n-body physics? We don't even have 2-body physics!

[Javster]

Anyone could mod a Jumbo Fuel tank to have the same mass as Gilly and then HyperEdit it into Gilly orbit.

Or just a probe core (that would be cool)

[Camacha]

Or, just maybe, everyone who disagrees with you isn't completely ignorant, maybe we just disagree with you on where the "big deal" line should be drawn in this case. De gustibus non est disputandum, you know?

It seems that most people do not know what n-body physics would bring, to be honest

I just realised that Universe Sandbox actually uses Unity (although the new version will not), so they deal with the same limitations as KSP does.

[samiamthelaw]

You could roughly compare it to that. Computational expensive should not be an issue, as games like Universe Sandbox do this on a huge scale with great succes. You should also not suffer from all the weird effects physics warp causes.

I'm pretty sure the weirdness of physics warp comes from the need to simulate intra-ship physics effects, not from the computation of gravitational forces. N-body gravity versus 1/2-body gravity doesn't matter there because they're both computationally free next to the work needed to simulate a large ship. Normal warp just suspends intra-ship physics simulation, but KSP requires that the acceleration on the ship be zero before you turn it on. This is basically a "balance" sort of design choice, as with zero acceleration we can safely assume that nothing should tear the ship apart.

[*Aqua*]

You're completely, utterly wrong. What autopilot is needed for is precise, long-term stationkeeping. Barring edge cases, orbits in N-body simulation are not chaotic, but they do slightly change over time. Unless you screw up, the orbit won't go out of control. Also, autopilot would be required for maintaining unstable, edge-case orbits (which you might want for whatever reason), but nobody says you have to use those.

Lets say I have a small sat with a kethane scanner and want to map Kerbin's deposits. I put it in a 75x75 km orbit and timewarp as fast as the kethane plugin can keep up. With the rails system this works. With n-body-physics it won't work. After 10 mins (RL time) or so Mun's influence will change the orbit to a like 69x80 km orbit.

Other example: I parked a Jool probe in Kerbin's orbit and timewarp to the transfer window (which can take up to a Kerbin year). Same result as above.

The Kerbol system won't work with these physics. The celestials have way to much mass. And if you remove the mass the whole system will break apart.

Clearly, you people don't know just how big deal N-body (or more accurately, a physics model based on integrators) is. It's not "only Lagrange points". There are sun-synchronous orbits, orbital decay, realistic planet motion, thrusting during timewarp, tidal forces, binary systems... With this kind of simulation, forces can be applied during timewarp just fine. It also fixes inaccuracies when crossing the edge of a SOI (currently one of Kraken's nests) and gives you a lot more options than with the current simulation.

I don't think it should be toggleable, either. It's just too big of a deal. Disable it, and say goodbye to ion engines, tidal stabilization, sun synchronous orbits, solar sails... The possibilities are so enormous that you wouldn't be able to stop using this once it's added.

And how is a normal player supposed to handle that? This needs ultra precise control of a craft. Say goodbye to the nav ball. That thing isn't accurate enough. Is the craft too light or too heavy? Sorry you just missed the orbit because of that.

No kraken is a fine thing. Lagrange points, sun-synchrounous orbits etc., too, I can't deny that. But the other stuff will make the game unplayable.

[Arsonik]

I came into this thread with the perspective that the current system is fine. But if everything Dragon said is indeed true.... Well, then I can say I agree with him. It seems like some people are afraid to admit they were wrong and instead want to debate the semantics of things like "where the big line is drawn". All of the stuff that would "trouble new players" is a joke. Most people launch a ship, don't finish one orbit, and then blast off at the target. Station keepers will have challenges to keep up with, that's about it best I can see. Other then that there are the claimed new game play options it brings as well as resolving several technical shortcomings of the game and fixing bugs. It's more then just Lagrange points.

[Camacha]

I'm pretty sure the weirdness of physics warp comes from the need to simulate intra-ship physics effects, not from the computation of gravitational forces. N-body gravity versus 1/2-body gravity doesn't matter there because they're both computationally free next to the work needed to simulate a large ship. Normal warp just suspends intra-ship physics simulation, but KSP requires that the acceleration on the ship be zero before you turn it on. This is basically a "balance" sort of design choice, as with zero acceleration we can safely assume that nothing should tear the ship apart.

