The war against lag: Anti-lag fairings

[woppeldopple]

Please leave the thread if you're not willing to discuss things properly. You've gone from discussing to "ROFL keep dreaming ^^" to calling someone who clearly knows a lot more about physics than you incompetent simply because you do not understand or agree with the matter at hand.

The physics is correct, and from a programmer's perspective the engine should be able to work this way.

If you are really certain that this is impossible, then the game developer who knows his way around the physics engine will be sure to prove me wrong wouldn't he?

Actually, if someone knows who that is, please invite him to this thread. I'd like to hear his opinion on this.

Hehe.

- Except center of mass (extremely trivial, i am not willing to argue about too) is the only thing that would be used in a phsyics engine.

- You were fixaded in the isolated word "inertia" i mentioned as prove how easily it would be implemented. This leads to severeal problems:

- Firstly i feel responsible to reply on this

- Secondly it shows you don't know what inertia is and why a point mass is not the same. (the formula will not help here)

- Your Chicaco friend posted an anonymous email adress and tries to disguise as someone he'd like to be (but cant afford a better pc). But he looks through the problem neither.

summa sumarum: youre arguing about the (wrong) implementation of a point mass simulation completely ignoring Squad will never create a physics engine from scratch. Everybody concerned with professional software development knows that even a simple 2D engine comes with huge effort. It is much more than 2 formulas (lets pretend the ones you posted were approriate). Why this effort? Because you strongly believe a point mass is the same as your 1000 parts construct? Sorry i have to LOL about this, too. May i ask you stop talking things like "from programmers point of view ... " - you dont know this view!

Reading further the arguments even go hair-rising....

Still all the contra arguments were completely ignored and are not understood by you and the professor in spe.

Quite rediculous - i didnt want to response but you guys started abusing "inertia" i mentioned.

You can take this as critique for your (sadly quite typical for KSP community) style of "discussion".

This is no discussion its a fight for wrong arguments.

Edited August 1, 2013 by woppeldopple

[Mr. Scruffy]

I just realized: There is neither a ´report post´ nor a ´ignore user´ button on this forum.

[nhnifong]

IIRC the Fairing Factory generator website that includes a DLL that struts anything attached to the fairing base. Unfortunately I never use FF parts since the fairings and base themselves are quite heavy.

Actually it just struts opposite fairing pairs to eachother.

[sjwt]

My personal favorite fairing is this one.. See how it neatly encapsuletes everything? .. OH I KILL ME!

I suggest everyone installs it with a dash of 'Staying nicely on topic'..

If symptoms persist, I'll see you in the morning!

[KerbMav]

I just realized: There is neither a ´report post´ nor a ´ignore user´ button on this forum.

Actually there is, but I think as adults we can handle this or leave the "room" in disgust if we so choose.

I consider the majority of the wobbling to be unintended, although I don't think it actually counts as a "bug" per se, more as a too-low spring constant between the joints.

However, this doesn't address the strength problem at all; a sufficient jerk is enough to break a payload apart and that can be produced during launch, during MECO, due to poor piloting, etc. It doesn't matter if the joints are all perfectly rigid; a payload should not be able to take an infinite amount of force without breaking, but this current proposal allows that to happen, so long as the fairings are attached.

If the "spring constant" (is that the correct term by the way?) is bound to the joining-force (just made that word up) in the Unity engine, then the wobble could only be suppressed if the crafts get magically more stable. But I dont know about these things, I am just guessing and letting my mind wander here.

To give the proposal a fair amount of thought: What happens if a second ship enters physics range? is it fully calculated or ever only the active ship? Is the lag that bad if the parts of the second ship start getting computed? Will have to make a rendesvouz today to see for myself.

But: Making Fairings into subspace fields suspending all but the sum of the mass of the parts contained therein - no. No for the reasons already called out here. Structural integrity would be 200% always and no approximated inertia would produce results that could be compared to the damage a craft/payload would take right now.

If a powerful engine had only to deal with the combined mass of the payload-parts the mass would still be really high and the connection between lifter and fairinged payload would most likely break - stopping this from happening would kill the prospect of having to design a sound rocket in this game and we could just expand the MS Space Flight Simulator instead.

