Biomedical Engineering Reference
In-Depth Information
4.2.3 Biomimetic and Bio-Inspired
Signal Processing
Probably the most widely used biomimetic
methods of signal processing are artificial neu-
ral networks, which are mathematical mod-
els of the manner in which the biological- and
cortical-based neural networks function in the
human body and the brain, respectively. These
are a well-known class of tools that have been
applied to a wide range of engineering problems
well beyond the field of robotics. Another class
of methods is based on the concept of fuzzy
logic, which was proposed in the 1960s to cap-
ture aspects of human possibilistic reasoning. It
must be said at the outset that, in spite of these
developments, researchers are still interested in
biomimetic and bio-inspired signal-processing
methods because most biological methods still
excel in real-time and real-world perception
and control applications, input data processing
from distributed arrays of sensors, and holis-
tic pattern recognition, compared with several
advanced functionally like-methods in engi-
neering systems.
From the concepts of learning and evolution
in biological species, the field of genetic algo-
rithms, capturing a host of Darwinian and
Lamarckian evolutionary mechanisms, has not
only arisen but is used in many engineering
applications. See Chapter 17 by Banzhaf on evo-
lutionary approaches in computation. From the
collective behaviors of elementary entities, new
concepts of swarm intelligence have been iso-
lated, leading to new methods of genetic optimi-
zation such as particle-swarm optimization and
ant-colony optimization.
Typically, when a biological sensor such as the
ear or the eye encounters a signal, the signal goes
through a range of processing steps before the
information is acted upon. In an engineered sys-
tem also, the signal undergoes a host of signal-
processing steps. These steps are initiated by
acquisition of the signal and followed by either a
sequence of steps leading to signal identification
Muscles can only exert a force while contracting.
On the other hand, a muscle can allow itself to
be stretched, although it cannot exert a force
when this happens. Thus, there is a need for a
pair of muscles, the agonist providing the force
by contraction while the antagonist stretches in
acting against the agonist. Thus, the antagonist
is responsible for moving the body part back to
its original position.
An example of this kind of muscle pairing is
furnished by the biceps brachii and triceps bra-
chii. When the biceps are contracting, the tri-
ceps are stretching back to their original
position. The opposite happens when the tri-
ceps contract.
The agonist and antagonist muscles are also
referred to as (1) extensors when the bones move
away from each other as the muscle contracts
(triceps), and (2) flexors when the bones move
toward each other as the muscle contracts
(biceps). Combined with the concept of con-
trolled compliance, the resulting biomimetic
actuators generally feature controllable stiffness.
The control of stiffness and, consequently, the
control of compliance can be achieved by means
of passive components or by employing feed-
back and active components.
A hybrid combination of both active and pas-
sive compliance is referred to as
semi-passive
compliance
and generally features the advan-
tages of both actively compliant and passively
compliant actuators. Whereas actively compli-
ant actuators use an unlimited power source
and passively compliant actuators use no exter-
nal power source, the semi-passive compliant
actuators use a limited power source, thus facili-
tating a limit variation of the stiffness. These
actuators are designed using smart materials
such as EAPs and SMAs, whereas their active
counterparts use electrohydraulic and electro-
pneumatic actuators. On the other hand, pas-
sively compliant actuators use passive
components only, and the variation of stiffness
in these actuators is achieved by increasing the
mechanical complexity of their design.
