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

[AtomicTech]

3 minutes ago, Rudolf Meier said:

... but, the outcome is just a rigid vessel and therefore a changed game experience. Which is something not everyone wants (including me).

Welp, I am happy with my experience and don't seek to force it on others. 

If they fix wobbly rockets soon, great!

If it's a while longer; that's fine. It just gives me more time to play my new Career playthrough at 10% science.

Edited October 3, 2023 by AtomicTech

[Infinite Aerospace]

7 minutes ago, Rudolf Meier said:

... but, the outcome is just a rigid vessel and therefore a changed game experience. Which is something not everyone wants (including me).

There needs to be a bit of common sense applied to that though, people need to ask themselves what should be rigid, and what should have the potential to deform, that's it. A solid rocket booster, which is made from several mated sections (which are not intended to move relative to one another) should 'not' deform unless in the most extreme circumstances. Now, a connection point where something is mounted (not mated) to something else 'should' be one of those variable locations. The issue is, Kerbal has always done an extremely poor job of simulating these things.

 

What I fully believe is needed, is a 'weld joint' tool. Where the attachment node/joint is for all intents and purposes removed and two parts become a single homogenous part, fir physics calculations and such forth. I don't really see a viable alternative to that. There does however need to be several 'hard rules' embedded within that, an example being one cannot weld a 1.25 metre tank to a 2.5 metre tank, as the contact point of the two tanks is the barrel section not the dome. So for that example you'd be forced, by design to use an adaptor of some form, which is what you'd want to do. It's what real rocket manufacturers do. The whole purpose of that is to psuedo-simulate the idea of having an actual physical connection, whether that's a weld, or bolts between two parts.

 

Where I think deformation is required is mounting solid rocket boosters for example, there should be a structure (strutting) that removes the torque created with having a booster attached to a single point like that. What I don't want to see is an 'auto-strut' setup again as it doesn't address the root cause, and we remove that second example (the solids mounted to the decoupler) from the equation.

 

I honestly believe we in the community need to accept that a solution is going to fundamentally alter 'something' about how the game plays and embrace it as something of a progression to Kerbal Space Program being a bit more realistic, from a physics standpoint. Don't get me wrong I don't think everyone will agree with the above, but I'd like to think I've weighed up the pros and cons of the whole concept.

 

[Infinite Aerospace]

Another area I think requires attention is joints between the base of a second stage, and the top of a first stage, the interstage. Right now I'd argue this is the biggest source of deformation in any rocket I build. The issue stems from the fact the game doesn't seem to treat the interstage fairing pieces as solid entities. You can often see slight clipping in those sorts of locations, and this is due mostly (I think) down to the singular attachment node that connects the two. Any mechanical system like that where there's two things connected at a single point is potentially prone to flexing.

[Pthigrivi]

Im also thinking a tutorial on strutting might be a good idea. I even see really experienced folks like Matt Lowne using them really inefficiently which is why they end up with so many. Its probably because so many players rely on autostrut and so they kind of stop thinking about structure. In fact I see a lot of players who look like they’re trying to mimic autostrut: just putting connections absolutely everywhere, which creates the same performance issues that autostrut has.  With a bit of education I think players could make their vessels even sturdier than Autostrut and with much less overhead. Ive never used Autostrut in KSP1 even for great big modded motherships. You just need to think about where these forces are coming from and counteract them in all 3 directions.
 

 

This isn’t a ‘get gud’ argument though. KSP2 feels spongier than KSP1 did even without autostrut, and I think everyone feels like single diameter stacks should not wobble around the way they do. So maybe treating those as rigid could be an option. If a player goes from 3.75m down to 1.25m that would still flex, but if you could solve that with a shroud. And I do still think radial attachments should still flex, though perhaps not as much as they do now. Nate brought up the robotics parts from KSP1 which famously performed like rubber and made for a really unrealistic and poor experience. 
 

