages, get about 1/2-2/3 of the value back. Put parachutes + heat shields on your second stage that slots you into a stable orbit, get 1/4-1/2 of the value back (just make sure to decouple from the last stage while your Pe is still under 50km).

Makes those early launches much more affordable / profitable since you can get back at least 1/4-1/3 of your launch costs right after launch.

[Anquietas314]

Stage Recovery (and yes it works with 0.90).

I do actually use that mod, but that doesn't seem to be how it works. The parts "land" much too soon for that to be what it's doing; instead, I think it simply waits until the parts hit the deletion cut-off and then computes terminal velocity based on all attached parachutes deploying and presumably something like MechJeb's atmosphere landing prediction. It also produces slightly incorrect results - if you fly the stages manually, you will very reliably get less return than if you let the mod take care of it. I imagine this is considerably worse with FAR, though I don't use that at the moment.

EDIT: The mod just uses the point when ships are deleted in the atmosphere. Confirmed both by observing debris fall back down and mention of implementation in the relevant thread.

You can easily get near 100% of the value back - you just need more parachutes; 2 radial ones for a 3x1.25m tank launcher seems to be about right.

Edited December 25, 2014 by armagheddonsgw