Biomedical Engineering Reference
In-Depth Information
Taylor, 1993), (Panait & Luke, 2005). Regarding
the controller, it is well known that nonlineari-
ties and uncertainties are very well managed by
computational intelligence approaches like fuzzy
logic, artificial neural networks or neurofuzzy
controllers. Literature is profuse and exceeds
the scope of this chapter. The interested reader is
referred to (Harris, 1993) or (Antsaklis, 1993). The
mission planner, sometimes referred as task and
path planner in a task decomposition perspective,
has to face a complex problem: the trajectory gen-
eration for the mobile robots. Indeed, autonomous
operation of mobile robots in real environments
presents serious difficulties to classical planning
and control methods. Usually, the environment
is poorly known, sensor readings are noisy and
vehicle dynamics is fairly complex and non-linear.
Since this represents a difficult problem to solve
due to the great amount of data required to take a
decision in an effective time, it has attracted the
AI community. Three main lines of activity have
arisen within this AI researcher's community:
(1) planning-based systems, (2) behaviour-based
systems, and (3) hybrid systems. Starting with
the intention of emulate partially some features
of human intelligence, AI techniques application
to control were gradually shifting to CI based
control, which is more descriptive of the problems
approached. The first two application examples
in the following section belong to this stage of
maturing of the knowledge in path planning.
Recently, streams of research and technologi-
cal implementations are going towards the use of
bio-inspired techniques. Particularly in robotics,
the research community is trying to take advan-
tage not only of human beings' problem solving
paradigms as part of the nature, but also from
other natural systems. A great deal of activity
currently growing is coming from evolutionary
systems and immune systems. In effect, several
researchers and practitioners addressed the ques-
tion of how biological systems adapt and interact
with the environment (Abbot & Regehr, 2004),
(Forde, Thompson & Bohannan, 2004). In par-
ticular they are interested in understanding and
synthesizing complex behaviours in the context
of autonomous robotics. Artificial Evolution was
proposed as a suitable technique for the study
of physical and biological adaptation and their
harnessing. In (Nolfi & Floreano, 2000), dynamic
modular hierarchies of neurocontrollers are pro-
posed to face stability and scalability issues, in
what they called
Evolutionary Robotics
(ER). ER
is a sub-field of Behaviour-Based Robotics (BBR)
(Arkin, 1998). It is related to the use of evolution-
ary computing methods in the area of autonomous
robotic control. One of the central goals of ER is
the development of automated methods to evolve
complex behaviour-based control strategies (Nolfi
& Floreano, 2000), (Nelson, Grant, Galeotti &
Rhody, 2004). Another central goal of ER is to
design morphological models. Some examples of
these works comprises evolutionary design (Bent-
ley, P., 1999), the evolution of robot morphologies
(Lipson & Pollack, 2000), and complete structures
from simple elements (Hornby & Pollack, 2001),
(Reiffel & Pollack, 2005). Scaling up in ER relates
to the creation of complex behaviours from simple
ones. The sequential and hierarchical organization
of a complex activity in systems with many behav-
iours, such as robot's path generation (Fernández-
León, Tosini & Acosta, 2004), (Maaref & Barref,
2002), (Nelson, Grant, Galeotti & Rhody, 2004),
remains one of the main issues for biologically
inspired robotics and systems neuroscience. It
is expected that understanding such systems in
a robotic context will also shed some light into
their natural counterparts, giving for instance a
better understanding of human motor capabili-
ties, their neural and embodied underpinnings,
and their adaptive and recovery properties. It is
being a statement that stability and scalability in
robot's behaviour can be addressed by mean of
the explicit incorporation of ideas derived from
neuroscience and the study of natural behaviour
(e.g., theories of evolutionary learning, skill
acquisition, and behaviour emergence (Dopazo,
Gordon, Perazzo & Risau-Gusman, 2003). ER
Search WWH ::
Custom Search