Modeling and Animating Human Figures - Computer Animation: Algorithms and Techniques - page 297

Once a motion has been determined to be unacceptable, it must be modified in order to bring its
comfort level back to within acceptable ranges. This can be done by initiating one or more strategies
that reduce the strain. Assume that a particular joint has been identified that exceeds the accepted com-
fort range. If other joints in the linkage can be identified that produce a motion in the end effector sim-
ilar to that of the problem joint and that have excess torque available, then increasing the torques at
these joints can compensate for reduced torque at the problem joint. It may also be possible to include
more joints in the linkage, such as the spine in a reaching motion, to reformulate the inverse kinematic
problem in the hope of reducing the torque at the problem joint.
9.3 Walking
Walking, along with reaching, is one of the most common activities in which the human form engages.
It is a complex activity that, for humans, is learned only after an extended trial-and-error process. An
aspect that differentiates walking from typical reaching motions, besides the fact that it uses the legs
instead of the arms, is that it is basically cyclic. While its cyclic nature provides some uniformity, acy-
clic components such as turning and tripping occur periodically. In addition, walking is responsible for
transporting the figure from one place to another and is simultaneously responsible for maintaining
balance. Thus, dynamics plays a much more integral role in the formation of the walking motion than
it does in reaching.
An aspect of walking that complicates its analysis and generation is that it is dynamically stable but
not statically stable. This means that if a figure engaged in walking behavior suddenly freezes, the fig-
ure is not necessarily in a balanced state and would probably fall to the ground. For animation purposes,
this means that the walking motion cannot be frozen in time and statically analyzed to determine the
correct forces and torques that produce the motion. As a result, knowledge of the walking motion, in the
form of either empirically gathered data [ 26 ][ 32 ] or a set of parameters adjustable by the animator, is
typically used as the global control mechanism for walking behavior. Attributes such as stride length,
hip rotation, and foot placement can be used to specify what a particular walk should look like. A state
transition diagram, or its equivalent, is typically used to transition from phase to phase of the gait [ 11 ]
[ 12 ][ 22 ] [ 29 ][ 52 ]. Calculation of forces and torques can then be added, if desired, to make the nuances
of the motion more physically accurate and more visually satisfying. Kinematics can be used to entirely
control the legs, while the forces implied by the movement of the legs are used to affect the motion of
the upper body [ 23 ] [ 63 ]. Alternatively, kinematics can be used to establish constraints on leg motion
such as leg swing duration and foot placement. Then the forces and torques necessary to satisfy the
constraints can be used to resolve the remaining DOF of the legs [ 11 ] [ 29 ] [ 52 ]. In some cases, forward
dynamic control can be used after determining the forces and torques necessary to drive the legs from
state to state [ 47 ] .
9.3.1 The mechanics of locomotion
Understanding the interaction of the various joints involved in locomotion is the first step in under-
standing and modeling locomotion. The walking and running cycles are presented first. Then the walk
cycle is broken down in more detail, showing the complex movements involved. For this discussion,