N-body physics

[goncaloeaguiar]

So with the awesome announcement of the sequel of our favorite game a question has to be asked... Will N-body physics à-là Principia be implemented? 

What do you think.

[Maravone]

I seriously doubt it.

N-body physics is a somewhat different beast from the traditional KSP mechanics, and somewhat of an hurdle for any new players in what is already a though learning curve for anyone new to the game who has no clue about orbital mechanics.

So no, I dont think so, even though I would love to have it. It would be cool if we had an option, but thats something that I dont see the devs focusing on.

Either way, KSP is and I hope will always be a basic platform where the community can add, improve and customize to their hearts content. With that in mind, with even better mod support, KSP 2 could quickly get a Principia version of his own from the community relatively quickly I suppose.

[SlinkyMcman]

24 minutes ago, goncaloeaguiar said:

So with the awesome announcement of the sequel of our favorite game a question has to be asked... Will N-body physics à-là Principia be implemented? 

What do you think.

If they have N-body physics you would also need to have a way for crafts to maintain their orbits when not focused. Juggling relays could get real tedious real fast.

[Geonovast]

If it does, and it's forced, I won't play it.

It's still a game, which is something that I think is lost on a lot of people that want a perfect simulator.  That kind of management would be a nightmare for most people.  I'm fine with any feature that you can turn off, but I think the amount of people that would use it is small enough that it's not worth dev time.

But... but... realism! 

Yeah, realism.  NASA has more than one person working for it.

This stuff should really stay in mod territory.

[sturmhauke]

I'd be happy if they at least added axial tilt. From what I understand, the few mods that do it currently have some weird issues, because the engine doesn't really support it well.

[Citizen247]

51 minutes ago, Geonovast said:

It's still a game, which is something that I think is lost on a lot of people that want a perfect simulator.  That kind of management would be a nightmare for most people.  I'm fine with any feature that you can turn off, but I think the amount of people that would use it is small enough that it's not worth dev time.

If there's automated station keeping it might not be so bad. It would give a reason to decomission satellites and replace them with newer models at least.

Having said that, I doubt there'll be N-Body physics.

[Xurkitree]

of course not, that falls in modding territory only. But i'm not sure how well one can adapt Principia to work with multiple star systems.

[Spaceception]

Nah, that learning curve would be too steep. But I hope modding support makes it easier for those who want it.

[DStaal]

Even if they manage to optimize the physics quite a bit, KSP is always going to be a resource-intensive game if it can model multiple large ships.  N-Body physics would add tremendously to that complexity, as it's *not* a solved problem, and you would have to continually compute the influence of every ship and body in the universe with every other ship and body.  Patched Conics is enough for 90% of the time, and an O(n) process instead of an O(n!) or worse - well worth the optimization for a game.

(Note: I'm guessing on what N-Body would be in Big-O.)

[Abyliviorn]

I did like the challenge Principia offered. But I don't think it should be forced on everyone. The learning curve is already steep enough for most.

And since the game is redesigned from the ground up, offering deeper layers modders previously did not have access to, maybe a Principia 2 will come out eventually.

[TimmyFifty]

They could fix most of the issues by adding a part that spends RCS fuel to keep a craft in its current orbit around around the nearest, largest body. This would work even if the craft was not being piloted.

[TheOtherDave]

I voted as if the poll said “should” instead of “will”, seeing as how I have no idea what Star Theory will do.

[Geonovast]

34 minutes ago, TimmyFifty said:

They could fix most of the issues by adding a part that spends RCS fuel to keep a craft in its current orbit around around the nearest, largest body. This would work even if the craft was not being piloted.

I think the biggest issue is "Most players aren't going to want to deal with that nonsense."

 

What's the point of even having a feature if it gets completely automated in the background anyway?

[TheOtherDave]

Just now, Geonovast said:

I think the biggest issue is "Most players aren't going to want to deal with that nonsense."

 

What's the point of even having a feature if it gets completely automated in the background anyway?

Because it lets you do cool things with gravity assists?

[Hay]

It should be added, as an option to make the game more challenging. This way peep that don't want it can turn it off.

[rridgway]

12 hours ago, DStaal said:

Even if they manage to optimize the physics quite a bit, KSP is always going to be a resource-intensive game if it can model multiple large ships.  N-Body physics would add tremendously to that complexity, as it's *not* a solved problem, and you would have to continually compute the influence of every ship and body in the universe with every other ship and body.  Patched Conics is enough for 90% of the time, and an O(n) process instead of an O(n!) or worse - well worth the optimization for a game.

(Note: I'm guessing on what N-Body would be in Big-O.)

 

 

N-body physics is O(n^2), each body has (n-1) interactions to compute, so it would be O(n*n). That being said, I've found a paper which gives an algorithm for implementing N-body physics in something like O(n log n) or O(n).

http://www.cs.hut.fi/~ctl/NBody.pdf

[Frozen_Heart]

I hope not. Yeah it's technically more accurate, but it's not fun.

 

Having to constantly juggle your 200 varying missions to adjust their orbits constant would get old and boring fast. KSP is successful as it takes the basics of spaceflight and makes it accessible to the masses. N body physics goes against that.

[farmerben]

I'd be willing to play it, but I doubt my computer is up to it.  That many background calculations running all the time would probably overheat ordinary laptops and summon the Kraken.

 

That said a partial implementation would be kinda cool.  It might be possible to implement Lagrange points as places to put telescopes, and for trojan asteroids without worrying about instabilities everywhere else.

Edited August 20, 2019 by farmerben

[magnemoe]

 

58 minutes ago, farmerben said:

I'd be willing to play it, but I doubt my computer is up to it.  That many background calculations running all the time would probably overheat ordinary laptops and summon the Kraken.