 

Edited October 3, 2023 by Pthigrivi

[Rudolf Meier]

welding -> ok, I like the idea... alltough, I'm not sure if it's needed... e.g. the Saturn-IC (first stage of Saturn-V) had 2 tanks inside and a specialy formed hull where the gap was between those tanks. This part is not as stable as the rest (that's why they reinforced it). Now... if you weld things together, you don't have this potential point of failure anymore. And... well, there may be other situations in which it is definitely the right thing to have and to do (for the game)... one solution would be to make the resulting part weak, if one of the welded parts is weak (KJR Next does make the extra joints weak, if the light part which it tries to bypass is weak... the idea is, not to make a strong connection between two huge tanks, if they're connected via pudding).

But I agree, this could be a good thing to have... and it would improve the performance.

12 minutes ago, Infinite Aerospace said:

Another area I think requires attention is joints between the base of a second stage, and the top of a first stage, the interstage. Right now I'd argue this is the biggest source of deformation in any rocket I build. The issue stems from the fact the game doesn't seem to treat the interstage fairing pieces as solid entities. You can often see slight clipping in those sorts of locations, and this is due mostly (I think) down to the singular attachment node that connects the two. Any mechanical system like that where there's two things connected at a single point is potentially prone to flexing.

This is mostly due to how joints work. It is not like in reality and it is not intuitive to understand what's going on (joints do for example also have anchor points which let them behave as if they're way outside the connecting bodies and stuff like that). But adding more joints between those parts does not help. It increases the load on the computer, but doesn't produce better results. You only get an improvement, if you connect parts behind and after the interstage directly to each other (with one single joint, that's enough, when the mass distribution is right).

[Infinite Aerospace]

12 minutes ago, Rudolf Meier said:

welding -> ok, I like the idea... alltough, I'm not sure if it's needed... e.g. the Saturn-IC (first stage of Saturn-V) had 2 tanks inside and a specialy formed hull where the gap was between those tanks. This part is not as stable as the rest (that's why they reinforced it). Now... if you weld things together, you don't have this potential point of failure anymore. And... well, there may be other situations in which it is definitely the right thing to have and to do (for the game)... one solution would be to make the resulting part weak, if one of the welded parts is weak (KJR Next does make the extra joints weak, if the light part which it tries to bypass is weak... the idea is, not to make a strong connection between two huge tanks, if they're connected via pudding).

But I agree, this could be a good thing to have... and it would improve the performance.

This is mostly due to how joints work. It is not like in reality and it is not intuitive to understand what's going on (joints do for example also have anchor points which let them behave as if they're way outside the connecting bodies and stuff like that). But adding more joints between those parts does not help. It increases the load on the computer, but doesn't produce better results. You only get an improvement, if you connect parts behind and after the interstage directly to each other (with one single joint, that's enough, when the mass distribution is right).

My experience with virtual joints is very limited, near zero infact so I'd have to defer to real world examples in my work for example. So that's useful insight regarding increasing joints not yielding increases in 'stiffness' in the virtual world.

[Meecrob]

So this is interesting. It seems that @Rudolf Meier is the ONLY one to have implemented some sort of fix for the wobbly rockets out of any of us (including the Devs), and we are all gonna not listen to what their experience was?

 

For what its worth, I enjoyed reading your insight into the limitations of the rigid body system in Unity. You went further in depth than the Devs did, Rudolph. Thanks!

[Superfluous J]

10 hours ago, Rudolf Meier said:

I'm not talking about old KJR... I'm only talking about KJR Next v4.2.x

If your mod is significantly different than KJR you may want to consider a new name. KJR has some baggage that you're inadvertently taking on by using the name.

Kerbal Integrated Structural Soundness (KISS) or something like that?

[Scarecrow71]

@Nate Simpson

It was mentioned by you that, since the video was created, Dave had been working on a tool to do comparisons of wobbly rocket fixes (or, rather, potential solutions).  Is there any chance we can see the tool in action?  Or, well, maybe not in action, but maybe (assuming it isn't entirely proprietary) shots of what it looks like, and a shot of the output when a comparison is done?

Secondarily to that, is there potential for this tool to be used for other issues/bugs that crop up?  Like, say, why parts simply fall off, or loss of control?  You know, if there is a potential fix and you wanted to see the results using Dave's tool, is that something that can be done?  Or would that tool require way more coding work than it has already taken, time which would be better spent on coding the game?

Either way, the video was a great watch.  More like this please!

[Vl3d]

Glad to see Nate engaging with the community a bit more.

