Biomedical Engineering Reference
In-Depth Information
A GENERAL LOCOMOTION CPG
ARCHITECTURE
control the locomotion and transition of all their
gait types. We are not able to say, in our study,
that an animal with less signalling bits is a lower-
level animal since the meaning of the number of
descending CNS signals is twofold. First, even a
centipede, which is a relatively low-level insect,
has much more CNS signals than in this study,
however, these still only occupy a very small
percentage of its neural processing part. Second,
higher-level animals such as the biped have much
more neural activities at other neural processing
mechanisms, besides the CNS-CPGs controlling
signals, while lower-level animals may have less
neural processing activities besides CNS-CPGs
signals. Therefore, the number of parallel signals
of any animal model in this study is not directly
related with its neural processing complexity.
It is known that locomotion speed is defined
by both coordinated phase relation and duty
factor, which is the relative proportion that a
flexor neuron is firing in one period. As an
animal's speed increases, the firing time of the
extensor (corresponding to stance) will decrease
dramatically while the firing time of the flexor
(corresponding to swing) keeps basically constant
(Pearson, 1976). The duty factor can be modified
by changing the reversibilities of two coupled
nodes in a simple OBB module as per the coor-
dinated phase relationship.
The Architecture
A general architecture of a locomotion system
may have three components: (1) the oscillation
system as driving motor, (2) the CPGs as pattern
source, and (3) the animal's CNS as command
source. The first two parts formulate the proposed
locomotion architecture in this chapter and can
be constructed by a model with reciprocally
inhibitory relationship between the so-called
flexor and extensor motor neurons, at each leg as
driving motor and the general locomotor CPGs as
pattern source. The CPG is a core of the locomo-
tion system which converts different conscious
signals from the CNS to the bioelectronics-based
rhythmic signals for driving the motor neurons.
These motor neurons are essentially relaxation
oscillators. They form a bistable system in which
each neuron switches from one state to another
when one of two internal thresholds is met by one
neuron and back when another threshold is met
by the other neuron.
For a 2n -legged animal, the general locomo-
tion model has the topology of a complete graph
M(N,E) with ||N||=4n and ||E||= 4 C . Each of the
upper layer macroneurons drive a flexor muscle
while the lower layer ones drive a extensor muscle,
see Figure 9. In this architecture there is a con-
nection between any two macroneurons with the
strength in the range [0,1] . Different gait pattern
has a corresponding, different connection weight
set. It sounds reasonable for the gait transition
to be controlled by the biological signals from
the CNS for avoiding risk or following animal's
willingness. The size of signal flow is equal to the
number of total couplings depending on the type
of gait patterns and the number of animal legs.
For those specific models derived from the uni-
form framework, the biped, quadruped, hexapod
animals have 4, 16 and 30 bits of parallel signals
from their CNS to CPGs , respectively, in order to
Simulation of Hexapod Gaits
We have shown the remarkable potential of asym-
metric SMER-based neural network on rhythmic
pattern formation. Now we can build an artificial
CPG example from the general gait pattern model.
Some basic criteria on macroneuron interconnec-
tion will be followed.
1.
Any two macroneurons in a CPG network
should be coupled directly if their activities
are exactly out of phase.
2.
The ipsilateral macroneurons should be con-
nected to form a cyclic undirected ring.
Search WWH ::




Custom Search