 

That said a partial implementation would be kinda cool.  It might be possible to implement Lagrange points as places to put telescopes, and for trojan asteroids without worrying about instabilities everywhere else.

Main problem would probably be calculating the trajectory lines as you modify them. 
Pretty sure they would to plenty of shortcuts like don't use bodies with minimal impact and don't do then in low orbit there most stuff is put. 

[chaos_forge]

N-body gravitation is more computationally expensive than patched conics, yes, but it's not that computationally expensive, especially considering that ships in time warp are treated as point masses, and the gravitational effect ships have on other bodies is ignored. All in all, simulating n-body gravitation almost certainly takes fewer computations per second than, say, flying a 100 part ship through an atmosphere.  Further proof of this is the fact that, as far as I know, the Principia mod for KSP that adds n-body gravitation doesn't have a significant effect on performance.

As far as difficulty, AFAIK as long as you don't try to do any weird stuff it behaves pretty similarly to patched conics. The biggest problem I see is stuff moving out of its orbit over time, but an automated station-keeping feature would be enough to deal with that, which shouldn't be too hard to add. I'd even be okay with it if there was some sort of easy mode option where station-keeping doesn't consume delta-v.

All in all, I'd be pretty excited to see n-body gravitation in KSP 2. I'm not sure if it will happen, but there's a few features that make me think we might get it, such as the presence of multiple star systems, and the Rusk and Rask planets, which seem to be similarly sized and orbiting their common barycenter. Both of those scenarios are extremely difficult if not outright impossible to simulate using patched conics.

[Nuke]

would be nice but some work would be required to make the kerbol system stable for at least long enough to complete the game. if it starts to fall apart after that, well that's cool too.

[TheFrizz]

6 hours ago, chaos_forge said:

N-body gravitation is more computationally expensive than patched conics, yes, but it's not that computationally expensive, especially considering that ships in time warp are treated as point masses, and the gravitational effect ships have on other bodies is ignored. All in all, simulating n-body gravitation almost certainly takes fewer computations per second than, say, flying a 100 part ship through an atmosphere.  Further proof of this is the fact that, as far as I know, the Principia mod for KSP that adds n-body gravitation doesn't have a significant effect on performance.

As far as difficulty, AFAIK as long as you don't try to do any weird stuff it behaves pretty similarly to patched conics. The biggest problem I see is stuff moving out of its orbit over time, but an automated station-keeping feature would be enough to deal with that, which shouldn't be too hard to add. I'd even be okay with it if there was some sort of easy mode option where station-keeping doesn't consume delta-v.

All in all, I'd be pretty excited to see n-body gravitation in KSP 2. I'm not sure if it will happen, but there's a few features that make me think we might get it, such as the presence of multiple star systems, and the Rusk and Rask planets, which seem to be similarly sized and orbiting their common barycenter. Both of those scenarios are extremely difficult if not outright impossible to simulate using patched conics.

I assume you are proposing that the spacecraft's position is calculated using N-Body physics, but the planets themselves are not calculated using N-Body and use the traditional SOI simplification (with some exceptions). If the planets themselves also orbit with N-Body I think trajectory plotting on the map mode and time warp might be somewhat challenging. I agree that Rask and Rusk almost guarantee some amount of N-Body in the game. I have to imagine there will need to be some level of cheating though. Like the feature in KSP where as you mouse over your trajectory you see where your target body will be at the time you are at that point in your trajectory. This would be brutal on true N-Body because you can rapidly jump between massively different timestamps. Easier done if you can just cheat and lookup where the planet is at that time because it doesn't obey N-Body.

On second thought, maybe not that bad. As you calculate your trajectory you could store the locations of all the planets at every point calculated in addition to your craft's location. Given that there won't be millions of planets the RAM requirements wouldn't be too intense.

Edited August 20, 2019 by TheFrizz

[Optionalcoast]

I think n-body physics would add a nice progressive gameplay element. Your first orbital vehicles will only be able to stay up so long if you dont keep fixing their orbit but as you progress as a designer and maybe though researching new tech you can impliment atomated velocity adjustments that can keep your vehicle in its orbit. Like maybe timed burns or a rcs systme that tries to keep your apoapsis and periapsis constant. But this would probably work best as a difficulty option when creating a game because i understand not all player want the added complexity having your communication sat fall out of orbit after a few hundred passes with the mun.

I think it will also fun for advanced players to learn about how to fly spacecraft in and environment with multiple bodys influence and it would be more rewarding knowing that this behavior is very close what scientists and engineers deal with, with real spacecraft.

Edited August 20, 2019 by Optionalcoast

[chaos_forge]

1 hour ago, TheFrizz said:

I assume you are proposing that the spacecraft's position is calculated using N-Body physics, but the planets themselves are not calculated using N-Body

That's what I was thinking, yeah. I know Principia calculates the planets' orbits using n-body too, but I feel like having the planets be "on rails" is a reasonable middle ground between realism and performance.

Mostly, I just wanna be able to use Lagrange points and fancy gravity assists and stuff

[TBenz]

22 hours ago, Geonovast said:

What's the point of even having a feature if it gets completely automated in the background anyway?

N-Body physics entails a lot more than just unstable orbits. I'd love to be able to park a station at a lagrange point.

And, for what it's worth, requiring RCS to automatically stabilize the orbit still has relevant gameplay. Namely that it requires you to design stations and satellites with the facilities to stay in orbit, and either regularly supply them or plan around them having a limited functional lifespan.

I seriously doubt N-Body physics will happen, because it will add a lot of complexity to both the code and gameplay, but there could be some nice benefits nonetheless.

Edited August 21, 2019 by TBenz