Wobbly Rockets with David 'Trigger' Tregoning - KSP2 Dev Chat

[Superfluous J]

1 hour ago, cocoscacao said:

This game has been under development for a long time (officially). I fear that short-term solution may cause additional bugs, forcing dev team to waste time hunting and fixing issues that shouldn't be there in the first place

That's a very real problem, but it's one of those tradeoffs in Early Access.

Your game must jump from playable beta to playable beta, so your fans are happy but you continue to make progress.

You make progress at the expense of making fans happy, and make fans happy at the expense of progress.

It's a balancing act and whatever you choose people will be unhappy.

[Poppa Wheelie]

I'm walking down this train of thought (everything in quotes is actually me, paraphrasing - not actual quotes):

• I'm hearing that "we apply physics at each joint", and that "the physics that we apply to each joint is pretty accurate"
• There seems to be a fairly universal agreement that there should be no wobble in certain parts of the stack.  Maybe that's "no wobble per stage", or "no wobble per stack of same size per stage", or maybe it's something else.
• One of the short-term options the team is looking at would be to reduce the wobbly impact of physics applied to some of the joints.  But I understood Trigger's concern with this would be that this would actually incur a performance hit (that would not be scalable to larger vehicles) because the implementation of such a solution (which I'll call "auto-strut-like") would look like more joints to the physics engine (more connections at each joint looking like more joints, requiring more work from the physics engine).  Thus, the problem with this is that it would apply even more physics to each joint.

I think what I've said so far is pretty accurate.  If so, then I would like to suggest a possible solution that should actually reduce the physics load, should be scalable to much larger vessels, and so could be a potential long-term solution:  determine joint-by-joint which joints will not have physics applied to them at all.  In effect, this would convert some sets of parts into "blocks" of parts that, as far as the physics engine was concerned, would each appear to be a single part.

This could take one of several forms.   I'm sure there are probably many more ways to implement this, but here are a few thoughts:

• When loading a vehicle at launch, apply a "no physics" tag to each applicable joint
• When loading a vehicle at launch, "auto-weld" each applicable set of parts into a single part

I love this game!  Press on, team!

[VlonaldKerman]

6 minutes ago, shdwlrd said:

So much this... Don't get me wrong, auto struts was useful in certain situations for experienced players. (Ex. Large to huge space planes.) But it doesn't need to be a crutch for bad design choices.

I’m not an expert, so please correct me if I’m wrong, but I’m pretty sure that in real life, rockets very rarely fail due to structural wobble. This is either because it’s such a solved problem that it never goes wrong, but in principle it might, or because the way rockets are made minimizes wobble naturally and easily.

In the dev chat, they alluded several times to “realism”, and to be sure, in reality things bend under torque. But I’m reality there is also n-body physics! Would it be accurate to say that, in many or most cases, a jointless or nearly jointless approximation would yield sufficiently realistic results, in the same way that Newtonian gravity simulation does? I think the answer may be yes.

I like the solution one forum member came up with wherein torque was calculated for certain joints within a vessel, but rather than bend, when a joint experienced torque beyond a certain threshold, the joint would fail. Again, this would mean no bending, so less realistic, but as they said in the chat, “[we] could make a perfectly realistic physics simulation that runs at one frame per five seconds” or something like that. If we’re already resigned to approximating, as far as I’m concerned, the aforementioned approximation is fair game.

Again, I say this with little actual knowledge. But if I’m wrong, someone is yet to explain why, I think.

[chefsbrian]

12 minutes ago, VlonaldKerman said:

Again, I say this with little actual knowledge. But if I’m wrong, someone is yet to explain why, I think.

I wouldn't say your wrong, rockets don't really wobble IRL for a variety of reasons (Internally pressurized tanks just so happen to really dislike being flexed along their length which certainly helps). Part of the problem is that the player doesn't really have any engineering tools to control those joints, or the ability to make any of the engineering tradeoffs that could be used IRL. We've got lego bricks, and the bricks go together as the developers designed, and we have no ability to sneak in with a bit of superglue to fuse that lego together. We just got a system where we get what we get, but what we get isn't presented or explained to us, we just get to observe improvised slinkies and question what we could have done to prevent this.

We can change the nature of the problem and the thresholds, but there's no great way to go around any system that has failure states that can't be effectively communicated or influenced by the player. I think that's probably part of why there's so much uncertainty about where this should go so far as a solution design is concerned - It appears the goal is that the player shouldn't have to worry about this actively, but also the player shouldn't be able to get away with clownishly impossible rocket designs.

