Collision with hierachical nodes problem

[user_deleted]

I'm seting up cloth collision objects on a robot character. I have a few standin lowpoly nodes attached directly to the rig as children - either under a joint, or a locator/group.

For some reason the arse and the legs work just fine, yet the chest and arm segments act strangely:
When connected to the joints (chest is just a child of a simple locator) - they are ignored - have no effect.
However, if I simply un-parent these poly objects and re-simulate, then they work perfectly as colision objects.... reparent, and bing! not colliding again.

It feels to me as if thier effect is somehow being transformed, although the mesh is not moving.

The functions of my cloth set up are all cool - the thickness and scale and all the actual simulation parameters are all doing thier thing well... its just this darned parenting thing....

I'm at a complete loss as to how to resolve this one... I'll try constraining them and stuff, sure - But how can this be? whats going on? It's particularly wierd, as the legs and arse are working fine like i mentioned.

Anyone seen something similar? Got any ideas?

cheeers!

[kubba]

I have same problem with IK handles, as long as I have them turned on cloth penetrates object.. A/W support was no help for me so it seems I'll have to find walkaround on my own..

BTW maybe you could point/orient constraint your geometry to skeleton, but I would not expect any miracle. If nothnig works you could try to bake anim into your bind skeleton and run simulation without rig..?

[user_deleted]

Well at least I aint going crazy!!

Its a bit worrying that IK might be the cause, as you'd think the use for Maya cloth is 95% of the time going to be IK related.

The wierd thing is I have a complex rig for my characters lower half, yet that works fine.

Baking out is one thing, but it's a real hassle, particularly as the garment could limit the movement in some way.... so I need the interactivity while designing the movement.

[scribbilyg]

Just to pose a two-cent related question... what version of MAYA are you guys running... we use 4.02 at school and i'm having a helluva time as well... are there more 'stable' versions that handle cloth functionality better than others....?

although it seems as you are both more skilled than i, the problems remains and runs a little deeper than seems normal...

just a thought...

g

[user_deleted]

I'm on 4.03 - Its a good point. I believe that even different BUILDS of the same release can exhibit considerable differences in cloth results, but this is probably more simulation related (according to the maths involved) than being related to what I'm starting to think could be major flaws in functionality.

I agree that it seems there are BIG issues involving IK and cloth like KUBBA says... I've spent the last MONTH looking at this problem, trying all sorts of things - and I can only say its pretty grim. Once I can nail down the issues a bit more I'll take it up with AW, cos something seems very wrong, as you suggest.

• I'd be very interested to hear what AW said to you KUBBA (although you said they were no help)

For example: I had my collisions woking well today briefly - until I used an expresion to drive the weight of a constrain that constrained the IK pos. of a chain that had my collision node on (phew!).. then the colisions stopped working..

I've just been sorting thru posts, moving old posts to the new forums, and I've seen similar problems keep arising, and no answers (to my colision issuse, that is).

[kubba]

well, I did small progres, I've deleted all handles and than when recreating IK's playing simulation after each step. As long as I dont have pole vectors, everything runs fine. When I delete them, simulation runs fine again... I have tryed to change my rpIK to scIK and use aim constrain instead, but again, simulation stops working..

10 minutes later - Boys you owe me a beer if it works for you- It WORKS, heres recipe:

1) create dummy
2) pointconstrain it onto your original polevector manipulator (!do not freeze transforms!)
3) conect translate of dummy into polevector..
4)let me know if it works for you..

-guy who was assigned to my problem was some Paolo Berto.
During a week I got replies like- change collision offset, increase base resolution, even got poly sphere with bone and IK inside and piece of cloth and words "look its working for me", and the end was "I did not make your setup, I dont understand it"

[user_deleted]

I've been testing too. it's definately certain rigs which mess it up.

I made a new test rig with a basic skeleton and spheres - i tested increasingly more things in the rig, simulating step by step like you have been doing.
It all works good for me directly keying the ikHandle Pos, and the body pos - but my character to be clothed is far more complicated rigs. I dont know How im gonna find a workaraound.

Tests got as far as having it work fine with the ik handle point constrained to two locators.
I could manually adjust the weights for the constrain, and it worked ok.

As soon as I drive these weights with an expression (or set driven key) then HEY-PRESTO! the arms suddenly stop being cloth constraints - disconnect, and the sim works again(like your pole vectors)

I have earlier been thru countless different attempts at getting this to work. But it always seems to come down to the same issue. There are some things which trip it out.

the tests continue...

(BTW... great help from support you got there...)

[kubba]

this morning I got bit further- I can use PVconst BUT not on original controler. On one leg when I point constrain new PV to original controler it works smoothly but not on other. When I just parent them it's ok. ...strange... anyway now it seems that it works for me, my problem is that we have around 5 months of animation from group of animators hanging above this rig so I just cant start over with anything simplier..

One more thing, you could connect them via reverse node? Once I had problems to get output from nodes into expression..

[user_deleted]

What is the character you are animating? How about your colision objects?
I have many coll. objs - a chest, upperarm, lower arm... etc- - as my character is a robot.

Given this, I was thinking of cutting my losses, and just making a small bake script that i would run before cloth simulation. The script would just burn the transformations of the IK to a seperate hierachical object group - so each clision object would just be moved and rotated. And it wouldnt interrupt work flow too much - just a quick bake before every sim - in fact I'd put a bake-n-sim ™ button on my shelf, i guess...
- But of course if you have a true skinned character then this is no use to you.
Is your rig entirely constraint driven, or do you have any expresions in there?

My test rig (which I thought I'd at least pinpointed the problem on) now works fine!?!? and I was just about to send it to AW.

Its interesting what your getting with the PV tests... and its good to have someone working with the same problem!

Incidentally does your scene report cycle check errors?

BTW - nice idea for the reverse node.

[kubba]

It's a project involving realistic human characters, collision object is one smooth skinned object with several lattices, influence objects and corrective blendshapes as collision object is unsmoothed mesh..

It's both, expression and constraint driven.

For Cycle check, there are two on arms but it works anyway.
One cycle on spine manips but it's only manipulators expression that is not transfered onto skeleton (cloth worked fine with that before).

so.. now it works (poleV on locator parented under original manipulator) as soon as I'll have some spare time I'll take some more in depth look onto this prblm to avoid it next time..

[user_deleted]

hmmm... seems like you have the lucky rig..

Even with a new ik handle and no Poleconst my arms just aint doing it right As usual switch off the IK, and the cloth collision works fine

Its certainly some culmination of rig events that cause this, cause its hard to recreate the specific problem...

Back to the drawing board - thanks for the discussion anyway, its certainly shed a bit of light on this for me...

[kubba]

4.5 works for me, try it..

[user_deleted]

for the record...

I've since then rebuilt my 'entire' rig (keeping nothing from the orignal) and it is working perfectly. It seems that some wierd problem within the IK particular to my rig was screwing it up.

Not sure if its important, but I made sure not to freeze positions of things in the rig too. I would generally use locators to drive IK handle position/orientation thru cnstraints - but I skipped that too, and im now animating the real rig elements driectly, saving lots of (needless) constraints.

With this clean 'bare bones' rig, the multiple cloth constrainer objects work perfectly, and I'm very happy with the performance. I was going to post a few tests I've done, but the test character animation underneath was thrown together real quick, and makes me look like a crap animator. Hopefully I'll get some movies up next month when things are more polished. Just quit my job, so no axcess to 4.5 'till I get to England