Validation of Predictive Dynamics Tasks - Human Motion Simulation: Predictive Dynamics - page 213

coefficient of determination R 2 , the degree of correlation between the predicted
and experimental data at the distinctive key frames. All benchmarks are consid-
ered valuable because substantial effort could be saved if the predicted motions
could be configured during the early benchmark tests.
The validation methodology performs each benchmark test by comparing the
predicted motions against the average motions obtained by performing experi-
ments for the same task. One critical issue that needs to be considered in the vali-
dation process is the inconsistency between the experimental data and the
predicted data spaces. The experimental 3D displacement data are expressed in
the Cartesian space, while the predicted dynamic tasks are expressed in the joint
space. Therefore, the experimental data are first transformed from the Cartesian
space to the joint space using a global optimization-based IK scheme ( Lu and
O'connor, 1999 ; Xiang et al., 2011 ).
Due to the large amount of information in the resulting motion, the validation
process becomes cumbersome, and therefore it only considers the determinants of
the motion. Furthermore, the validation methodology selects a more restrictive
subspace of these determinants at distinct key frames. The distinct key frames
represent the magnitude of the determinants at well-defined times in the determi-
nants' time history. For a lifting task, for example, the magnitudes of the knee
angle at 0%, 20%, 40%, 60%, 80%, and 100% of the lifting height can be used as
distinct key frames. In addition to validating the kinematics of the motion, the
validation methodology also checks the kinetics of the motion.
Two tasks will be validated in this chapter to demonstrate the method, process,
and accuracy of PD.

9.5 Validation of predictive walking task
9.5.1 Walking task description
Following the ROM, the subjects performed the normal walking trials. The nor-
mal walking trial was set up so that the subjects walked continuously back and
forth across the laboratory. The first and final steps of each pass were considered
acceleration and deceleration, and the middle stride was analyzed as a steady-
state gait cycle for normal forward walking.
The normal walking trial was set up so that the subjects walked forward at a
comfortable speed, stopped, and walked backwards to return to the starting posi-
tion. On average, the subjects took 4

5 steps forward depending on preferred

step length.
The trials were repeated to ensure that the results were usable and to examine
the consistency of the velocities chosen by the subjects. In addition, every trial
began and ended with the subject in the same position—the T-pose described ear-
lier in this chapter. The reflective markers seen on the subject in Figure 9.6 corre-
spond to the Santos s marker placement protocol.