Simulating a concave space within a ragdoll

I’ve talked with oogaboogaman a little about this, and whilst he kindly provided some interesting hints about making ragdolls as a whole, this concept sadly seemed a little too “out there” for any plans to be made for it, I guess spending time making stuff that isn’t entirely experimental might be a wiser.
Also I don’t recall saying thanks for the help, so if you’re reading this, thanks. Also before I go any further, I’m using 3DSmax 2011.

Still, this is my problem:

I have made a chest of sorts, one that is empty and has a lid on it. I’ve made it a ragdoll so that it can be opened and closed in a more analogue, Amnesia/Penumbra way rather than the robotic func_door_rotating option.
This opening thing works using two bones, one for the lower section, and one for the lid. Bone02 and Bone01 respectively.

Everything compiles without problems, the top opens and closes properly, the limits work and altogether all is well. With the possible exception of the “cavity” being physically solid, but visually hollow.

Now, After some latent research and some changes to model to make it empty rather than full of junk, and therefore with the need of the lower part not being a solid collision model, I’ve found that source actually does not support Concave phys models on ragdolls. Blast.

Being a determined little bastard, I haven’t quite given up, as one, it would be a useful model if done right, and secondly I’ve spent long enough on it already.

After some thought, I found one interesting thing; Source does support concavity *between *bones, meaning that If I was to split each side of the chest into a separate bone, the “inside” would be considered between the bones, and would as a result be a containing space.

So that’s what I’ve done, I’ve split the model into a total of 6 parts, the top, left side, right side, front, back and bottom. Each has a bone assigned to it. As well as Bone01 and Bone 02, there is now RightBone, LeftBone, BottomBone and FrontBone. Each are correctly skinned and the like. I have used the linking tool to link them to Bone01, which I have made the parent bone in the .Qc.
The bones look like this (I’ve removed the front bottom and sides for this picture, obviously)

So if you’re with me so far, this is what the schematic view of the bone hierarchy looks like:

As far as my limited bone-knowledge goes, that seems about right.

I also make the appropriate physmodel parts, and rig those up and export everything as normal.
The result is of course, awful:

All of the physmodel parts that where attached to the new bones have jumped to Bone01.

Here’s what it looks like if I don’t link the bones:

Which is of course, perfect. Only problem is that when I go ingame with that, this happens:

Which is as tragic as it is hilarious.
This proved really handy for another model I made months ago, but I digress.

Upon closer inspection, GUIStudioMdl is “Collapsing” the new bones to Bone01, so that the different panels use only one bone, which is understandable.
The only problem with this, is that now that it’s one bone, the concavity is lost and the extra bits of physmodel jump to the single bone.

So, after wading through all of those sentences which are about as concise as someone desperately trying to meet a word count, here are my actual questions:

1.) How would I prevent StudioMDL from Collapsing the bones?
2.) If there is no way, how would I make it so that the model’s different parts work as separate props, but so that they act like they’re “Parented”, or immovable from each other?
3.) Am I crazy for trying such a thing? Is there a reason why I haven’t seen anything like this before?

Any bits if insight and I’d be hugely grateful

drop bone01 and make bottom your root (or any bone really)
Also are the other five bones actually rigged to the mesh itself?

All 6 of the bones are rigged to a part of the mesh.
Here’s the idea:

It looks like you’re doing it correctly. (I had this same issue before) with rigging the sides of the chests to different bones. I don’t know why it’s collapsing the phys boxes onto your root bone though. A puzzling quandary. How did you setup the physics constraints for the wall/floor pieces?

Only the bone for the lid is constrained at the moment

Otherwise, the other bones aren’t constrained or mentioned in the .Qc

I think I once tried to constrain them all, to try and get any sort of different result, but I can’t remember anything interesting happening

Try adding the other bones in, but with zero’s for min/max values.

Can’t constrain the root. Would throw you an error otherwise.

I just tried containing all the bones (I’d left all the constraints in the Qc, just commented out, Hell if I’m pasting all that stuff again) and it turns out I hadn’t tried constraining the Root the first time.

Something of a problem however.
It comes out with this error: “ERROR: Rotation constraint on bone “<One of the new bones>” which has no parent!!!”

Complete with multiple exclamation marks and everything. It must really not like what I’m doing.

Still, this brings on more thing up.
When I gave the model to Oogaboogaman to look over, he had found that the Root, Bone01, wasn’t there.

The bones that apparently have no root are all linked to Bone01, which more importantly, has been made the Root within the Qc.

So, apparently the root doesn’t exist, when it functions as the root just fine ingame.

Well if it claims it doesn’t exist, I get the impression it’s not rigged to anything and being truncated out.

As to rotation contraint on bone which has no parent - Like I’ve said before - you cannot put a rotation constraint on a root bone (first in the chain)

The bone that seemingly doesn’t exist is the parent and root, it’s rigged to the back panel of the chest.

I hadn’t put any constraints on the root either, that error is for all of the bones that aren’t one of the hinge bones (Bone01 and Bone02)

I remember that once, whilst experimenting, I’d got it so that all of the panels where still connected, but really loose, like welding two props together in Garrysmod with only one weld.
I can’t remember how I’d done it though. If I’d applied constraints then something good might have happened.