Overview
Forward Kinematics (FK) and Inverse Kinematics (IK) are standard tools used
by animators to set joint angles for character animation. FK allows
animators to explicitly set the angles while IK allows the animator to
choose a target end position and the IK system solves for a set of joint
angles that reach the end-effector. Most IK systems can solve for joint
angles under a given set of constraints. For example, the animator may
constrain an elbow joint to lie between 45 and 180 degrees.
Current IK systems solve for constrained angles along
a single chain. It is sometimes necessary, however, to
constrain joint angles along different chains. A very
familiar example of such a structure is the human hand.
The hand's anatomical and neurological structure create
an intricate system of inter-joint dependencies. Try,
for example, to bend your ring finger without moving any
other finger.
Another shortcoming of current IK systems is that a multiple-chain setup
contains multiple goals that are each associated with a single end-effector,
and thus solved for sequentially, rather than simultaneously. In other
words, a single chain and goal combination have no context of the other
chains in the system.
We therefore aim to build a model able to
capture the subtle intricacies of the human hand in motion. The model
should be able to display realistic animations of the human hand performing
a dexterous task. Multiple-goal reaching targets are exemplefied in many
typical hand manipulation tasks, and excellent example of which is playing a
musical instrument. Thus we also present a flexible system for generating
target fingering positions from guitar tabulature.
Acknowledgments
We would like to acknowledge the hardware and software contributions
from NDI and
Side Effects Software.
This research was partly funded by NSERC
|