Biomedical Engineering Reference
In-Depth Information
larly recent interest in the functional architecture
of the insect brain, point towards a hierarchical
organisation of behaviour (Webb, 2001; 2002;
2004). The basis of behaviour is the direct reflex.
In the context of walking, direct reflexes utilise
tactile and proprioceptive information to make
immediate adjustments to leg movements. While
it has been argued (Cruse, 1991) that local reflexes
alone may be sufficient to generate co-ordinated
insect gaits, there is also evidence for CPG involve-
ment in walking, and the interaction of these two
mechanisms is an interesting problem to explore.
Most “insect-inspired” robotics to date has been
limited to exploration of reflex control, but in
fact there is clear evidence that insect behaviour
is modulated by secondary pathways that carry
out more sophisticated sensory processing such
as adaptive classification and associative learn-
ing. An intriguing hypothesis is that one of the
main secondary pathways, which passes through
the highly organised brain structures known as
mushroom bodies, effectively implements a func-
tional equivalent of some learning algorithms. The
learning system can underpin the acquisition of
anticipatory behaviour. For instance, the learning
can capture the visual depth stimuli that predict
an impending tactile or proprioceptive stimulus.
A learning system can thus make an appropriate
motor response, to generate a reflex and (or) CPG
modulation of the walking pattern.
The model is reconfigurable in the sense that it is
easy to modify the sharing resources within the
OBB modules to change the connection strength
between motor neurons, thus to make different gait
patterns for a smooth gait transition. The model
benefits from the modular OBB modules which
are amenable to VLSI implementation.
In principle, two research trends will co-ex-
ist in the future on analysing and modeling the
biological rhythmic pattern generation mecha-
nisms. The low level, traditional approach uses
the stringent mathematical methods such as
dynamical system theory to describe and solve
the phenomenological problems (Wilson, 1999;
Izhikevich, 2007). This approach usually focuses
on the microscopic, cellular level aspects of pattern
dynamics, with emphasis on the onset, develop-
ment, propogation, transition and stability of the
activity of a neuron population.
Another is a high level, macroscopic approach
dedicated to mimicking the observable motion
of the animals. This approach usually involves a
mathematical description of the target prototype
geometry and a simplified neural network model
(Golubitsky et al., 1998; Yang & França, 2003)
to produce the same dynamical results with the
advantage of modularity and quick development
without any loss of accuracy. It is a viable bio-
inspired methodology to implement a biological
dynamical system. For instance, we have used
the SMER-based OBB modules to integrate, in
a single model, the axial and appendicular move-
ments of a centipede in its fastest gait pattern of
locomotion (Braga et al., in press). The similar
motor mechanism can also be generalised and
applied to the swimming animals such as the
lamprey.
CONCLUSION AND FUTURE
TRENDS
After a brief review of the history of studies on
the legged locomotion and state of the art on
computational models of CPGs, we present a
simple CPG architecture to generate a full range
of rhythmic patterns for a specific legged animal.
This general CPG model features some attractive
properties. The model is a purely distributed
dynamical system without the need of a central
clock due to the adoption of the OBB modules.
ACkNOWLEDGMENT
This research has been supported in part by a
Brazilian CNPq grant No.143032/96-8, a Brit-
ish BBSRC grant BBS/B/07217, an EPSRC
Search WWH ::
Custom Search