What I could imagine, would be a difficulty setting suspending physics altogether - why not go all the way if big contraptions is the only goal for a player?

I for one do want to feel the pain and the relief if something weird is finally in orbit.

[imasquare]

While sending up another huge floppy space station component, flailing around in <1 FPS on top of more Mainsails than anyone is supposed to use, it hit me:

We don't have to render and emulate physics of parts inside fairings.

Now of course we all love the Procedural Fairings mod, something that should become stock immediately in my opinion.

This idea would expand on that, making them more useful and providing a real solution to a very difficult problem (game performance).

How it works:

Parts inside a fairing are not rendered (optional), nor physics emulated. (Perhaps they don't even exist at all, until spawned when the fairing opens.)

The weight of the parts is calculated into the fairing base. Of course the center of mass shifts accordingly. (Don't forget angular inertia as well.)

Because of this, the fairing is a perfect substitute for the parts, it physically represents all of them - in one single physics entity.

Effects of doing this:

-A 1000 part "thing" can be sent up by a huge lifter without any lag. (Now we can launch anything.)

-The fairing "protects" the part from structural failure as a result of wobbling or G-forces. (For realism, you can think of the parts being structurally attached to and supported by the fairing.)

-Less flailing, wobbling, wiggling and whatnot that makes attitude control a PITA. (We can actually steer.)

-There is a good reason to use fairings other than aesthetics. (Because pretty.)

And I will repeat;

Now we can launch anything.

I agree with the principle of what this suggestion is trying to achieve, and that the Procedural Fairings mod needs to be made stock.

There is a great debate over the physics and wobbling of the game, and the inherent issues with the physics engine and realism etc. etc. and I side on that of simplifying things for balance and stability of medium and large payloads.

After many frustrating failed launches of large or complex loads, I pondered briefly on how real life rockets can be much larger, deliver delicate, intricate and expensive payloads without anywhere near as much issues.

Basically we're given lego blocks to try to put together a high powered launch vehicle, simulated using single points of physics between parts on a single cpu core - and it just doesn't work properly without putting struts all over it like piano wire spider webs and even then it just isn't fun any more.

I would rather treat the object(s) inside the fairing with simplified physics, because it's not enjoyable to try and do all sorts of re-enforcements with the given tools in KSP.

I can't re-create a useable version of the 2800 ton Saturn V rocket in KSP because I am constrained by the 'KSP Laws of PhysX' which are terrible - not the laws of real physics. Frankly I would rather see rigid whole body physics for each whole ship in KSP until such time that the ship actually impacts with the ground, another craft or a celestial body.

I understand people can see this as rewarding 'bad rocket/aircraft design'. Again, bad design using simplistic Lego pieces? ...actually you know what? I think lego rocket/aircraft would be more stable because you'd have at least 4 to 8 points of interconnect between components for PhysX to calculate.

If a load is too complex or delicate, placing it inside the fairing should be like saying 'too complex/impossible to make launchable with KSP tools'. KSP is an arcadish simulator with a simplistic physics engine only capable of supporting relatively simple space craft without strutting it all over the place - but customers yearn for large, complex and sometimes absurdly powerful rockets.

Enough of the physics point of view. Let's look at it from a commercial/consumer point of view - because that's what matters most.

I am a customer, I have purchased this game, I enjoy this game - but I want to launch bigger and better rockets and payloads because I enjoy doing so. I want big, complex and intricate spacecraft with plenty of lights and moving parts, why? It's fun! But I can not enjoy the game that way because the PhysX engine is not fit for this purpose.

On top of enjoying large and intricate spacecraft, I want reasonable frame rates; to help achieve higher frame rates the developers could load in the model for the payload inside the fairing gradually through the ascent or just before fairing jettison or in whatever what is deemed optimal.

I have to admit if this game were developed by a larger software house or overseen by a large brand name publisher they would probably put a restriction on both spacecraft mass and number of parts to prevent performance and gameplay experience issues.

Then again if they had it were being developed by a larger software house they'd probably look at multithreaded physics engine/GPU acceleration, or maybe clever ways to simplify the physics calculations simulated but still keep the simulated physics satisfactory.