[Kerbart]

56 minutes ago, VlonaldKerman said:

I’m not an expert, so please correct me if I’m wrong, but I’m pretty sure that in real life, rockets very rarely fail due to structural wobble. This is either because it’s such a solved problem that it never goes wrong, but in principle it might, or because the way rockets are made minimizes wobble naturally and easily.

In the dev chat, they alluded several times to “realism”, and to be sure, in reality things bend under torque. But I’m reality there is also n-body physics! Would it be accurate to say that, in many or most cases, a jointless or nearly jointless approximation would yield sufficiently realistic results, in the same way that Newtonian gravity simulation does? I think the answer may be yes. (...)

Again, I say this with little actual knowledge. But if I’m wrong, someone is yet to explain why, I think.

I think that the fundamental problem isn't wobble; it's failure. Yes, it's very easy to modify the game in such a way that rockets are infinitely stiff and don't wobble. Problem is that it's also possible to make rockets that are super aerodynamic (think shaped like broomstick instead of a soda bottle) and offer very little design constraints.

The reason we don't have infinite stiff joints is the same as we don't have 2500 kN engines with 95,000 Isp, gimballed and weighing only 50 kg. Wouldn't it be wonderful? At the same time, it kills any challenge the game might offer.

Real rockets don't wobble; you're right about that. They just explode. And usually they don't because they're designed by engineers who know what they are doing. Wobble is the way the game gives feedback that you're flying an unsound design. People freak out when their rocket, shaped like a hot air balloon, flips over when they yank it to a 45° angle at 10km. Imagine the complaints of people when their rocket explodes for no reason. Oh yes, the reason is there: your design is too slender, there's too much weight on top, you're using way too much control authority, etc. But now the game doesn't give any visual clues up to that point, just boom. Can we have rockets that don't wobble? Yes, but it would either be not fun (the rapid unplanned disassembly), or, ironically, unrealistic (build and fly what you want without regards for real life physics). That's why it's such a hard problem.

Edited September 29, 2023 by Kerbart

[cocoscacao]

11 minutes ago, Kerbart said:

Imagine the complaints of people when their rocket explodes for no reason.

I just had a vision of NASA engineer making angry rant post on reddit... Rocketry sucks

1 hour ago, Alexoff said:

A long-term solution may also add bugs. In general, any change in the game can create an unpredictable number of bugs.

And it probably will. My point was, if they need to deal with them, they don't have to deal with them twice.

Edited September 29, 2023 by cocoscacao

[Alexoff]

28 minutes ago, cocoscacao said:

And it probably will. My point was, if they need to deal with them, they don't have to deal with them twice.

Well, if a long-term solution will take more than a year, then it’s better to give something to users now. At least make the connections between the tanks more rigid. Chats alone won't make the game better.

[PDCWolf]

I just spent 18 minutes of my life listening to what's essentially a recap of the wobble discussion I've been reading in these forums and anywhere the game is talked about, for the best part of a decade. I understand explaining stuff for the new players and those who might not even know the forum exists, but I seriously expected more than just that... Even if at least just telling us which solutions you're picking. It's like the video stopped right when it was about to get interesting.

As for the discussion in the thread, pfft. I can't believe we're still talking as if wobble was a proper feature. It allows parts to slide into each other, it requires that parts be able to slide into each other, it's exaggerated, unrealistic, and not an indicative of an inherently bad design at any level. It's wholly inappropriate for parts being used on a different way than originally intended (which is a big part of the game), and now it's become clear that not even new players like it or perceive it in any positive light.

Wobble is over, it needs to go.

 

[NoMrBond]

It almost seems like the physics needs a fix akin to the krakensbane solution?

When the part mass difference becomes too large, the smaller mass object starts getting calculated relative to the larger mass object rather than all joints being the same kind of joint

So you only end up grouping constellations of parts with acceptable (to the physics solver) mass scales together, and everything else is relative

So lets say, you have two large objects bridged by docking ports, in the current system physics would go HUMUNGOUS-tiny-tiny-HUMUNGOUS, but what if (at the physics level) the game just did the joint as HUMUNGOUS-HUMUNGOUS with the docking ports visually present (placed by relative position), but not in the physics hierarchy.

