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

[Stoup]

Maybe the argument should come less from the perspective of realism, and more from one of, is it FUN to require the player to use struts in this specific situation? As in, does it contribute to the stable rocket puzzle? Or does it get to a point where we're being forced to strut excessively? It's important to keep strutting down, if we can, to reduce part count and ease physics interactions between additional joints.

There are cases where struts should definitely used, but like even Nate has said, vertical rocket stacks probably shouldn't need to be strutted to themselves to stay stable. I think that's the biggest gripe when people talk about wobbliness, though I welcome different perspectives.

[Guest]

I don’t think there’s all that much disagreement about the objective here really. @Kerbart describes its gameplay function really well: with no wobble at all badly-designed vessels would just fail for no visible reason.

There are a couple of people who would prefer that, but I think most of us (including Nate!)  just want it brought down to a level where it’s intuitive. It’s clearly wrong that a conventionally designed three-stage rocket on a conventional trajectory flexes visibly, and if it flexes so much that it makes it more difficult to make orbit that’s a major problem. On the other hand, some crazy designs ought to collapse on the launchpad!

Edited October 1, 2023 by Guest

[Delay]

I can't really think of a good way to phrase it, so...

Something something soft-body dynamics? Could that potentially work? I mean, it's a discipline that specifically deals with deformable objects and there are games that use it to plausibly simulate local damages and flexing (Strangely they're all vehicle simulations/racing games)...

Yes, I do not have any expertise. So this doubles as an opportunity for me to ask why such an approach is infeasible should that be the case.

[gussi111]

13 hours ago, Kerbart said:

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

Where do you see vertical struts?

[The Aziz]

38 minutes ago, Delay said:

I can't really think of a good way to phrase it, so...

Something something soft-body dynamics? Could that potentially work? I mean, it's a discipline that specifically deals with deformable objects and there are games that use it to plausibly simulate local damages and flexing (Strangely they're all vehicle simulations/racing games)...

Yes, I do not have any expertise. So this doubles as an opportunity for me to ask why such an approach is infeasible should that be the case.

Soft body physics are potentially even more expensive hardware-wise. In Beamng you have to frequently go in slow motion so that the game won't stutter (or freeze) while it's trying to calculate the crash. And every so often the physics break anyway.

[Delay]

8 minutes ago, The Aziz said:

In Beamng

I was more thinking Wreckfest, but you make a good point with BeamNG. Scalability would probably be a problem with that approach.

What I personally view as a big problem would be how to make soft-body physics play nicely with the part-by-part construction in KSP and docking/undocking.

[Kerbart]

1 hour ago, gussi111 said:

Where do you see vertical struts?

I've gone over my posts five times, I haven't seen mention of vertical struts anywhere.

Do I use vertical struts? Yes, when connecting a large tank, small engine + decoupler to another large tank. And I think it's a stunning lack of mechanical insights to expect the game to not require extra bracing in such a situation. To connect similar sizes parts? Never. It's just a matter of keeping diameter in check with vertical size, aka good (mechanical) design practices.

Now I'm curious where on earth you need vertical bracing between (same sized) parts. I certainly don't

[Guest]

13 hours ago, Periple said:

There are a couple of people who would prefer that, but I think most of us (including Nate!)  just want it brought down to a level where it’s intuitive.

Pretty much. Getting a badly-designed unaerodynamic turd into space should be a real accomplishment of skill rather than just the game letting you do it for because. I originally thought the wobble should be entirely removed but after KSP1 and its autostrut, well it makes the game incredibly easy. For that matter, I'm of the opinion that space tape should have some real downsides in order to penalize using too much. Wobble is about the only real puzzle challenge to solve in this game.

[PDCWolf]

14 hours ago, Kerbart said:

snip

And if you looked at a real picture and not a papercraft model, and did the math, you'd realize those struts are not strong enough to transmit the trust of the boosters. The real, trust transmitting connection happens through almost solid rings at the bottom, linked through a big, central, decoupler structure. Those linkages at the top are merely hydraulic helpers for the thruster separation (which are also repeated at the bottom). If those are there for any structural motif, it's to stop the levering-in momentum on trust, not to transmit trust or prevent the boosters from slopping out.

 

Wobble is not realistic, not difficulty, and not a mechanic. It's a glorified bug that required the devs to permit parts to slide into each other magically in an engineering game.

[The Aziz]

Sure, a singular attachment at the very bottom of a stack near 18 roaring engines would surely keep the booster in place. Mhm.

[PDCWolf]

32 minutes ago, The Aziz said:

Sure, a singular attachment at the very bottom of a stack near 18 roaring engines would surely keep the booster in place. Mhm.

Thankfully, this is not about hmms or umms. We have sources we can check.