Currently, the whole thing is cobbled together in a rather awkward fashion. I mean, 4x physics warp, having to switch to 0x to switch to rails warp which then has a different scale... Really? Not to mention not being able to save in flight, the physics going crazy when under tension and warping and whatnot. From a user experience point of view the system is a disaster. KSP players learned to work with it, but looking at it from a neutral perspective and from a distance, it is really a kludge. I would much prefer a system that does not have all the weird artificial boundaries and scales seamlessly across the whole board. That it is also closer to real life is only a benefit.

And how is a normal player supposed to handle that? This needs ultra precise control of a craft. Say goodbye to the nav ball. That thing isn't accurate enough. Is the craft too light or too heavy? Sorry you just missed the orbit because of that.

Where do you get all those wild ideas? Like it was said before, orbits might change slightly in some cases and barely in other cases. It is not as if you randomly shift through the universe. They used the navball in real life, I really do think it will suffice for KSP.

Edited March 12, 2014 by Camacha

[*Aqua*]

Like it was said before, orbits might change slightly in some cases and barely in other cases.

Prove it a LKO can be stable for at least 10 Kerbal years. No mechjeb allowed.

They used the navball in real life

They use computers in real life.

Edited March 12, 2014 by *Aqua*

[samiamthelaw]

Currently, the whole thing is cobbled together in a rather awkward fashion. I mean, 4x physics warp, having to switch to 0x to switch to rails warp which then has a different scale... Really? Not to mention not being able to save in flight, the physics going crazy when under tension and warping and whatnot. From a user experience point of view the system is a disaster. KSP players learned to work with it, but looking at it from a neutral perspective and from a distance, it is really a kludge. I would much prefer a system that does not have all the weird artificial boundaries and scales seamlessly across the whole board. That it is also closer to real life is only a benefit.

The UI system for setting time warp levels isn't relevant to whether N-body versus 1-body physics would allow KSP to simulate thrust under time warp conditions. I don't dispute that changing to N-body simulation would permit possibilities that don't work now, but some of the examples you specified (thrust under time warp, ion engines, and solar sails) are not among the possibilities N-body simulation would enable.

[Camacha]

The UI system for setting time warp levels isn't relevant to whether N-body versus 1-body physics would allow KSP to simulate thrust under time warp conditions.

It is the other way around - because Squad does not really know how to deal with physics warp versus on-rails warp, they cobbled up this weird method, leading to the strange UI system and unintuitive limitations.

I don't dispute that changing to N-body simulation would permit possibilities that don't work now, but some of the examples you specified (thrust under time warp, ion engines, and solar sails) are not among the possibilities N-body simulation would enable.

I could well be misunderstanding something as I am no expert, but I think those should all be possible. Since you are not sliding along some fixed rail, computations will influence each other, meaning you can apply forces like thrust under warp, and thus ion engines and solar sails (which basically is low thrust for a long time).

Maybe someone with a little more background regarding the integration could fill us in, but I think it should be possible to apply thrust along with all other relevant forces. Without that option, the whole system would be moot, as it would still be an on-rails system.

[Lack]

@Camacha,

Lets say I have a small sat with a kethane scanner and want to map Kerbin's deposits. I put it in a 75x75 km orbit and timewarp as fast as the kethane plugin can keep up. With the rails system this works. With n-body-physics it won't work. After 10 mins (RL time) or so Mun's influence will change the orbit to a like 69x80 km orbit.

He's not wrong. After 10 min Honeyfox's n-body integrator has my orbit move from 75kx75k to about 53kx100k.

[Camacha]

He's not wrong. After 10 min Honeyfox's n-body integrator has my orbit move from 75kx75k to about 53kx100k.

I am sure it does that, I am just not sure where that comes from And again, doing it right probably means more than just implementing purely the physics.

Edited March 12, 2014 by Camacha

[Lack]

I am sure it does that, I am just not sure where that comes from

I suspect it's from the close proximity of the Mun from Kerbin (11.4Mm compared to ~384Mm with the Earth-Moon system). The low radius compared to gravitational strength probably plays a part as well. I'll have to try in with RSS sometime.

Took 11 hours, 23ish min before I smashed into the Mun.

And again, doing it right probably means more than just implementing purely the physics.

Almost certainly, they'd probably have to smudge the n-body calculations or alter the orbits.

Edited March 12, 2014 by Lack

[Whirligig Girl]

Harvester talked on UNITE 13 about how and why KSP works how it does.

