Is there a mass based rig? Opinions?

[queequeg]

Could you elaborate on this? It would make setting this up a somewhat easier if I could use IK, but I have not found a way to do so. Where should I use it and how?

I've added an explanation (see my last post) of what the rig actually does, so maybe it becomes a bit more clear that it's not IK. It's not implemented with IK, and I highly doubt it could, though I may be wrong.

[failure]

Let's stop for a moment and let me ask you a simple question. Please answer as precisely as you can, I want to be sure that we are talking about the same thing.

What is 'center of mass'?

[queequeg]

I think this is a very good suggestion, because after all I think most of the we do not talk about the same thing.

In a particle system (I'm talking physics here, not Maya), a particle i is defined as having a location in space, let's call it p_i, and a scalar mass, m_i. The center of mass of the particle system is a point in space being located at

P = (m_1*p_1 + ... + m_n*p_n) / (m_1 + ... + m_n),

where n is the number of particles.

Look at this if I didn't make myself clear.

Your definition?

[joojaa]

Yes yes yes... but the animation isnt realy good at whats going on. The real question is that if you wnat the thing in a SPECIFIC shape is it intuitive for you to put it in that shape, and then then the next, the next next.

Theres no way i can tell form your plablast how teh controlls work, and how intuitive they are. In reality yu sould be considering a few other things to such as angular momentum, momentum joint stiffness etc etc. But tehn you start animating forces wich is not intuitive.

Center of mas insnt that special, after all its just a spatial vector integral over density/mass. BAsic machine eingeneering stuff from the 1800s. Im a mechanical eingeneer by education. mechanical eingeneering is by the way where the concept of ik and fk came form (theres a special area of physics called kinematcs taught mainly for mechanical eingeneers).

[queequeg]

Joojaa, I already wrote in the post what controls there are, and what they control. It's the paragraph that starts with "Controls".

As for rotation resistance... um, how exactly do fk/ik rigs have control for rotation resistance? As I said like a million time before: IT'S NOT MEANT TO BE A COMPLETE RIGID BODY SOLVER! There is no support for torque, and I really can't think of a way to have a UI for torque-related stuff.

And I know that the concept of CoM isn't "special." But normal FK/IK rigs don't have support for it, and I'm looking for a way to change that.

If you want a specific shape, and the shape is much more important to you than where the CoM is: USE FK! I said that before, too.

[failure]

Yes, you described the controls but you "wisely" skipped the most important part. How the com constraint actually works, what is interactive and what controls what. You described the parts which we all know already but left out the new stuff that you're suggesting, namely the infamous com constraint.

Make a detailed description utilizing relevant physics and math. As things are now, too much things in your proposal/examples is heavily undefined. And I can't understand how it's supposed to function, therefore I wouldn't go into deeper discussion ( until you present it better ).

Concerning com definition, I think that we are talking about same thing after all ( What Joojaa wrote is actually general level definition. Particle system com is just a special case of it ).

Remeber: you can't spatialy constrain the com of the system to some part of the same system. It's because you need to know exact spatial state of the system in order to calculate the com. It'll cause a cyclic dependency in the calculation. X depends on Y, Y depends on X. No defined solution.

Also; if you know the position of system's com you can't calculate exact spatial distribution of the mass ( discrete as n - particles or continuous ) since infinite number of states of the system can yield same com vector.

That's why rigs don't support it in a way you want it. They can't.

Com is abstaction of the system of masses. It's aggainst the logic to constrain ( in a general way ) the abstraction to a part of the concrete entity that abstraction represents.

[queequeg]

Hi failure!

While I tried to write an explanation, I think I realized why we don't understand each other. I hope I'm right.

There are two points of view on a particle system:
1. There are the world space coordinates of all points and those of the CoM.
2. There's the world space position of the CoM, and the position of the points relative to the CoM.

You seem to think mainly in terms of the first PoV: The CoM is part of the system. I OTOH think mainly in terms of the second: The CoM IS the system. I believe this is why we don't understand each other. Both PoVs are valid, we only have different preferences on them. This is probably why you asked what I think the CoM is.