[Alexoff]

8 minutes ago, Scarecrow71 said:

Is there any chance we can see the tool in action?

Eh, how many videos have we seen from the developers about the game (not talk show) after release? I think the fingers of one T-Rex paw will be enough

[Scarecrow71]

On 10/4/2023 at 2:27 PM, Alexoff said:

Eh, how many videos have we seen from the developers about the game (not talk show) after release? I think the fingers of one T-Rex paw will be enough

My question was really two-fold.

1. If they are developing a tool to help them with bugs and solutions, then they are taking time away from actually working on the game code itself.  I think it would be beneficial to see exactly what they are working on that is taking time away from coding the game.
2. Although I may be in the minority, I think it's interesting that they are simultaneously developing something that could help them decipher and test bugs and solutions, and I'm curious to know if that is something that could be used for a wider range of bugs/solutions.  Seeing the tool in action, and seeing what they are working on, might help the community understand what is going on.

As I said elsewhere, I'm just throwing out my 2 cents on stuff, for whatever that is worth. 

[Lowi_Sace]

Not wanting to be negative, but after 7 months I was kinda expecting more progress on the wobble problem. This problem, after hunting us in ksp1, should not have existed in ksp2. I rather have a rigid body (also better for performance) than having that noodle mess

Edited October 6, 2023 by Lowi_Sace

[Gargamel]

1 hour ago, Scarecrow71 said:

My question was really two-fold.

1. If they are developing a tool to help them with bugs and solutions, then they are taking time away from actually working on the game code itself.  I think it would be beneficial to see exactly what they are working on that is taking time away from coding the game.
2. Although I may be in the minority, I think it's interesting that they are simultaneously developing something that could help them decipher and test bugs and solutions, and I'm curious to know if that is something that could be used for a wider range of bugs/solutions.  Seeing the tool in action, and seeing what they are working on, might help the community understand what is going on.

As I said elsewhere, I'm just throwing out my 2 cents on stuff, for whatever that is worth. 

The creation of such tools does not take time away from the end goal, but makes it more efficient.    When I’m woodworking/machining/ making in general, sometimes I have to stop and make a fixture / jig / or tool from scratch to finish the project.    I could finish the project without the tool, maybe, but it’s much faster overall. to take the time and make the right tool. software wise, when I was making a motion activated thermostat, I had to create a device from scratch to emulate the various inputs, rather than waiting to install it and then have to rip it out again when I found a problem.   
 

Tool creation is a key skill in any creative process and doesn’t detract from the efficiency of a project.   

[Nate Simpson]

1 hour ago, Gargamel said:

Tool creation is a key skill in any creative process and doesn’t detract from the efficiency of a project.   

David's instinct to measure twice and cut once is one of the traits that makes him an extremely effective engineer. It was clearly honed over his years of dealing with similar challenges in KSP1. I'm always impressed by his calm and analytical approach - he never jumps to a prescription before we understand every dimension of the problem we're trying to solve. Trust me when I say that my own eagerness to solve a problem in the "obvious" way is almost always proven misplaced once we've done a deep dive on the causes of an issue. David is a treasure and we all love working with him. 

[Alexoff]

1 hour ago, Nate Simpson said:

David's instinct to measure twice and cut once is one of the traits that makes him an extremely effective engineer. It was clearly honed over his years of dealing with similar challenges in KSP1. I'm always impressed by his calm and analytical approach - he never jumps to a prescription before we understand every dimension of the problem we're trying to solve. Trust me when I say that my own eagerness to solve a problem in the "obvious" way is almost always proven misplaced once we've done a deep dive on the causes of an issue. David is a treasure and we all love working with him. 

What about "show and tell" about some bug?

[Rudolf Meier]

You want videos? I have videos...

I have disabled AutoStruts completely in KSP 1 to show that it is as bad as KSP 2 (still today). Then in the second part of the video I have activated only one mode of KJR Next: the "reinforce inversions" mode! This mode adds only 9 joints to this stock Saturn V.

 

 

Edited October 6, 2023 by Rudolf Meier

[MechBFP]

22 minutes ago, Rudolf Meier said:

You want videos? I have videos...

