I have worked with hierarchical blended animation systems for a number of years, but I have never been quite sure how to handle scale correctly. The introduction of non-uniform scale is additionally puzzling. Does anyone have any insight into the right way of doing it? What I am trying to determine is when in the hierarchical concatenation that scale is included. My components are stored as a follows:
vector3 position; vector3 scale; quaternion rotation; quaternion stretch_rotation;
What I am finding is if I include scale into the matrix during concatenation, I run the risk of grossly amplifying the accumulated errors down the hierarchy. Does anyone have any insight or articles here? To be clear - I know how to do hierarchical animation just not the right way to handle scale.