[asmi]

5 hours ago, regex said:

What I think could work pretty well would be node-to-node welding of similar size, so a 2.5 -> 2.5 node would weld but a 1.25 -> 2.5 node wouldn't weld. Makes adapters useful, makes a sort of intuitive sense, reduces rigidbodies, doesn't mess with wheels (if they're surface attached)... Naturally this wouldn't apply between decouplers.

Procedural fuel tanks would solve a lot of those issues, as each of those tanks will actually be a single rigid body, with no need for any welding. I know Nate doesn't like procedural parts for some reason, but once I tried them in KSP1 I could never go back to the LEGO stuff.

[Cryptobux]

If you haven't yet watched the video, let me summarize it and save you 18 minutes:

• Rockets are too wobbly
• They're working on it

[Meecrob]

 

12 minutes ago, Cryptobux said:

If you haven't yet watched the video, let me summarize it and save you 18 minutes:

• Rockets are too wobbly
• They're working on it

Also: We have not chosen a path to fix this issue yet.

 

I know the strategy is to "not say something that might later be held against us," but this is getting ridiculous. If you guys don't have the appetite to make wrong decisions in a timely fashion and iterate with experience, as opposed to correct decisions some time in the distant future, I'm sorry, but this won't turn out well. You can't be "risk-off" when you botch a launch so badly.

Also, admitting fault goes a long way, especially with reviews and player numbers as they are (shut up, I didn't post the actual numbers).

That is aimed at whoever makes the decisions, not the devs. Just to be clear.

 

Edited September 29, 2023 by Meecrob

[VlonaldKerman]

3 hours ago, chefsbrian said:

Part of the problem is that the player doesn't really have any engineering tools to control those joints, or the ability to make any of the engineering tradeoffs that could be used IRL. We've got lego bricks, and the bricks go together as the developers designed, and we have no ability to sneak in with a bit of superglue to fuse that lego together. We just got a system where we get what we get, but what we get isn't presented or explained to us, we just get to observe improvised slinkies and question what we could have done to prevent this.

RO allows players to choose the type of tank (structural, non-structural, reinforced in various ways) they use. Beyond that, and with respect to joint connections, would there ever be a benefit to making the joints as weak as they are now?

In other words: the game also just decides that your cabin is pressurized. This is not a reason for the cabin to randomly depressurize (rocket wobble), but for it to ALWAYS remain pressurized because the player would never choose to have it the other way (stiff rockets). There is no need for there to be a joint stiffness option if players would always choose rigid.

3 hours ago, Kerbart said:

Wobble is the way the game gives feedback that you're flying an unsound design. People freak out when their rocket, shaped like a hot air balloon, flips over when they yank it to a 45° angle at 10km. Imagine the complaints of people when their rocket explodes for no reason.

In the suggestion I’m referencing, the forum member gave examples of other ways of giving visual feedback ex. Sparks, visual vibration, an alert system, etc. Real rockets also have internal sensors which give data- I see no reason why this couldn’t be in KSP 2.

In the abstract: does it really seem likely that having rockets wobble is the best way of communicating to players that their design is bad? There has to be a better solution.

[Stoup]

I wanna mirror my sentiment posted elsewhere... I didn't pay 50 dollars for the game to be told that I need to "eat my vegetables" in hope of some unspecified future instead of being able to easily enjoy the product that has been released

[Pthigrivi]

17 minutes ago, Stoup said:

I wanna mirror my sentiment posted elsewhere... I didn't pay 50 dollars for the game to be told that I need to "eat my vegetables" in hope of some unspecified future instead of being able to easily enjoy the product that has been released

I don’t think thats what he meant. I think he meant the devs would need to eat their vegetables aka digging into a more robust and sustainable solution. 
 

The central problem is that neither spaghetti nor a flying brick are good solutions. I think Nates comment that these parts should behave close to our intuition is exactly right, but even that is a spectrum. Being an architect I absolutely expect things to bend and flex because thats what real materials do. People don’t normally see that because if a bridge or building is visibly flexing something has gone horribly wrong. You also wouldn’t expect a plane’s fuselage to flex, but it does slightly under lateral loads. You can also see wings flexing if you look out the window under turbulence. 
 

So Im happy to see the commitment to really solving this problem long term. Obviously what we’re currently seeing isn’t right. And if autostrut puts a finger in the dyke until the real solution is ready thats fine too. 

Edited September 30, 2023 by Pthigrivi

[Stoup]

Yeah, that is a fair take. I hate to indulge in the inevitable drama that results from this kind of ambiguous silly analogy, honestly, but regardless of what he meant, it still feels to me like that's what we're expected to do: Deal with what we have now, because sometime maybe eventually in the future things will be better.

