Validation of Predictive Dynamics Tasks - Human Motion Simulation: Predictive Dynamics - page 230

As mentioned earlier, the shoulder performed adequately during the first half
of the lifting cycle ( Figure 9.18 ); however, the coupling between the shoulder-
clavicle motion, the stretching in the trunk, the twisting in the wrists, and the
additional motion from the lower extremity may have affected the results. For
these reasons, the graph for the shoulder extension, shown in dashed lines in
Figure 9.19 , has been divided into two segments representing the first and second
halves of the lifting cycle. Interestingly, in the first qualitative benchmark test,
most subjects did not observe any abnormality in the shoulder motion because it
is too local.
9.7 Feedback to the simulation
Validation is not only necessary but essential to improving the simulation. When
there is a major difference between the predicted motion and human motion, it is
critical that the PD formulation be refined. This is accomplished by using differ-
ent objective functions in the optimization schemes. If unrealistic motion has
occurred at a certain joint, it may become clear that a physical constraint is miss-
ing or that an additional objective function is needed to render the motion more
natural or realistic.
For example, for the box-lifting task, the subjective evaluation showed that
the whole-body motion of the predicted model ( Figure 9.20 ) looks different than
the whole-body motion of the experimental model ( Figure 9.21 ) for the key
frames 0% height, 20% height, and 40% height, which could be related to the lift-
ing strategy. In this case, the objective evaluation ( Figures 9.18 and 9.19 ) may
not have the capacity or the information to relate that to the change in the strat-
egy, because the discrepancies in the motion may show in parts of some determi-
nants, which could be considered as a deficiency in the model. This is very clear
in the objective evaluation of the determinants ( Figure 9.18 ), where all determi-
nants were inside the interval of confidence except the hip flexion and shoulder
flexion, which were outside the interval of confidence for the first 40% of the
cycle. This is a good example for showing the significance of having the objec-
tive and subjective evaluations in the validation framework.
9.8 Concluding remarks
This chapter has presented a framework to validate the results of PD. The frame-
work was applied, as an example, to a 55-DOF predictive computer human model
performing two of the most important common tasks: walking and box lifting.
Validation of human motion is not an easy process; however, it is an essential
step of formulation development. There are two purposes for validation: (a) to
ascertain that the formulation for predicting a human motion is correct, and more