Major Shoulder Problems

[snogmo]

I've been working on a project for my Character I Class for a long while now and I am having some serious issues with the shoulder area. The main problem is that when I move the characters arm to grab something off his back (the object is on the shoulder of the same arm)the geometry twists right where the arm meets the the torso.

I have tried everything I could think of as well as tried fixing areas that my teacher thought were off. Things I've tried are Painting weights, Sculpt Deformers, rebuilding the IK after resetting the preferred angle on the elbow. I am really stuck and need some help rather quickly. Any help you guys could give me would be greatly appreciated. I'm somewhat of a beginner with maya so the more through of answer would be great.

-SnogMo

[mantrid]

Sure thing. Firstly, arms really are a lot better for the animator with forward kinematics, i.e. animating rotation arcs not using ik handles. You can of course use an IK/FK switch for this type of movement.

Secondly, your 'joint flipping' is NOT a result of deformation, as such, but a result of the skeleton trying to match rotations to your poses during interpolation. Many theories have been penned as solutions, one being quarternian joints, which asre intelligent joints that can analyse the interp propperly. This has been attempted to be included in Maya 4, but simply ain't workin propperly.

So, solutions...if you are using or switching to FK you can set your rotation limits and damping so that it is impossible for the joint to rotate past your defined figures, hence, if it can't roatate one way to match you pose, it must only go the other...the correct direction.

Now, if you really want to us Ik for this, and you are using a Scsolver for your handle, wel, this is not an easy thing to fix with this IK solution. What I can suggest is that you go to bind pose, make a cube at the shoulder joint, group the ik handle to itself and snap the centre pivot of the group to the cube. Now you can constrain the cube to the ikhandle group, and connect the ikhandle group rotation to the cub. Now, you have effectively created anIK/FK switch, as you can animate the rotation of the arm in a controllable fashion, as well as positioning the ikhandle itself within that new local coord space. Ofcousre, setting this up intelligently, with Ik handle for each part of the arm and itelligeent hierarching of the groups will give you really good roatation control using the same theories.

As for IRsolver IK, this is MUCH easier solution. Reason being, you can just use what is called a pole vector constraint. The pole vector, is essentially, the average 'normal' between the jonts within the start and end of the ik handle. I.e., the direction that the rotate plane controlling the joints is pointing.

So...you can select an object that you want to constrain the pole vector to be in the direction of. If you place a cone behind the elbow, select the RP IK handle and shift-select the cone, go constrain>pole vector. Now, if you translate the cone up, you'll notice that the elbow is pointing up towards it. Now, this is not ideal control. So....

Creat a cube and constrain its position to the ik handle or even the wrist joint. Now, make the cube a 'parent' of the cone. Now, when you move the arm, the cone will maintain its relative position. But when you rotate the cube, the cone rotates accordingly, and hey presto, you have direct control of orientation of the elbow.

Now, when you set those key frames, although you'll still get the flipping a little, all you need do to fix it is just make a single keyframe of the cube rotation at a strategic point between poses, and the computer will know not to flip the joint, cos it is following a rotation control from the pole vector constraint.

[snogmo]

I've tried what you said and it didn't work that well...I might not be doing something right but the geometry is still rotating with the should on the X axis. I really wish i could figure this out. If it would help i can email u the scene file and maybe that could show you more clearly what i need fixed. But then again i think i just can't grasp exactly what needs to be done.

[bumpsetspike]

First, the shoulder is the most difficult area of a bipedal character to get really good deformations.
There are several ways to approach the twisting (candy wrapping) problem that you are experiencing with the shoulder. The method I reccomend will require you to add some joints to your arm chain which will give you the ability to have the upperarm rotations in Y and Z controlled by all three bones but the twisting action will be reduced as your bones get closer to the shoulder.
There are books and videos for sale that describe how to do this in detail( some more clearly than others). Here is the basic workflow:
1. You need 2-4 joints between the shoulder and the elbow (depending on the complexity of you model) The more joints you use the less the geometry will shrink in volume when you twist.
2. The bone chain that you have now of just one bone from shouder to elbow will control this other joint chain. I will call this two joint chain the control chain and the multi joint chain the skinning chain. This control can be done many ways. As long as the joint chains are properly built in relationship to each other you can do direct connections with the connection editor for the shoulder in Y and Z. The elbow joint should also get a direct connection in the one direction of rotation. The additional joints between the shoulder and the elbow will follow their parent in Y and Z. The x rotation then needs to be achieved through an expression.
3. The twisting expression is very simple. If the control shoulder is named "rt_sho_ctrl" and the skinning joints are named "rt_sho_skin_a", "rt_sho_skin_b"... , the expression will read similar to this:
rt_sho_skin_a.rotateX = rt_sho_ctrl.rotateX * .2;
rt_sho_skin_b.rotateX = rt_sho_ctrl.rotateX * .3;
rt_sho_skin_c.rotateX = rt_sho_ctrl.rotateX * .5;

This will spread out the twisting action over the length of the arm. How you end up dividing the twisting will depend on trial and error. Generally, the joint closest to the shoulder will only take about 10% of the roation.

1. Now you need to assign the weighting accordingly to this joint chain. Since you have already skinned your character you can add these new bones as influences on your skin cluster and move the weight values to the new bones.

2. And now about that wierd twisting action at the wrist and forearm....see above.

alan

[snogmo]

Thanks a lot for the insight..Its sounds like a very solid solution. I will let you guys know if i get it working when i get to the labs after class tonight.

[mantrid]

OK, email me your scene file and I'll fix it up for you.

[twedzel]

Shoulders they sure are a bitch arn't they. You've allready found one of you're problems, controlling X deformation... as shoulders do not normal twist like a like a dog with ice cream on its ass. Solution... change rotation order. YZX can be quite sucessful in many situations to control that nasty X. But the underlying problem ultimately is, your shoulder doesn't simply rotate around a central pivot point, but is a complexe marriage of major muscle groups attaching your arm to your torso. Good deformation needs more advanced rigs, but this is good for now. Controling shoulder twist is the first step to solving shoulder issues in Ik based arm rigs. Wether you use the solution abouve or any other, this is the key to controlling these more advanced shoulder rigs.