Concluding Remarks - Human Motion Simulation: Predictive Dynamics - page 235

effect. As shown in the box-lifting example, a higher load will cause substantially
higher torques. An enormous load will cause the human simulation system to
respond with “I cannot do this.” The PD method will simply not converge and
will be stopped.
Cause and effect is one of the most important aspects of a simulator. It is
what differentiates an analysis system from an animation system. In gait analysis,
for example, the analyst obtains data for the physical gait from the person and
then employs software tools to analyze the motion and determine its performance.
However, if this same analyst wanted to examine a different condition, the analyst
would have to return to the laboratory, conduct another experiment, and perform
another analysis. With a human simulator, changing the conditions and determin-
ing the effects of, or response to, that change can occur immediately in the soft-
ware without the need to do additional experiments.
10.1.6 Predictive dynamics uses joint space, not muscle space
One of the most significant contributions from our work over the past 10 years
has been the realization that modeling and working with joint space is effective.
Instead of addressing the resolution of forces into every muscle or muscle group,
we focused our attention on generalized variables at the joint level—namely, rota-
tion angle, angular velocity, angular acceleration, and torque. Because each joint
is represented by one or more DOF, each DOF is systematically associated with a
set of generalized variables.
Predicting motion with PD was then simplified to focus on determining these
generalized variables at each DOF. Once the predicted motion is determined,
these variables are calculated as a discretized curve (a B-spline) that governs the
shape/velocity/acceleration of the motion, but also with the capability to calculate
torques for every DOF and reaction forces where needed. It is a fast, efficient,
reliable, and robust approach to predicting human motion.
Recovery of muscle forces, if needed, can be accomplished by first combining
the torques for every DOF at the joint into a resultant component, then resolving
the resultant component torques into muscle action forces.
10.1.7 Predictive dynamics uses dynamic strength surfaces
Because we have used joints as the key ingredient in our basic formulation for
PD, it is critical to identify strength requirements and, even more importantly,
strength limits. We have shown that answering the questions “What is the limit of
this person's ability?” and “Why can't Santos s do this task?” requires a deep
understanding of joint strength limits.
As a result, over the past 10 years we have conducted a substantial number of
experiments to measure the joint-level strength, focusing particularly on torque-
angle (isometric) and/or torque-velocity (isokinetic) data, also called dynamic
strength surfaces.