Enjoy!

Actually, I have seen that video, it's one of my favorites on the entirety of YouTube.

[cantab]

You need to do this both during timewarp and out of it, as in that case, integrators handle everything, including planetary motion.

It's been shown that the KSP system is stable. So to put the planets and moons "off rails", the system would have to be revised - and I'm not sure by how much.

The orbital drift others have described is probably because in stock KSP Kerbin does not orbit the Kerbin-Mun barycentre, whereas with the n-body mod your ship will. (Assuming the modder didn't fix this already). Similarly Kerbol doesn't orbit the Kerbol-Jool one. That can be fixed without needing to take the celestials off their rails, though the rails may become much more complicated.

[Lack]

It's been shown that the KSP system is stable. So to put the planets and moons "off rails", the system would have to be revised - and I'm not sure by how much.

The orbital drift others have described is probably because in stock KSP Kerbin does not orbit the Kerbin-Mun barycentre, whereas with the n-body mod your ship will. (Assuming the modder didn't fix this already). Similarly Kerbol doesn't orbit the Kerbol-Jool one. That can be fixed without needing to take the celestials off their rails, though the rails may become much more complicated.

Link to the KSP system under n-body conditions for anyone interested.

There's a few problems with the Jool system (Vall gets ejected pretty quickly). Don't think this one models it from Barycentres either, though. Does anyone know if anyone has calculated it using the correct barycentres?

Edited March 12, 2014 by Lack

[Dodgey]

Am I the only one who looks at this like the FAR of the solar system? I mean it's a feature that really should of been implemented from the start (I don't think people would complain if it was there from the beginning, we would be used to it). And one of the main arguments against it is that it would be too complicated for players, I say tough, it is just another step in the learning curve.

To say that ksp is a pick up and play game also. I just don't understand where this comes from, I just don't understand how you come to this conclusion.

Obviously due to the diminutive size of the ksp system optimization will have to be done but in no way is this a reason not to implement it into the game.

[m4v]

I am sure it does that, I am just not sure where that comes from And again, doing it right probably means more than just implementing purely the physics.

So how should it be done right? because I see a lot of players here bashing Squad for not being "smart" enough for implement n-body physics yet everyone only talks about the physics while skirting the gameplay implications.

Currently, the whole thing is cobbled together in a rather awkward fashion. I mean, 4x physics warp, having to switch to 0x to switch to rails warp which then has a different scale... Really? Not to mention not being able to save in flight, the physics going crazy when under tension and warping and whatnot. From a user experience point of view the system is a disaster. KSP players learned to work with it, but looking at it from a neutral perspective and from a distance, it is really a kludge. I would much prefer a system that does not have all the weird artificial boundaries and scales seamlessly across the whole board. That it is also closer to real life is only a benefit.

n-body isn't going to solve that, you can't thrust while in time warp because with higher time steps you lose accuracy, generating phantom forces and breaking your ship apart. You can probably figure out a solution, but you don't need to implement n-body for it, they are two unrelated problems.

[Xaiier]

Sigh, this topic comes up every so often and everyone gets all huffy about it.

First off, this is designed as a game, not a simulator! Squad is set on their vision and they will make it as they like, and there is 0% chance they will rewrite the entire physics system this late in the game. Sorry! If that bugs you, you are free to pursue more accurate simulations! But as we have already seen with resources, Squad will do their thing as they see fit, in order to create the game they want to!

Second off, the benefits and detriments of a n-body system have been discussed to death. Proponents talk about all the advanced maneuvers and realistic effects, opponents talk about how it would make everything too complicated. This typically ends in the proponents bashing the opponents for not being able to handle the increased difficulty and everyone gets in a big argument.

So in the end, it doesn't matter what you think. KSP is not a simulator. It will not become one. None of this will ever change. Arguing is pointless. Let those who want it investigate the mods that allow for it do so, and those who don't care for it continue with their current system.

If it's things like on-rails thrust that you want, those have already been solved in the current system, in fact I'll be adding that feature to my mod soon.

[Vanamonde]

Folks, N-body physics is on the what-not-to-suggest list because Squad has no plans to ever use it, and even if Squad ever should decide to convert to this model, it would be a major project and far, FAR down the priorities list. And since talking about it in the meantime tends to clutter up the forum endlessly and start arguments, it was decided to make this topic off-limit on the forum. The thread will be closed now.