Page 1 of 1

What are Soft bodies and Rigid Bodies?

Posted: Wed Feb 16, 2005 12:50 am
by mchs3d
I hear about these all the time, but I've always wondered: what are they exactly? :?:

Posted: Wed Feb 16, 2005 12:04 pm
by halibut
Softbodies, think, curtains, jelly, wobbly stuff? :?

I could probably do with an explanation too I suppose ;p

Posted: Wed Feb 16, 2005 2:26 pm
by rednelb
softbody does change its shape according to collisions with other objects or when exposed to forces (eg gravity); and rigid body does not.

Re: What are Soft bodies and Rigid Bodies?

Posted: Wed Feb 16, 2005 2:38 pm
by bjornmose
mchs3d wrote:I hear about these all the time, but I've always wondered: what are they exactly? :?:
Just buzzwords every one uses and everyone understands them a little different.

Rigid bodies dynamics deals with mass centers, torque of inertia, angular momentums ... and assumes the shapes of the bodies do not vary in time.
However the interaction of rigid bodies colliding , linked with joints or springs can give real complicated dynamics.
( pre build engines to simulate that are SOLID and ODE )


Softbodies can be
Elastic solids, fluids, even gas and anything glibberish in between. In CG such systems often are simulated with interacting particles.
Cloth are modelled as a 2D elastic solid.

BM

Posted: Wed Feb 16, 2005 2:52 pm
by Toon_Scheur
Indeed softbodies are things like cloth, jelly, clay, meat etc. Deformable objects.
Rigid bodies, well... the same says it all. Those are rocks, metal cubes etc. Nondeformable objects.

A rigid body would be just a object that is not elastic that moves around as if under the influence of real physical forces. You don't have to animate it by hand, the computer calculates how this objects moves and collide with other objects according to some parameters like dmaping factor, gravity, angular moment, initial velocity and so on.

Now softbodies is much more complex. The object tends to be elastic. Even if the elasticity is homogenous, it would take very complex calculations to determine the position and shape of the object. Consider the object as being a big collection of rigid bodies, all connected to springs internally. So if you want a nice cloth for example, you'll be needing a mesh with lots of verts for a acceptable simulation. Then you have the added difficulty of internal collisions. Most of the time you don't want the object to intersect with itself if it is getting deformed. So additional calculations will be needed to check this. And if the object is about to intersect, it should deform further again to avoid the intersection... and this new deformation should be check again for intersections and so and so.
So thats why the first few Hollywood CGI only movies like Toy Story didn't have any clothes to speak of. Those are present, are only visible for a couple of seconds of the total movie lengt and they don't behave like real cloth really.
So, Blender getting softbodies is surely a milestone in his existance and a step closer to maturity.
One great challange still lies ahead I think: a character dressing up. I've never seen a character putting his/her clothes on. That is more difficult than water some and fire simulation all put together. Even in Shrek 2 they use cut scenes to avoid that problem. For example when human Shrek robbed some guys for their clothes. I don't think that Pixar is about to tackle this problem, because their next project is called "Cars".

Posted: Wed Feb 16, 2005 11:17 pm
by cessen
Dynamics in general refers to animation that is generated from simulations of physics.

My understanding of the differences between the specific types of dynamics is as follows:

- Rigid body dynamics simulates objects that do not deform. For instance, rocks, bricks, or anything that you decide shouldn't deform.
- Soft body dynamics simulates objcts that do deform, but do not change their surface topology. For instance, cloth, hair, a cardboard box, a tire, or anything that you decide should deform.
- Fluid dynamics simulates things that do deform and do change their surface topology (or simply don't have a well defined surface at all). For instance, water, smoke, explosions, jelly, or anything that you decide should.... blah blah blah.

That it just my understanding of the terminology, though. I'm not sure if that is quite accurate.

Posted: Thu Feb 17, 2005 10:34 am
by bjornmose
cessen wrote: .. deform, but do not change their surface topology. ..
This definition would make a "zippered" jacket a fluid ?

