Game Development Reference
proposed here differs in the sense that the interior is explicitly created by an artist.
This gives you the flexibility of creating anatomically correct deformations, but
at the cost of additional setup work.
13.4.1 Muscles and Bones the Old Way
The previous section described ways of resolving the volume problem as it arises
in the skinning process. Unfortunately, avoidance of rubber hose skinning is only
half the battle. Additionally, non-linear volume changes can be observed in living
creatures as their skin collides with underlying anatomy.
This other type of volume change (we'll call it anatomical collision ) can be
approximated by simply adding more joints. These muscle joints are then manu-
ally animated to bulge and flex. While this is a common contemporary method,
there are two major issues with it:
The additional joints needed to simulate muscle bulging can severely clutter
the animation skeleton. It becomes much more time consuming to balance
the weighting between limbs and muscles.
These additional joints only provide bulging deformations. They com-
pletely lack the ability to give the impression of bone sliding under skin
(the bug-under-the-rug effect). This section describes a way to model this
behavior in a semiprocedural way.
13.4.2 Interior Collision Geometry
To capture the skin sliding over bones/muscles behavior, we start by creating some
geometry to represent the muscles and bones. Then after applying the skinning
algorithm, we loop through each vertex and check whether or not a collision has
occurred. If it has, we simply push the skin vertices outwards along their normal
such that they envelop the entire model.
Modeling anatomy is not a typical requirement of character rigs, but it's essen-
tial if we want to simulate this effect. It's never necessary to have a biologically
correct representation of muscles and bones. This is where artistic sensibilities
become integral to the successful implementation of an advanced character rig.
Typically, major muscle groups can be safely approximated by a single sur-
face. It may even be sufficient to model the radius/ulna as a single capsule. Having
the ability to collide with an arbitrary triangular mesh will enable artists to create
any desirable sub-skin collision shape.
That said, implicitly defined surfaces work well here and may be significantly
faster to compute collisions against. These collision shapes should be parented
to the skeleton, and their transformations must be updated prior to calculating the