Hello,

I am setting up a ragdoll simulation using kinefx. I am following the tutorial here (time stamped): https://youtu.be/-Gfjgzo0Zec?t=120 [youtu.be]

My question is how do you know which axis is the up axis? The docs say this:

"For Ragdoll, it specifies the twist(direction to the child) and up(direction to pole vector) joint axes needed for the Cone Twist Constraint. For example, a rotation order of Rx Ry Rz will assign the twist axis to X and the up axis to Y."

I am not sure what it means "the direction to the pole vector".

The rotation order for my scene will start with y (see photo), the question is just which axis comes next, x or z?

based on the video it seems like the order would be Ry Rx Rz, but that doesn't seem to work (see first photo). On the second the handles look correct when set to Ry Rz Rx. Why is this? How can I easily know what the up axis is to easily know the rotation order?

edit 1: oddly enough the arms actually need Ry Rx Rz instead of Ry Rz Rx, who ever did this rig must not have done something right...

I figured out a solution to working with a mocap clip where the orients are all out of whack. Obviously you cannot just re-orient the bones because when you plug into a bone deform your model is all messed up because the bone capture weights are based on the original orientations.

The way I solve the problem is to have two versions of the skeleton. The original skeleton and a version of the skeleton where I fix the orients. I keep my original skeleton in a t-pose and just do all my animation on the fixed skeleton. Once I am done I just use the retargeting nodes (map to points, full body ik) to transfer the animation back onto the original skeleton. Then I plug that into the bone deform and everything works perfectly.

Couldn't you just fix your rest skeleton in the same way as your animation skeleton? As long as they're both consistent, it should work in theory.

edward
Couldn't you just fix your rest skeleton in the same way as your animation skeleton? As long as they're both consistent, it should work in theory.

lol, thanks. I didn't know that would work. that is like 100x easier/better than what I did. Thanks

edit. at first I thought it did not work. turns out the issue was the animated pose had extra parts not in the capture pose, so when I tried to compute orientations it was slightly different causing the bone deform not to work. I just deleted those extra joints and it worked.

A very easy way to set up the joint limits is to use a motion clip.

I find this a lot easier, setting them up by hand can be very time consuming. This is my result

willh
I find this a lot easier, setting them up by hand can be very time consuming. This is my result

yep, you can do the same thing (well similar) with agents by auto computing the rotation limits from the animation. The issue is that my rig I downloaded had bad orientations so that feature of auto-computing the limits didn't work correctly(until I fixed the orientations).