I really hope that I'm right with this, because TBH I started to think that you wouldn't understand me on purpose! (sorry) And you probably thought the same of me. And this PoV thing would be an explanation of how it's possible that we were not.

So before I get into further details on the rig, I'd really love to know what you think of this. Maybe then we can begin to talk in the same world.

Cheers,
QQ

[failure]

I think the problem is that I'm pointing out some conceptual difficulties you are not aware of yet. And it's because you didn't get deep enough into the problem. Please, try to write it down and maybe you'll understand what I was blabering about in my previous long and boring posts.

And, no, coordinate spaces has nothing to do with this.

Com is not the system. It's a certain simplified representation of the system. If you use it that way, you'll have to agree with some limitations, namely: that you'll treat your system as an infinitely small rigid body.

But important thing is: com is consequence of the current state of the system and cannot be used as an input parameter to the system. You cannot force the com upon the system. You cannot walk into the middle of the particle system and tell it: 'Your com should be there! Obey!'. There are no mathematical means to do that.

As I already said; try to write detailed explanation of your proposal. Do a reality checks. If you bump into a wall of non-feasibility you'll undarstand what I'm talking about.

And I'll be more than happy if it proves that I'm all nonsense. In that case we'll all benefit from having a new and powerful tool for posing.

Btw, why would I want to not understand you on purpose. That's plain silly.

[joojaa]

umm the nimator is interested in how the character goes to certain poses.

And yes just like a two bone ik has infinite number of solutions its costrained to a plane and theres still 2 solutions so the ik solever needs a prefeered angle to kow wich way to bend.

Now the comdirven animation clotroll has also a unlimited number of solutions. Im guessing tough that theres some order of execution that limist the coms you can solve true enough. This has the potential of becoming the riggers nightmare.

Also what do you mean if you want to pose use ik. What is ithe usefullnes iof a rig that you cant pose. If the director says put the hand higher you put the hand higher, he doesnet care if i your rig doesent allow you to. Dos os without standing on toes, wich he then says no i dont want him on toes. DOH!

Altough certain intelligence is in the fact that some com based approx happens between keyed poses. But ist still not foolproof way. The animators might still get anguish becasue they again cant controll the movement.

[queequeg]

failure, I agree that it's completely silly, but I never said that I'm not silly! I try to explain myself for some time now, and to no avail, so I was looking for a reason, and, since (I thought) I really tried, that was the next best explanation. (As I said, I'm sorry)

Alright, here's how I got Maya to see things from my point of view: I have the concept of local and proxy space. Basically, both are representations of the particle system. The local space is what I called the first PoV: The CoM is computed in this space.

The proxy space is just an instance of the local space, but here's the trick: When the CoM moves (in local space), the entire proxy space is moved (for example by an expression) in the exact opposite direction: Let's say the pointConstraint moves the CoM in the local space from position P to P+v, then the proxy space is moved from P' to P'-v. And if I additionally parent the proxy space to a control object, I effectively get explicit control over the CoM.

Here is a script that creates such a setup, and, just in case, a movie to show how the setup is supposed to work.

Normally, both the local space and the CoM are hidden. So the user works only with the proxy space, by translating A, B, C, and the CoMTarget. Usually, when I think "CoM," I think in the proxy space. So I hope that makes clear what I mean by this whole PoV thing. And this is what I mean by "pointConstraining the CoM." It's not how it really works, but OTOH it's what it look-and-feels like.

Now: If you imagine how this can be used in a rig, namely by having those mass points parent(Constrain)ed to joints, and rotating the joints, you get a way of effectively controling the CoM of the rig!

And BTW I don't think your posts are long or boring. In fact I really appreciate that even though we don't seem to understand each other most of the time, you keep posting. And I think I learned quite a bit so far, so: thanks to you.

[failure]

Now how this can be used for better posing and more realistic character weight?

