Posture Prediction and Optimization - Human Motion Simulation: Predictive Dynamics - page 42

calculating the joint variables of a lower extremity to allow the foot to be posi-
tioned on a pedal, for example, is also a posture prediction problem.
For a posture prediction problem, the design variables are the joint angles
of the body. The constraints are the location and possibly the orientation of an
end-effector (usually the fingertip) and the joint ranges of motion. Commonly,
optimization-based inverse kinematic methods focus on minimizing some form
of discomfort ( Abdel-Malek et al., 2004b; Jung and Choe, 1996 ) or minimiz-
ing perturbation from a neutral position (Case et al., 1990; Porter et al.,
1990). If more than one objective function is used, they can be combined
using multi-objective optimization techniques like the objective sum method,
the min max method, or the global criterion method (Yang et al., 2004a d).
Other objective functions that have been tried include minimizing joint
displacement, minimizing change in potential energy, minimizing the distance
to the target, and maximizing reachability ( Abdel-Malek et al., 2004a,c ;Yang
et al., 2004a d).
Optimization can be used not only to find the most realistic posture to reach
a point in the workspace, but also to find the best placement of the human
within a workspace or the best placement of the target relative to the human
( Abdel-Malek et al., 2004a,c ; Abdel-Malek et al., 2006; Abdel-Malek et al.,
2001a d; Abdel-Malek et al., 2005; Marler et al., 2009 ; Mi et al., 2002a ; Yang
et al., 2006a,b,c,d; Mi et al., 2009 ; Yang et al., 2004a d).
Motion prediction broadens the approach and finds the optimal motion
between two target points. For motion prediction, B-spline approximations or
polynomials are used to approximate joint displacement with respect to time (Mi,
2004). Furthermore, additional objective functions are formulated that deal with
the change in joint displacement over time. Examples include minimizing incon-
sistency, minimizing joint acceleration, and minimizing velocity at the beginning
and end of the motion ( Abdel-Malek et al., 2004b ; Mi, 2004).
In this chapter, the treatment of an open kinematic chain is addressed as
opposed to closed-loop systems. For the human body, we consider a variety of
open kinematic chains, typically all beginning in the waist and extending to the
hand, the foot, or the head. Similarly, the kinematic chain can begin with the
ground (foot); then extend through the body to the arm, hand, and fingers. Note
that the word prediction is often used rather than calculation since predicting real-
istic human motion is not an exact science, but rather a prediction of human
behavior. For a human to touch a given point with a finger, several postures
exist

but what is the “best” way to accomplish this task?
Motion prediction refers to the calculation of joint variables with respect to
time (often called a motion profile) to enable motion from an initial configuration
to a final configuration. Posture prediction is typically associated with the calcula-
tion of static postures, which characterizes the final positions and orientations of
segmental links without much consideration of how this posture occurred. The
challenge in predicting postures and motions is evident in that there are a large
(infinite) number of solutions.

...