And I acknowledge the irony of saying this in the wake of a video where they specifically talk about short term solutions, but... I mean, why would we need to talk about short term solutions? I don't feel the damage they could cause would be outweighed by the sentiment of neglect that I feel has grown, at least in my own experience

[cyberKerb]

Is there much in the current design scope to put the amount of wobb joint flexibility in the hands of the user that builds the ship? 

By that, I don't means leaving autostrut on by default. I mean getting it down to a level similar to Nates description where certain parts would have an indicator to advise they can only be 'welded' to other parts (no physics cals, but still a break point for collisions), versus a part that is just 'attached' and has physics applied?

or is this temp fix more of a game design / behind the scenes fix with no user input/ choice?

Edited September 30, 2023 by cyberKerb

[chefsbrian]

1 hour ago, VlonaldKerman said:

Beyond that, and with respect to joint connections, would there ever be a benefit to making the joints as weak as they are now?

Possibly - If the players decisions around strength had an impact on the weight of the parts, or perhaps resources consumed, then it would be a decision to be made. You might get more Delta V or use less Structural Steel if you tie them together with Twine and hope for the best.

Its very much a game design decision on that front though. If wobble is intended to be a gameplay mechanic, the player should have gameplay methods by which to interact with it. If its a realism decision, then the player should have realistic means by which to address it. If its not intended to be either and its just a passive side effect of the simulation, then I'm of the opinion that the aspects of a simulation that cannot be influenced by the player should not be hostile to the players goals

[Cryptobux]

As Nate says, "that work is ongoing right now, so we should have some cool stuff to show you soon".

What he means by "cool stuff", and "soon" can be only be guessed at, informed by looking at his track record so far.

[regex]

16 hours ago, asmi said:

Procedural fuel tanks would solve a lot of those issues, as each of those tanks will actually be a single rigid body, with no need for any welding. I know Nate doesn't like procedural parts for some reason, but once I tried them in KSP1 I could never go back to the LEGO stuff.

Even worse, Chris doesn't want procedural tanks in the game either. We're getting LEGOs and everything that comes with them. Real shame.

[jacksawild]

This is such a non issue to be wasting time on. The player designs the rockets, the "engineers" builds them to hold together. The deign is player driven, the "engineering" is from the underlying software.  If they fly is down to the design and never down to the engineers putting them together. Get the software out of the way of the game and concentrate on the things which matter. If the wobbliness is that important to you, come back to it after you've completed the milestones but no one is going to miss it in the meantime.

[Kerbart]

3 hours ago, jacksawild said:

This is such a non issue to be wasting time on. The player designs the rockets, the "engineers" builds them to hold together. The deign is player driven, the "engineering" is from the underlying software.  If they fly is down to the design and never down to the engineers putting them together. Get the software out of the way of the game and concentrate on the things which matter. If the wobbliness is that important to you, come back to it after you've completed the milestones but no one is going to miss it in the meantime.

While they're at it, I hope they give us infinite fuel. That's another engineering problem. And batteries that never drain. And microscopic antennae that work across the galaxy. Those are all things designers shouldn't have to deal with.

[Alexoff]

It seems to me that most of us would be satisfied if wobbling was reduced to the level of KSP1.

[jacksawild]

3 hours ago, Kerbart said:

While they're at it, I hope they give us infinite fuel. That's another engineering problem. And batteries that never drain. And microscopic antennae that work across the galaxy. Those are all things designers shouldn't have to deal with.

That's just obtuse. The player puts the fuel tanks on the rockets and we give it enough to go where we want it to go, same for batteries, just like real rockets do. For a player to solve the wobble we have to ruin our design with stupid struts all over the place just like real rockets don't.

[Kerbart]

1 hour ago, jacksawild said:

That's just obtuse. The player puts the fuel tanks on the rockets and we give it enough to go where we want it to go, same for batteries, just like real rockets do. For a player to solve the wobble we have to ruin our design with stupid struts all over the place just like real rockets don't.

So... wobble is not an issue in real rockets and struts are not used? I think that "making your rocket strong enough" is am essential part of designing it, just like adding enough fuel and Smaking sure you have enough batteries. Your engines need to be strong enough, so why not your rocket?

SpaceX seems to be using struts. But maybe their designers don't know what they are doing?