Let's suppose that green locator represents arm and other two locators legs. When animator moves the arm to a new position, the legs go completely berserk. And all this just to keep some semi-relevant abstract point called com in the same place.

Besides, from the animator's point of view, this is very unpredictable and unintuitive kind of behaviour.

[queequeg]

No no no! The body does not only have one arm and two legs. It would give you very bad result indeed if you'd do it like that, because arms and legs don't have a common joint. And you still don't seem to understand what I call the second point of view: See, when you change one of the points, you do NOT change it's world space position, but only the position relative to the other points. You move the point in the local space, not in world space.

What that means for a rig: You can still put the character in a pose like you're used to, because the pose is only defined in relative terms. Two steps: Create the pose, place the CoM. As opposed to: Create the pose, place the root bone.

Note: I use CoM and what I call CoM target as synonyms. They are always located in the same place, so this should be allowed.

The example the script creates would be useful iff the character is jumping, falling, an astronaut in space, etc. I.e. when there are no influences (forces) from outside on a _single_ part of the system (like floor-feet interaction), but only on the system as a whole (like gravity).

Let's say the character did just jump off a cliff. The animator would work in two steps:
1. Create a quadratic curve, and create a path animation for the CoM. The rig makes sure that, whatever the pose of the character is, the CoM stays on the path.

1. While the character is falling, use FK or IK, plain and simple, to change the character's pose. Or rotate the character, with the CoM as the pivot.

The only difference between this and a rig that has no knowledge of mass is that the animator controls the CoM, as opposed to the root bone. Physics tells us that the CoM is also the pivot for any rotation of the system, so this is where the rig is helpful.

Let's say the character is in an "O-shape" while he's falling. The root bone would be located at the perimeter of the character, while the CoM is, well, right at the center. Now if you rotate the root bone, you get a pretty weird result, because the pivot is weird. If you rotate the CoM, you get just what you would expect. This is why I think the CoM (target) is a better representation of the system than the root bone.

[failure]

In reply to:

1. Create a quadratic curve, and create a path animation for the CoM. The rig makes sure that, whatever the pose of the character is, the CoM stays on the path.

2. While the character is falling, use FK or IK, plain and simple, to change the character's pose. Or rotate the character, with the CoM as the pivot.

Problem is: you can't freely pose the character once you have enabled the com constrain. Whole posing mechanism starts to behave really nasty ( read again my last post, and look again at your playblast )

[queequeg]

You can freely pose the character, please read again my last post. I've edited it after I realized that, again, we were not talking about the same thing.

[joojaa]

Well i can se thet comm anim is usefull if you animate characters forating in space

[failure]

I understand you pretty well. I even understand flaws in your logic.

And I will not issue further comments until you show something that actually works.

What you've described so far is doable with couple of expressions and doesn't bring any relevant improvements in terms of better posing or character weight.

[failure]

Yes, sounds really ironic.
Behold our new com based rig! It's best for achieving better illusion of weight in weightless environments.

[joojaa]

I think the general concept problem is that he seems to think that pose is a function of com, while the truuth is taht com is a function of pose. But this isnt a function that works solvably both
ways.

Also fluid com movement doesnot mean fluid overall movement. In fact dancing can be fluid even if the com ovements in dancing are quite nonfluid.

for example a human walking the feet and body arent functions of com but rather the com is a function on how the bodyparts move. Theres a attempt to do so tough made in the sony thing, but it relies on pinning certain pars of the body and let thos non pinned move with the com, thisway you actualy pose and the com moves with the posing. Se posing is still ethe main ussue. not where the center of your mass is. But its oly a posing tool the actual movement still relies on conventional interpolation.

Try making a good walkcylce with your com anim constraint. If its easier than without it then it passes 1 isntance. then make a tumble etc.

The way i see it is that it works quite well as long as the charecter is in air but as soon as it hits a surfcse com becomes a slave of contactpoints that dribve it to different places. But even thee air wacum thing has problems with roll.