Biomedical Engineering Reference
In-Depth Information
1998). An experiment was conducted in which,
a volunteer stood on a pair of force platforms
and voluntarily rocked back and forward slowly
in time with a metronome. CoM was measured
using a 6-camera Vicon system and full-body
BodyBuilder model (Eames et al., 1999). By the
above logic, one would expect to see a phase (time)
delay, probably around 150-250ms, between CoM
and CoP since the motion is consciously gener-
ated by the volunteer. However, the experimental
results show no time delay between the signals
of CoM and CoP.
One reasonable explanation may be that, if the
volunteer is conscious on his motion from quiet
standing, it seems there should exist some phase
lag because of the conduction delay from CNS;
otherwise his behaviour may be controlled by
some mechanism like CPG, a biologically plau-
sible, preprogrammed pattern organ or equivalent
structure most probably located in the spinal
cord, in which case the signal conduction phase
lag may be trivial. Based on this assumption,
the CPG model introduced in this chapter will
be especially valuable for retrieving biological
patterns as we expected.
As we described, this chapter provides a
general
architecture and coherent methodology
on gait modelling. Some additional modifications
may, however, be expected if this method is used
for a particular development of gait related issues.
For example, on the design of a robot with each
leg having multiple degrees of freedom, one may
consider the arrangement of an OBB net at each
joint and make connections among all OBB nets.
A more detailed CPG architecture could be con-
ceived in this
way,
nevertheless it would be still
based on
the general asymmetric Hopfield-like
neural network under SMER.
prospects on the modelling process and its ap-
plications. One of the possible novel works is to
determine the dynamic features of OBB models
in a more detailed mathematical framework. An
immediate example would be the investigation
of system stability by analysing the symmetry
in the system parameter matrices.
Another interesting direction is that of the
general methodology of asymmetric Hopfield
neural network combining SMER can be applied
and improved. One of the goals of this study is to
provide a
reconfigurable network structure using
simple and composite OBBs. A specific coupled
CPG architecture should
be created according
to different specimen under research. Generally
speaking, the more detailed information is avail-
able on a locomotor CPG architecture, such as
the number of neurons or coupling situation, the
more insights can be fed back for a better model
implementation, such as the better error tolerance.
For instance, the bipedal gaits can be retrieved
with a very simple model consisting of only four
macroneurons for the relaxation oscillation of two
legs. However, it is possible and better to work
out a much more detailed model consisting of
more macroneurons coupled with the different
weights. In this case a bipedal CPG architecture
may have macroneurons for its toes, heels, ankles,
knees, back and even two arms respectively, with
the stronggest coupling strength between a pair
of toe and heel while the weakest one between a
pair of arm and knee. This more detailed model
is expected to be more physiologically reasonable
in the sense that, intuitively there exist some weak
coupling relations between a pair of arm and knee
during human walking, and no great impact on
locomotion as two arms are amputated except that
some kind of instability is introduced.
Another direction for future research of CPGs
is in the field of sensorimotor integration. The
proposed model requires adaptive legged locomo-
tion; thus the pattern generation parameters must
undergo continuous modulation by sensory inputs.
The study of insect sensorimotor systems, particu-
FUTURE RESEARCH DIRECTIONS
As a newly
proposed approach to CPG design
and biological computation, one could see some
Search WWH ::
Custom Search