BM

hmmm

Posted: Thu Feb 17, 2005 10:36 am
by bfvietnam
Rigid Bodies are as some have said, they are objects that have a rigid
shape, they can't deform.. Rigid Bodies first were made accessible to many people on Maya.. Some of the first tests I saw done with it were
thwoing a bowling ball at a trashcan and watching the lid warble on the ground.. You can tell if something is not rigid body dynamics, if its
missing the torque effect, like spinning a football on the ground will eventually cause the footbal to spin on its tip and oscillate.. Its a good test to see how good the dynamics system is..

Soft body dynamics is for fleshy stuff.. I've seen it used for making
skin, tails, and marshmellow men. Soft bodies seem to retain their shape
and are affected by torque just like rigid bodies.. I mean, if you hit one
side of a soft body, the rest of the body will flip preserving angular momentum. Softimage foundation and Cinema 4D both have soft body dynamics, so does blender.. Blender had rigid body, which Softimage and Chinema 4D don't.. But there isn't very many cases where you are going to want to have rigid body dynamics, since you can't really shatter things
without sculpting them as shattered.. Soft body dynamics are used for a lot of stuff, and tend to made characters look more realistic..

http://www.marcodilucca.com/porfolio.htm

Look at the Dog Videos. At the end of the last "making of" video,
there is an animation of the dog, and I believe the cheek flaps and
tail are soft-body dynamics.. Though I think recently there was special
kind of springy spline chain added to the animation of things like
floppy ears and tails.

http://www.marcodilucca.com/temp/dogueWIP.avi
http://www.marcodilucca.com/temp/dogue_rig.avi

Posted: Thu Feb 17, 2005 2:22 pm
by Toon_Scheur
I don't know if this is accurate, but in every Pixar movie there is a rigid body test. You'll always see a pencil or pen or crayon being dropped. I can confirm this in the case of The Incredibles and Monster Inc.

Side step: As long as we talking about Pixar. The trademark Pixar eyes has a exception in every movie, namely small black beady eyes. In Toy Story it is Ham, in Bug's Life it is the Queen's pet, in Monster Inc. it is the young monster that bit Mike's fingers... but I didn't see it in the Incredibles... anyone know who it is?

Posted: Thu Feb 17, 2005 3:56 pm
by Toon_Scheur
I don't think that fluids can be classified as "softbodies". Although in real life they are soft indeed, the mechanism inside a computer program are different.
Actually, fluids, smoke and fire are computationally similar and different from softbody computation.
For example, in fluid dynamics they use the Navier-Stokes equation for the simulation. Of course there are more clever and faster algorithms nowadays that is much faster than the Navier-Stokes equations (see latest Siggraph papers).
Also a voxel grid is used for this instead of a mesh. Not totaly true..... an ocean view can be done with a plane and some bump textures. BUT if you want splashing water and things falling in the water or floating on it... then you should use fluid dynamics. I think Finding Nemo did it differently. They used some kind of particle deflection system similar to Blender's and wrap a mesh like a flypaper on top of this particles in the scenes when water is pouring on the fish.

I think this should be the next thing after softbodies. Voxel rendering should be fairly easily to program. Its is just an area of stacked invisible cubes with variaring degrees of opacity. Rendering could be a problem if you want refraction and shadow. There is a Pixar paper on this to obtain actractive looking shadows using deep shadow maps. If I understand the paper correctly, deep shadow maps can be used instead of the current shadowmap genaration techniques with no disadvantages. Faster rendering, motionblurred shadows etc etc

Posted: Thu Feb 17, 2005 9:47 pm
by cessen
bjornmose wrote:This definition would make a "zippered" jacket a fluid?
Only if you consider that a change in topology. I'd say that's a fairly weak argument against my definition.

However, that does bring up a much better argument against my definition: what about objects that break apart or tear? That would certainly throw my definition into question.

I tend to think of the differences more in terms of what types algorithms are used, because it's really an issue of what effect the artist wants to convey.

Anyone know the formal definitions for various types of dynamics? Care to share? ;-)