I see. I guess that would make sense. Although, I can’t understand how it becomes such a technical problem.
Surely, the model could go from a motion-captured stance / action (e.g. attempting to land a blow), to a ragdoll model, inheriting the dynamics of the former for a period of time?
e.g. the AI attempts to land a blow, the player avatar avoids it and counter-punches to the AI head. The AI models head then goes to ragdoll mode, and snaps back, pulling the remainder of the torso with it using IK. If the AI had momentum when the blow was struck, an algorithm takes this into account. An model running forward is going to react to a kick differently to a static model. If we can calculate combined collision trajectories for cars, surely it’s not too much to extrapolate limbs?
Mind you, I’m not much of an organic animator, and not a physicist!