First off, Elon mentions the redesign on the center core to accommodate the loads.
https://twitter.com/elonmusk/status/732329443776024576

This was later echoed in the SES-10 conference.

https://www.reddit.com/r/spacex/comments/62i6m1/recap_of_the_elon_musk_and_martin_halliwell_press/

Another source explains that the center core houses all the separation equipment

https://www.wevolver.com/specs/falcon-heavy-block-5

An animation by SpaceX themselves

Great NSF thread where there are a couple exclusive pictures showing the attachment ring at the bottom of the booster, as well as the connections to the side boosters. Coincidentally shows other rockets, proving that nowhere in real life do KSP struts exist.

https://forum.nasaspaceflight.com/index.php?topic=43255.20

 

[Opus]

Give us a welding tool that can be player configured to instruct the game to treat assemblies as defined by the player as single rigid bodies until acted upon by some outside force IE a collision, decouple extreme inertial force etc.  At which point physics can be reassessed on a part by part basis within the assembly and craft as a whole. Allow the players to define the assemblies as they see fit. If a casual player wants to treat an entire 1000 part craft as a rigid body, fine. If another player wants to treat side boosters as separate assemblies, fine. If another player wants all parts calculated individually at all times, FINE. Continue to improve struts and reconsider implementing at least some level of auto strutting.

 

I stepped away from this game six months ago and said "this isn't for me right now, I will come back in a year and try again." Watching the development team has me... Lets just say, concerned that the game will be anywhere near ready in any reasonable amount of time. The world economy is in a really bad place and the purse strings are going to become tighter and tighter. These devs need to deliver results or risk KSP2 becoming abandonware. LETS GO  I am rooting for this team and this game!

Edited October 1, 2023 by Opus
Paragraphing

[MechBFP]

Maybe one day those hydraulics will fail and then you will see what moments of inertia would do to those connection at the bottom.

[Alexoff]

It seems to me that in this case the two boosters have the same thrust and a fairly close mass with the central core. Therefore, the force transmitted to the central unit is not too great. At KSP we often fly on boosters alone, and launch the central unit only after separation. Or the boosters produce enormous thrust.
But as far as I understand, there are no plans to introduce strength into KSP2, because we can set the tanks up to Minmus and nothing except bending will happen. In my opinion, the best solution would be to reduce wobbling by a hundred times and make accident reports more informative. And also add the ability to view the recording of the disaster in slow motion from all angles. Isn't that how it's done in the real world?

[Pthigrivi]

9 hours ago, PDCWolf said:

Thankfully, this is not about hmms or umms. We have sources we can check.

First off, Elon mentions the redesign on the center core to accommodate the loads.
https://twitter.com/elonmusk/status/732329443776024576

This was later echoed in the SES-10 conference.

https://www.reddit.com/r/spacex/comments/62i6m1/recap_of_the_elon_musk_and_martin_halliwell_press/

Another source explains that the center core houses all the separation equipment

https://www.wevolver.com/specs/falcon-heavy-block-5

An animation by SpaceX themselves

Great NSF thread where there are a couple exclusive pictures showing the attachment ring at the bottom of the booster, as well as the connections to the side boosters. Coincidentally shows other rockets, proving that nowhere in real life do KSP struts exist.

https://forum.nasaspaceflight.com/index.php?topic=43255.20

 

Dude they’re still holding the side boosters in place prior to separation. Do you have any idea the moment that you’d see if it was only attached at the bottom? 

[Pthigrivi]

14 hours ago, gussi111 said:

Where do you see vertical struts?

Really no one should be using struts that are parallel to thrust, as even in KSP that doesn’t help with wobble. Struts are there both in KSP and in real life to resist lateral forces. They can be angled (mine usually are to form trusses) but if they’re straight up and down they’re not resisting the right forces. 

Instances where you might see struts within the center stack would be like the Soyus hotstage ring or maybe securing a heavy payload with a small base attachment. In both cases you’re looking to strut diagonally to form a truss. 
 

 

Side boosters absolutely need to be attached in at least 3 places. For smaller boosters I often put the decoupler near the top and two diagonal struts near the base. The decoupler force is enough to kick them out and away from the center stack. For larger boosters I’ll use the doupler near the base and struts at the top with sepratrons  around 3/4 the way up the booster. This usually creates a nice balanced kick-out so the base of the booster doesn’t over-rotate inward and clip the center engines. Again attaching in three places is what gives you your structural triangle for resisting lateral loads. 

Edited October 2, 2023 by Pthigrivi

[Jacke]

I started KSP in v0.25.  It had very wobbly joints.  I finally fiddled enough and manually flew a rocket to orbit.  Once.  It was one of the worst gaming experiences of my long life.

Thereafter, I used MechJeb as it's Ascent Autopilot had a finer hand in those days and I didn't have to deal with this stupidity.

Later, as those stupid wobbly joints were tightened up, I discovered manually flying a rocket to orbit was much easier and enjoyable.

In KSP 2, fix those damn wobbly joints.  Find a solution perhaps better than in KSP 1.  Maybe what KJR mod uses, which is I think making what appears to be a single joint actually a ring of 3 joints.  Or something.  Soon.

[dr.phees]

That's where the huge parts catalogue bites the development. It is beyond me, why they didn't start with a smaller catalog while implementing systems and only flesh out the parts list when ready. Starting with design before systems is a very weird decision.

[Rudolf Meier]

@Intercept Games  @Nate Simpson

I was wondering, if you did look at the latest KJR Next and how I tried to solve the problem.