So by implementing the ideas suggested by Physcix there is a good chance of increasing performance and gameplay experience of the customer. Better performance and gameplay experience should hopefully lead to more gamers recommending it, and hopefully a flow on in commercial return on sales.

[PolishRenegade]

From all the talks about physics and the lag problem, there is very little talk about how game engines work and the source of the lag.

Due to the nature of the internet, here's a bit of a background; I'm a Unity developer professionally, mostly tackling optimization issues with rendering and memory. Ensuring object culling from the GPU is as efficient as possible.

That said, Unity's physics engine is outdated while also being a heavily customized version of PhysX (part of the reason it's outdated; difficult to support).

From my experience with the game, in order to save on computing time, a lot of the parts (rigidbodies) can start 'sleeping'. This means that until some delta value is applied to them, no physic calculation is applied other than for the parent transform value changes (like pivot points, there is a hierarchy of objects made from the parts).

This raises the question on how much KSP is really a 'realistic simulation' and how much is 'emulation'. I believe, in order to optimize as much as possible, physics-wise, the latter is probably more true. In that perspective, I also believe that most of the discussions in this thread are moot. The OP's idea is definitely something that can be implemented and could be emulated correctly, just like how most of the computations are emulated now. Striving to go _100%_ realistic in KSP is asking to have a super-computer do the calculations with custom software. That's not even close to the realm of the possible.

On the other hand, rendering wise, 'showing' what's inside the fairings, is _very light_. It's the physics that makes this whole engine lag beyond any recognition. As you might guess, 'emulating' the physics inside the fairings, with 0 rigidbodies inside, is definitely a huge-boost in performance.

At the end, my personal opinion is that I prefer a lag-free experience. KSP's emulation is darn good enough and can be tweaked but I don't expect nor believe it can ever be a simulation simply due to the engine's limitations in that particular domain.

*Edit; The poster before also tackles most of those points. +1 to him.

Edited August 1, 2013 by PolishRenegade

[aNewHope]

Squad!.... JUST ADD THIS!

[Dunbaratu]

There is a possible problem with this that hasn't been addressed yet, which is how the payload would be affected by acceleration and jerk during the early launch stages.

With the current system, the payload has to be designed to survive the forces, moments and impulses created by the launch vehicle; it must be solid enough to not flex uncontrollably, it must not break under the stresses imposed.

I would consider this challenge of the game to be a lot more fair and realistic only if we were allowed to attach parts properly and not have to resort to a tree structure, which is NOT a realistic limitation, but is the primary reason that payloads' attachment points are weak and bendy during launch. The single point of attachment is not something you can cite as a point of realism in the game, and it IS the chief cause of this being such a challenge.

In a real rocket design, the payload isn't a load bearing part of the launch vehicle. The load is handled by the structure that surrounds the payload. In the Saturn 5, the lunar module wasn't supporting the weight of the capsule, pushing up against it during launch. It was attached underneath it, and being pulled up by the command module, not pushing against it. The structure AROUND the lunar lander was holding the weight. But that requires multiple attachment points of substantial material (not just struts). And because this game doesn't allow that, you get this purely artificial limitation that payloads have to be load bearing because they have to be part of the main trunk of the craft.

[ferram4]

I would consider this challenge of the game to be a lot more fair and realistic only if we were allowed to attach parts properly and not have to resort to a tree structure, which is NOT a realistic limitation, but is the primary reason that payloads' attachment points are weak and bendy during launch. The single point of attachment is not something you can cite as a point of realism in the game, and it IS the chief cause of this being such a challenge.

A single point of attachment isn't realistic, but sufficiently strong struts add multiple points of attachment for any payload. Besides that, the main difficulty isn't that there is only one attachment point, but that those points are too flexible, which is a different problem entirely.

In a real rocket design, the payload isn't a load bearing part of the launch vehicle. The load is handled by the structure that surrounds the payload. In the Saturn 5, the lunar module wasn't supporting the weight of the capsule, pushing up against it during launch. It was attached underneath it, and being pulled up by the command module, not pushing against it. The structure AROUND the lunar lander was holding the weight. But that requires multiple attachment points of substantial material (not just struts). And because this game doesn't allow that, you get this purely artificial limitation that payloads have to be load bearing because they have to be part of the main trunk of the craft.

