In the meantime I have come up with the following rules, which seem tolerable:
- The pelvis_L/R bones are constrained to only rotate around their single axis and copy 15% of the rotation of the thigh control.
- upperleg01_L/R copy 85% of rotation of the thigh control (i.e. lagging behind by 15%)
- upperleg02 rotates and scales to bridge the end of upperleg01 and the end of the thigh control.
For shoulder I'm less certain, so:
- clavicle and shoulder rigidly rotate as a chain with the single shoulder control bone - not sure at all how the angle between these two bones should behave.
- shoulder additionally copies a single axis of rotation of the arm, like pelvis.
I would like some hints if this is not how it's intended to be used.