Jason Winnebeck -- Skeletal Animation Project -- May 14, 2003

How Does Skeletal Animation Work?

The skeleton is strictly a hierarchy of transforms, and is affected directly by its single parent. Each joint has 5 matricies:

• Reference Matrix: the initial starting position for each joint.
• Offset Matrix: transforms mesh coordinate space into joint space.
• Transform Matrix: the current transform of this Bone, in Bone's coordinate space.
• Combined Matrix: this matrix is equal to this bone's parent's combined transformation followed by this bone's transform matrix.
• Skinning Matrix: This matrix is the transformation from the combined matrix followed by the transform of the offset matrx, and it is this matrix that gets sent to the hardware

Binding Types:

• Rigid Binding - used in Half-Life for speed, used now for mechanical objects. One bone influcence per vertex.
• Smooth Binding - used now to model biological actors. Each vertex has blend weights adding to 1.
  • FM = M1 * w1 + M2 * w2 + ... + Mn * wn

Rendering in hardware:

• Non-indexed blending : model split up into segment, for each segment, load small set matrices and render segment.
• indexed blending : each vertex has indices pointing to a large matrix palette containing all bones' matrices.

back || Implementation Design