So this comes from this post
Why a lot of people are having this issue?
I did some test and find out that, in DCCs, to make animating easier, we tend to make pivot of limbs on two side of body appears to be axial symmetric, one uses left hand pivot, another side of limb use right hand cord pivot. Therefore, if you select two should bones and rotate, their animation will appear symmetrical.
However, this is based on the fact that these DCCs support IK calculations in global spaces. So as long as pole vector is provided, the IK will work whatever f*** orientation the bone’s pivot is.
In UE’s current modular rig system, I think the arm/leg IK uses a local coordinate system which only works in right hand coordinate system. It requires that skeleton should uses centroid symmetrical bone orientation.
So after I reorient the bone in DCC, and then reimport that in UE, the goofy IK relieved a little bit.
See, you should not let bone’s orientation interfere with IK calculation, in a classical 2-bone IK system, you can solve mid point position by using trigonometry
After reorienting one side of limb bones so that limbs bones’ pivots on two side of body are in centroid symmetrical pattern, and manually reverse primary and secondary axis vector, the twist problems improved a little bit
Another problem is that in modular IK, hand controller’s rotation also propagates backwards, which is not always an ideal behavior. We just need the translate channel to back propagate, while the rotation only propagates to child bones. It even happens to Manny
