Each set of joint animation information is basiclly a transform for that joint. That information is directly set as the local transform for that joint in panda, no adding or anything is being done with the previous frame of information.
Blender gives you lots of different animation information; poseMatrix is sightly different then “quat” and “loc”, and localMatrix is different again. Panda will want the “localTransform” for that joint (which is relative to the parent joint) for each frame. There are many methods to figure out the bones’ local transform, I played with data until I found a method that gave me what was needed.