I have disabled AutoStruts completely in KSP 1 to show that it is as bad as KSP 2 (still today). Then in the second part of the video I have activated only one mode of KJR Next: the "reinforce inversions" mode! This mode adds only 9 joints to this stock Saturn V.

 

 

Yup this makes sense.

I never used auto-strut in KSP 1, so perhaps I just got used to working around it so much and that is why I don't have really any issues with  wobble KSP 2.

Without the ability for those who do have issues though to easily fix it in KSP 2 is really the problem right now, regardless of what the solution is.

Edited October 6, 2023 by MechBFP

[Alexoff]

33 minutes ago, Rudolf Meier said:

You want videos? I have videos...

I have disabled AutoStruts completely in KSP 1 to show that it is as bad as KSP 2 (still today). Then in the second part of the video I have activated only one mode of KJR Next: the "reinforce inversions" mode! This mode adds only 9 joints to this stock Saturn V.

 

 

What about the second stage engine, why does it bend so much? I don’t remember having any problems with 3-meter parts; they held up without struts.

[Scarecrow71]

19 minutes ago, MechBFP said:

I never used auto-strut in KSP 1

I try to never use autostrut in KSP1, instead focusing on correct object placement (same-sized pieces connecting, using adapters) and struts where they are truly needed (radial connections).  I can't say that I never use it - I'm not perfect - but I try not to.

[kdaviper]

5 minutes ago, Scarecrow71 said:

I try to never use autostrut in KSP1, instead focusing on correct object placement (same-sized pieces connecting, using adapters) and struts where they are truly needed (radial connections).  I can't say that I never use it - I'm not perfect - but I try not to.

I wonder if this is why the 6-way adapter in ksp1 is so heavy...

 

[Rudolf Meier]

1 hour ago, Alexoff said:

What about the second stage engine, why does it bend so much? I don’t remember having any problems with 3-meter parts; they held up without struts.

This stock vessel doesn't have a second stage like the real Saturn V, it only has what is the third stage in reality. That's why it is not bending there. And further up, they used a lot of struts around the engines to make it stable.

Edited October 6, 2023 by Rudolf Meier

[GGG-GoodGuyGreg]

12 hours ago, Nate Simpson said:

David's instinct to measure twice and cut once is one of the traits that makes him an extremely effective engineer. It was clearly honed over his years of dealing with similar challenges in KSP1. I'm always impressed by his calm and analytical approach - he never jumps to a prescription before we understand every dimension of the problem we're trying to solve. Trust me when I say that my own eagerness to solve a problem in the "obvious" way is almost always proven misplaced once we've done a deep dive on the causes of an issue. David is a treasure and we all love working with him. 

Looks like David is up for a promotion, gg

[Infinite Aerospace]

On 10/6/2023 at 8:57 PM, MechBFP said:

Yup this makes sense.

I never used auto-strut in KSP 1, so perhaps I just got used to working around it so much and that is why I don't have really any issues with  wobble KSP 2.

Without the ability for those who do have issues though to easily fix it in KSP 2 is really the problem right now, regardless of what the solution is.

I rarely used auto-strut as it was something of a 'cheat solution' to a problem in my opinion (Plus I played the game for a significant period of time before I ever knew it existed!) The wobble in KSP2 is fairly manageable and if I'm honest I've never understood how someone like Matt (Lowne) with the experience in the game he has, has so much trouble with it. I suppose the biggest complaint I have personally, with how I like to do things in game is the lack of structural 'strength' in docking adaptors. There's a few parts that don't seem to work as intended mind, such as engine plates, those are very, very wobble inducing but I can imagine that will get better.

[The Aziz]

24 minutes ago, Infinite Aerospace said:

The wobble in KSP2 is fairly manageable and if I'm honest I've never understood how someone like Matt (Lowne) with the experience in the game he has, has so much trouble with it. I

It's manageable when there's no change in diameter. Go watch the latest @ShadowZone's video to see what it's all about.

... while on the subject of said video, there was a mention of a real rocket that has radial attachments that don't flex - true, but are they connected to the stack by a single attachment point, like in KSP(1 and 2)? Because I don't think so. Any more than one should prevent any flex, but because of the way the construction works in-game, the only solution for multiple attachments are struts. And honestly, one strut per radially attached part is enough.