While all of that is true, the Lunar Module still had to be able to handle the forces and impulses applied to it; it might not have had to hold up the command and service modules, but it did have to handle 4gs and had to withstand the sudden jerk from the S-IC engines cutting out. Psycix's proposal wouldn't even require it to withstand that. Or, to make the issue more obvious (and properly absurd), if you were to take a Lunar Module and fit one of these magic anti-physics fairings around it, and then placed that on top of instant-detonation explosives (or perhaps a ton of high-ejection force decouplers) that could create an instant 100g force, the Lunar Module would survive. Removing the fairings would prevent the Lunar Module from surviving. Even if there were millions of attachment points between the "launcher" and the payload the payload should still be crushed; Psycix's idea prevents that from occurring, which makes me very wary of it.

[TimMartland]

Well, I think this is a splendid idea. It doesn't cause any realism issues (The fairing would act like the payload was secure inside, a realistic, strutmageddon-free and aesthetic way to fly) and would help people with less powerful computers. I don't have lag issues, but I have to play with lower graphics than I would like due to PhysX being wobbly and laggy.

[tntristan12]

Realistically, how many times have you had a rocket fall apart because of G-forces being placed on it? If you apply enough struts then it will always hold steady, so why can't we assume that any payload inside the fairing is strutted sufficiently that it won't break? You could include the weight in the weight of the fairing, etc.

[Dunbaratu]

A single point of attachment isn't realistic, but sufficiently strong struts add multiple points of attachment for any payload. Besides that, the main difficulty isn't that there is only one attachment point, but that those points are too flexible, which is a different problem entirely.

But those struts are the reason for the physics lag. You have to use 10-20 parts to simulate the strength of of 2 or 3 more substantial parts, because the substantial parts can't be connected in a "loop" like struts can.

While all of that is true, the Lunar Module still had to be able to handle the forces and impulses applied to it;

The point is that it ONLY had to have a strong enough connection to support its OWN rather light mass through those forces, rather than the mass of itself plus the entire command module above it being crushed into it, as it would in KSP. In KSP, the rocket would push the lander up into the command module and the lander's connections have to be strong enough to handle that. In real life, the rocket pushes *around* the lander through the outer walls to the command module and then the lander "hangs" off the command module. It's that "around the lander" bit that KSP's tree structure doesn't support without extremely laggy physics of lots and lots of struts.

My point being this: don't forget how much of the challenge is an artificial artifact of the simulation rather than being a real-world problem. In the real-world, your payload does not need to bear the weight of what's above it because you can attach substantial structure around it attached in more than one location. In KSP it does need to be load-bearing unless it's the topmost thing exposed to the air (in which case the complaints about it being unrealistic to replace it with one part should go away since it's not going to break anyway in that scenario.) The only place where its load-bearing ability matters is in precisely that place where the challenge of building a load-bearing payload is an artifact of the simulation's limits anyway.

[imasquare]

...My point being this: don't forget how much of the challenge is an artificial artifact of the simulation rather than being a real-world problem. ...

That's the crux of it, well said. It seems that some people want super realistic simulations right down to the scale of atom interactions requiring a football stadium sized super-compute-cluster, whilst others just want...an approximate emulation that their humble PC can play. I'm not sure what Squad wants to deliver.

[SyberSmoke]

I like the idea that the fairing would treat the object as just a mass while removing the physics calculations for each spring/attachment. This though is more of a work around for a fundamental issue with the engine. That being objects in the game only have one connection point. While this is good for putting things together (in most cases at least), it is not so good for calculating the stresses on a physical structure that would be "knitted" together.

The wobble is caused by these single connections, something that (by my observations at least) the struts were put in to fix by adding additional physics connections and linkages. But part of the problem could also be fixed by adding more discrete connection points. Points that treat connections between tanks as a rigid weld making them a solid single body. While others that simulate flex between smaller more fragile structures. These structures would most likely need struts to compensate.

But for the interim, if implemented properly and drag was properly balanced out (*Cough* Nosecones *cough*) then this would be a good way to make assumptions about the cargo with out making the player place 100 struts in the process.

