N-Body Physics in KSP

[Sovnheim]

Hi all,

I seen some threads discussing the implementation of n-body physics (rather the conics-thingy what the system currently uses) to KSP.

Out of curiosity, I have read part of the official thread for the N-Body Gravitation Mod for KSP, but I have a problem - it often looks like this :

The predictions are currently using the same integration method (McLachlan and Atela's 1992 optimal 5th order method), with the same splitting of the Hamiltonian (kinetic + potential energy), this is somewhat usable but unacceptably slow. I am currently implementing as many symplectic integrators as I can find in order to compare them, and I will also be comparing various splittings (as as nonsymplectic methods in use for computing ephemerides).

I have a degree in literature, not theoretical physics.

I would be curious to understand the actual benefits in terms of gameplay would such implementation have ?

[Superfluous J]

The benefit is it's more realistic.

The problems are so many I won't bother listing them. There are at least as many old threads about this as there are problems with the idea.

[Xavven]

N-body physics would be bad for gaming enjoyment possibly because all of your vessels would be subject to orbital perturbation (http://en.wikipedia.org/wiki/Perturbation_%28astronomy%29). In short, you could timewarp ahead in what you thought was a stable orbit, only to find you've been ejected from the system, or crashed into something else, etc.

Not to mention Vall becomes a planet in short order:

[Whirligig Girl]

Not to mention Vall becomes a planet in short order:

Only if you don't account for the center of mass of the Joolean moons. If you do that, then the laplace resonance of the moons is stable and Vall doesn't fly out to infinity.

[Eric S]

The benefits: orbits would be more like those in reality rather than the mathematically perfect ones we have in KSP. You could park things at L4/L5, or at L1/L2/L3 if you want to babysit them. Note however, that you can already fake L4 and L5 if you're precise enough. L1-L3 would mostly just get used to simulate real life missions where you want a satellite in earth's shadow (telescopes, usually) or never in earth's shadow (solar observatories, I think). Those Lagrange points aren't actually stable, so some station keeping would be required.

We would get some other interesting things that most players wouldn't notice, like a planet and moon would rotate around their barycenter rather than the center of the planet.

The cost would be a major overhaul of the map mode and maneuver node system, in addition to the work necessary to convert the game over to n-body. Orbits wouldn't be mathematically precise, so you couldn't draw an ellipse to represent the orbit you're currently in on the map, since that orbit wouldn't be a perfect ellipse and even if it were, would shift over time, something the current map just can't represent.

Basically, a lot of work, mostly for things that your average player wouldn't notice.

Which isn't to say that I'm opposed to it, just that I think there are other things with a better cost/reward ratio.

[Kerbart]

I would be curious to understand the actual benefits in terms of gameplay would such implementation have ?

There is never a dull moment; keeping up with station-keeping will be quite the challenge.

[samstarman5]

I would love when the time comes when a regular CPU can calculate N-body physics to a degree resembling them as we understand them in real life, but it is just not today.

Of course, about the same time will come the AI capable of handling every vessel you have on mission to keep it in orbit. But, even with that, you will need to keep an eye on them to make sure they have the delta-V to allow for station-keeping. The ISS doesn't maintain its orbit with just the happy thoughts of the world, after all.

[Eric S]

Oh, and one more thing. There's no arguing that n-body physics would simulate reality better. However, patched conics approximations are good enough for quite a few things even in reality. In fact, the Apollo program was planned out using patched conics. Basically, unless you're spending much time away from the gravity wells, patched conics works fine.

[bakanando]

What your quote says is basically: we are making it very precise, but at the same time there's so many calculations that your pc might just spontaneously burst in flames. He wants to make it faster without sacrificing accuracy, but hasn't finished applying it.

Also floating point errors already suck in-game, going n-body would make it even more noticeable.

[Fel]

I would love when the time comes when a regular CPU can calculate N-body physics to a degree resembling them as we understand them in real life, but it is just not today.

Of course, about the same time will come the AI capable of handling every vessel you have on mission to keep it in orbit. But, even with that, you will need to keep an eye on them to make sure they have the delta-V to allow for station-keeping. The ISS doesn't maintain its orbit with just the happy thoughts of the world, after all.

*Spewing nonsense that is somewhat true*

Well... CPUs are absolutely horrible when it comes to doing anything math related. 64-bit is a pointless "upgrade" as we do have the technology to go much higher... much much higher... and get crazy benefits from SSE code.

Of course, the CPU's main purpose is running small programs and keeping quiet. Number crunching? Nahh... just enough to keep the system running. Now it is the GPU that does that, and does it well...

http://arstechnica.com/security/2012/12/25-gpu-cluster-cracks-every-standard-windows-password-in-6-hours/

You can use a modern CPU, just streamline 25 GPU's and you can then calculate real time N-Body Physics (maybe) ;p

[Linear]

*Spewing nonsense that is somewhat true*

Well... CPUs are absolutely horrible when it comes to doing anything math related. 64-bit is a pointless "upgrade" as we do have the technology to go much higher... much much higher... and get crazy benefits from SSE code.

Of course, the CPU's main purpose is running small programs and keeping quiet. Number crunching? Nahh... just enough to keep the system running. Now it is the GPU that does that, and does it well...

http://cdn.arstechnica.net/wp-content/uploads/2012/12/radeon_city.jpg

http://arstechnica.com/security/2012/12/25-gpu-cluster-cracks-every-standard-windows-password-in-6-hours/

You can use a modern CPU, just streamline 25 GPU's and you can then calculate real time N-Body Physics (maybe) ;p

Why does everyone think nbody is so intensive?!?!

[Joonatan1998]

Why does everyone think nbody is so intensive?!?!

It itself isn't that bad, but at 100000x timewarp, and orbit predictions...

[Superfluous J]

Why does everyone think nbody is so intensive?!?!

Because it is. Unless you do it improperly in which case you may as well do patched conics.

The total number of interactions is X*(X-1). In the Kerbolar system there are 17 bodies, so just to figure out how gravity affects the PLANETS (before you've even launched a single ship) it'd take 272 calculations. Right now, with the planets on rails using patched conics, the total number of calculations is 17.

But really for a modern computer that's not the problem. The problem is it'd be no fun and the planets would eventually pull each other out of their orbits.