Old KJR's did simply add an insane amount of joints, that's true. And the game was different after that because almost nothing was breakable anymore and stuff like that.

But because I didn't like that, I wanted to go to a "let's fix the problems of Unity without touching the idea of the game" system that fixes the problems with the least amount of extra joints (and optionally with the old behavior with a lot of extra joints, but that's just an option you can turn on or off). What I found out was, that in most ships you don't need more than 2-6 extra joints and everything is already much more like what you expect from real rockets and what you want for the game.

I think the solution is not bad today and it is also not very complicated.

https://github.com/meirumeiru/Kerbal-Joint-Reinforcement

[PDCWolf]

I've seen this discussion play too many times in this forum. Same drawings and imaginary conjectures without sources, and all. So I'm out.

For those who are new to the discussion, or are actually interested in reading up and getting properly informed instead of blindly defending a glorified bug:

1. How SRBs work, attached at only 2 points (decoupling rings) + video:

Spoiler

 

2. Super tall payloads attached to very small points in real life do not wobble nor need struts:

Spoiler

In fact Atlas has a really tall fairing with an entire rocket stage inside and is stabilized by a single ring on a single point.

3. The ISS only flexes half a centimeter at 109 meters length:

Spoiler

 

4. Why wobble is a glorified bug and not a mechanic (it requires making the physics less intuitive, and hacky workarounds like allowing parts to ghost into each other, from KSP1, mechanically identical in KSP2):

Spoiler

5. How the creator of KSP1 solved wobble in his new game:

Spoiler

 

6. Other counterpoints to wobble (personal takes):

Spoiler

1. KSP2 will have an even more limited part budget than 1 [1]. Asking players to waste their precious part budget in struts is just punishing them for the game's own flaws, and it punishes lower end hardware even harder.
2. Joints are one of the most obvious performance hogs. Struts add joints between parts, thus compounding the impact of #1.
3. Wobble is clearly not a mechanic, as it is never onboarded for new players, explained, or exposed by the game at any point.
4. For a game where parts are supposed to serve more than 1 purpose (think tanks as space station parts) wobble as a mechanic is not appropriate, as it doesn't discriminate use cases. It's also nearly impossible to make it do so.
5. Struts, the in game "solution", is not a reflection of anything from real life other than the cords that'd held up biplane wings in the 20s. Real life rockets use common load bearing structures, hydraulic separators, and so on, as clearly demonstrated. Thus struts create another point of unintuitiveness that's never explained for new players. 

 

[kdaviper]

What if they treated decouplers as a part of the parent part instead of as a discrete entity? Just add their mass to the base mass of the attached part to avoid trying to simulate a large mass differential. Maybe even have a function that merges other high differential connections or applies a smoothing function to their masses somehow. 

Also would it be possible to treat only part of a craft as unpacked? For example if you have a payload inside an undeployed fairing, could you treat the contents therein as packed so that they are treated as one object?

[Rudolf Meier]

3 minutes ago, kdaviper said:

Maybe even have a function that merges other high differential connections or applies a smoothing function to their masses somehow.

the latest KJR Next does exactly this... and it works much better than I first thought (in KSP1, but this would also work for KSP2)

[Starwaster]

On 9/29/2023 at 12:53 PM, regex said:

Please don't implement autostrut, it completely trivialized KSP1. Similarly, it would really suck if a craft was treated as a single rigidbody.

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.

vOv At least you guys aren't taking the easy out.

You can turn autostrutting OFF if you don't want it in KSP1. So turn it off and don't try to deny it to people who want it.

[Guest]

25 minutes ago, Starwaster said:

You can turn autostrutting OFF if you don't want it in KSP1. So turn it off and don't try to deny it to people who want it.

My dude(ette), we're talking about KSP2 here, where the developers are at least trying to construct a coherent game with inherent challenges and actually thinking about what they want that game to look like, not a decade-old barely-designed set of poorly-integrated incoherent ideas someone called "a game". Let's stay on the same page.

[Rudolf Meier]

26 minutes ago, regex said:

My dude(ette), we're talking about KSP2 here, where the developers are at least trying to construct a coherent game with inherent challenges and actually thinking about what they want that game to look like, not a decade-old barely-designed set of poorly-integrated incoherent ideas someone called "a game". Let's stay on the same page.

That's why we need to talk about the problem. We need to understand it and then we (or, the developers) can try to implement a solution.

So... what is the problem? ... the answer is easy: it is Unity. As long as you use Unity + RigidBody + Joints, you will have all the problems that you have, when using Unity + RigidBody + Joints. And those problems are well known, fully understood and we also know what can be done to against it. What works, what not. So... it's not like flying into a completely unknown part of the universe. We know why it is what it is and what can be done! Why are we talking about it, as if nobody has an idea what is going on?

And by the way, for the "autostrut" discussion: we also know autostruts. We know how they work and why they are not a good idea. Why are we then talking about "autostruts yes or no?" instead of "why are they not good?" "what is the problem with them?" "can we find a better implementation?" "are there ideas?" ... it's so annoying to read about this without seeing a real discussion about solutions!