As for loading the objects...it is remarkable how little people know about games some days. The objects would not need to be unloaded from memory and could in fact remain as a static mesh with textures in the fairing. The object would remain in the fairing and would not be rendered by the GPU, and really it should not be rendered by the GPU if it is not visible any way. If it is being rendered...that is just sloppy programming.

When the Fairing is popped, the object will be there and render as normal. The engine should also have time, even if a delay of a couple frames is added, to reintegrate the constructs physics into the system with out an issue. Especially since the fairing will be deployed in the relative calm of space and not during the launch window.

Really this is all game technology 101.

[sumghai]

My cursory understanding is that KSP/Unity allows the definition of various joint types, such as MERGED_PHYSICS and NO_PHYSICS.

Would it be possible to, say, load the payload under the fairing as normal, temporarily change all the connections between the individual components to MERGED_PHYSICS so that they would be treated as a single mass, and then re-enabling the original joint types when the fairings are jettisoned?

[pizzaoverhead]

As far as the delay on resuming physics for the payload is concerned, we have an example of this already: Exiting timewarp.

Assuming a payload is fully strutted to the fairing base, is there any difference in simulation with a surrogate single-part or welded-part payload? The weak-points become the fairing covers and the attachment to the fairing base, which would be susceptible to acceleration bumps from launch and MECO. If any of these break, the payload becomes simulated again.

I would like to see fairings which are weak to horizontal aerodynamic stresses:

[nhnifong]

Aerodynamic refactor is the new Planets.

[Holo]

Maybe the problem is poor payload design.

[ferram4]

The point is that it ONLY had to have a strong enough connection to support its OWN rather light mass through those forces, rather than the mass of itself plus the entire command module above it being crushed into it, as it would in KSP.

This is what I'm talking about as well, however, based on what I've seen presented here, this idea would allow that payload to take an infinite amount of load. How would it handle when a player creates a monstrosity with a TWR of 20? Will it crush under its own weight, (with nothing above it, only its own mass) or will it manage to survive regardless of how flimsy it happens to be?

Further, with this idea, what benefit is there to keeping any part of the rocket outside of a fairing? Wouldn't it make the most sense for my rocket to consist of successive fairing layers, like a Matryoshka doll? Even if there is some manner of structural failure checking, the removal of wobbling inside the fairing will allow flimsy rockets to launch that would tear themselves apart if they weren't encased in a fairing.

The problem is not perfect simulation vs. approximation; the problem is (what I perceive to be) inconsistent physical laws inside and outside the fairing. Inconsistency in a game is a recipe for player confusion and frustration ("Why would that even matter?" "It doesn't make sense that it would break if there wasn't a fairing there; everything else was the same!").

[Dunbaratu]

This is what I'm talking about as well, however, based on what I've seen presented here, this idea would allow that payload to take an infinite amount of load. How would it handle when a player creates a monstrosity with a TWR of 20? Will it crush under its own weight, (with nothing above it, only its own mass) or will it manage to survive regardless of how flimsy it happens to be?

Further, with this idea, what benefit is there to keeping any part of the rocket outside of a fairing? Wouldn't it make the most sense for my rocket to consist of successive fairing layers, like a Matryoshka doll? Even if there is some manner of structural failure checking, the removal of wobbling inside the fairing will allow flimsy rockets to launch that would tear themselves apart if they weren't encased in a fairing.

The problem is not perfect simulation vs. approximation; the problem is (what I perceive to be) inconsistent physical laws inside and outside the fairing. Inconsistency in a game is a recipe for player confusion and frustration ("Why would that even matter?" "It doesn't make sense that it would break if there wasn't a fairing there; everything else was the same!").

But leaving it as it is isn't a good solution either, for the reasons I mentioned. Trying to design a payload that's connected in such a way that not only do the connectors take the stress of holding onto the payload but they also have to take the stress of the crushing mass of everything else above the payload too is a purely artificial challenge caused entirely by KSP's imposition of a tree-structure for everything that isn't a strut. Remember that this was about trying to get rid of FPS lag, and that FPS lag is being caused BY the massive amount of strutting going on, which is being done purely for this artificial reason that you can't break the tree structure rule with anything else other than struts.

[Psycix]

This is what I'm talking about as well, however, based on what I've seen presented here, this idea would allow that payload to take an infinite amount of load. How would it handle when a player creates a monstrosity with a TWR of 20? Will it crush under its own weight, (with nothing above it, only its own mass) or will it manage to survive regardless of how flimsy it happens to be?

