|
Results:
-
Extension of the Modeling Field
Theory neural network for the classification of objects (as
seen in ICANN 06)
the system can be applied to the
classification of action patterns in the context of cognitive
robotics view results
Description:
This consist of the
7 main data (X, Y, Z, and rotations of joints 1, 2, 3, and 4)
for each of the 6 segments of the robot’s arms (right shoulder,
right upperarm, right elbow, left shoulder, left upperarm, left
elbow). As training set we consider 5 postures: resting position
with both arms open, left arm in front, right arm in front, both
arms in front, and both arms down.
Click for full size image
Figure 1: Evolution of fields
in the robot posture classification task. The value of the field
corresponds to equation (7). Although the five fields look very
close, in reality the individual field values match very well
the 42 parameters of the original positions.
Up -
Back
|