The fairing itself will break, causing physics to enable, causing catastrophic failure the way you are used to.

Further, with this idea, what benefit is there to keeping any part of the rocket outside of a fairing? Wouldn't it make the most sense for my rocket to consist of successive fairing layers, like a Matryoshka doll? Even if there is some manner of structural failure checking, the removal of wobbling inside the fairing will allow flimsy rockets to launch that would tear themselves apart if they weren't encased in a fairing.

The weight of the fairing prevents you to Matryoshka doll everything. Every fairing will cause you to lose a certain percentage of your payload fraction.

Perhaps fairings could scale their weight based on what is in it. Heavy payload - heavy fairing as to support this payload properly.

The problem is not perfect simulation vs. approximation; the problem is (what I perceive to be) inconsistent physical laws inside and outside the fairing. Inconsistency in a game is a recipe for player confusion and frustration ("Why would that even matter?" "It doesn't make sense that it would break if there wasn't a fairing there; everything else was the same!").

So what if we have flimsy fairings that are more easily broken than strutpocalypse crafts?

This way you can't use it as a "cheat". The old fashioned way is potentially sturdier - as long as your PC likes to simulate hundreds of parts supported by hundreds of struts.

ANY flimsy payload can be launched conventionally by adding perpendicular beams on the craft that connect struts to everything, turning the ship in one giant girder with a rocket on the bottom. Not exactly realistic, fun to fly or fun to build.

Wouldn't fairings be a better* alternative than having to do stuff like this:

http://i.imgur.com/5phhGoQ.jpg

Picture from google images.

* Better in the sense of:

-Better game performance

-Better realism

-Better allocation of time: building and flying rockets rather than having to go through endless trial-and-strutting cycles.

Edited August 17, 2013 by Psycix

[SyberSmoke]

Phycix, if Procedural Fairings is still current and in development, why not talk to the person developing it to see if a "proof of concept" can be made. I think some of the ideas need a little tuning, like I do not think the fairing base should be the point of mass, but instead the physics skeleton should remain to provide accurate mass calculations. That skeleton is just made inert when it comes to calculating flex and stretch at the joints. This would be the most reasonable alteration to prevent obscene payloads.

As a thought also, given the weighted nature of the fairing as described. It may be an idea to have it also have built in counterweights. This would give it mass but also allow less balanced payloads to be balanced out by the fairing weight. There should be limits of coarse, but it is a thought.

[Technical Ben]

But then it would have to load it once the fairings were separated, and that could take a big chunk of time, depending on the amount of parts under there.

Perfect solution: Load a "Super duper launch camera view" so that on load, the game pauses, zooms in, and the fairings puff, open and the part floats out dramatically. Basically, if it's not a bug, make it a feature!

[ferram4]

The fairing itself will break, causing physics to enable, causing catastrophic failure the way you are used to.

Fine, that works then. But what prevents someone from strutting the fairing to hell to prevent it from failing and physics suddenly being enabled on the payload?

The weight of the fairing prevents you to Matryoshka doll everything. Every fairing will cause you to lose a certain percentage of your payload fraction.

Perhaps fairings could scale their weight based on what is in it. Heavy payload - heavy fairing as to support this payload properly.

Why would two payloads, with the same sizes, have different masses? And if we did this, if there are fuel tanks in there do we base the fairing mass on the full mass or empty mass?

So what if we have flimsy fairings that are more easily broken than strutpocalypse crafts?

This way you can't use it as a "cheat". The old fashioned way is potentially sturdier - as long as your PC likes to simulate hundreds of parts supported by hundreds of struts.

ANY flimsy payload can be launched conventionally by adding perpendicular beams on the craft that connect struts to everything, turning the ship in one giant girder with a rocket on the bottom. Not exactly realistic, fun to fly or fun to build.

Wouldn't fairings be a better* alternative than having to do stuff like this:

http://i.imgur.com/5phhGoQ.jpg

Either the fairing would allow you to launch that flimsy payload without the huge girder mess, at which my "inconsistency" complaint still stands, or it would break, at which point we're back to where we are now.