154 Environments of intelligence
robots to be discussed in the next section will typically be), but with the aim of
being genuinely anthropomorphic in its functional architecture. Relevant features
of the human being as an organism are replicated in reasonable functional detail.
Such a robot, like evolutionary robots and all other robots built under the paradigm of behaviour-based AI, will not have to be fed with explicit models of the
environment or the situations in which to act. If such explicit models were used,
any attempt at sensu-motoric integration would lead to combinatorial explosion.
Instead, a robot will learn to cope with its environment, partly by observational
learning and partly by refining its skills in the course of executing some task,
where that learning is guided by a set of basic semantic rules, allowing the robot to
infer higher-order regularities in the activities it observes (Ramirez-Amaro et al.
2015). Skills of communicating or interacting with human beings as intentional
beings are not part of this endeavour.
A key notion employed by, and embodied in, cognitive robotics is that of
“sensory-motor maps”, to be understood as the mapping of cortical structures
onto certain areas of the body and their sensu-motoric properties (Cheng 2014;
Mittendorfer and Cheng 2011). What makes this notion relevant is the observation that processing of sensory input and motor output is not organised along
sensory modalities (visual, tactile, etc.) but as an integrated cortical representation
of the information provided by various modalities with respect to the body regions
involved in an activity. Motor maps also account for the phenomenon that similar
brain areas will show similar activation patterns when a subject A observes subject B’s activities (Chaminade and Cheng 2009). This specific property of motor
maps is cited as the reason why it was possible to transfer an entire motor map
dataset developed by one robot to another not exactly identical but similarly built
robot who adapted that motor map within a day. In an earlier experiment, a monkey was able to control the walking of a remote robot without walking himself
(Nicolelis and Chapin 2002).
A prima facie similar notion employed by cognitive roboticists is that of a “body
schema” (Cheng 2014; Wieser et al. 2011), which goes back to Head (1920) and
is to be understood as dynamic representation of one’s posture – the shape and
relative position of one’s limbs – used for the spatial organisation one’s activities.
Unlike motor maps, however, a body schema does not make reference to cortical
mapping, and hence may be organised differently. A motor map is the set of cortical structures in which one’s body parts, their sensations and their movements are
organised, whereas the notion of a body schema is genuinely phenomenological,
qua being a representation of a perceiving subject’s posture of which he or she is
supposedly aware, and which he can readily use in the organisation of action. Notably, one’s body schema can be modified or “extended” by using tools. An artefact,
once one is trained in its use, becomes part of his or her body schema. According
to one experiment, artefacts such as additional (rather than prosthetic) artificial
limbs, connected via brain-computer interfaces, can be used roughly as naturally as
natural limbs. Most notably, a robotic exoskeleton was used by a paralysed person
to accomplish the symbolic kick-off at the 2014 FIFA World Cup Brazil, as a demonstration of the accomplishments of the “Walk Again Project” (Lin et al. 2014).
robots to be discussed in the next section will typically be), but with the aim of
being genuinely anthropomorphic in its functional architecture. Relevant features
of the human being as an organism are replicated in reasonable functional detail.
Such a robot, like evolutionary robots and all other robots built under the paradigm of behaviour-based AI, will not have to be fed with explicit models of the
environment or the situations in which to act. If such explicit models were used,
any attempt at sensu-motoric integration would lead to combinatorial explosion.
Instead, a robot will learn to cope with its environment, partly by observational
learning and partly by refining its skills in the course of executing some task,
where that learning is guided by a set of basic semantic rules, allowing the robot to
infer higher-order regularities in the activities it observes (Ramirez-Amaro et al.
2015). Skills of communicating or interacting with human beings as intentional
beings are not part of this endeavour.
A key notion employed by, and embodied in, cognitive robotics is that of
“sensory-motor maps”, to be understood as the mapping of cortical structures
onto certain areas of the body and their sensu-motoric properties (Cheng 2014;
Mittendorfer and Cheng 2011). What makes this notion relevant is the observation that processing of sensory input and motor output is not organised along
sensory modalities (visual, tactile, etc.) but as an integrated cortical representation
of the information provided by various modalities with respect to the body regions
involved in an activity. Motor maps also account for the phenomenon that similar
brain areas will show similar activation patterns when a subject A observes subject B’s activities (Chaminade and Cheng 2009). This specific property of motor
maps is cited as the reason why it was possible to transfer an entire motor map
dataset developed by one robot to another not exactly identical but similarly built
robot who adapted that motor map within a day. In an earlier experiment, a monkey was able to control the walking of a remote robot without walking himself
(Nicolelis and Chapin 2002).
A prima facie similar notion employed by cognitive roboticists is that of a “body
schema” (Cheng 2014; Wieser et al. 2011), which goes back to Head (1920) and
is to be understood as dynamic representation of one’s posture – the shape and
relative position of one’s limbs – used for the spatial organisation one’s activities.
Unlike motor maps, however, a body schema does not make reference to cortical
mapping, and hence may be organised differently. A motor map is the set of cortical structures in which one’s body parts, their sensations and their movements are
organised, whereas the notion of a body schema is genuinely phenomenological,
qua being a representation of a perceiving subject’s posture of which he or she is
supposedly aware, and which he can readily use in the organisation of action. Notably, one’s body schema can be modified or “extended” by using tools. An artefact,
once one is trained in its use, becomes part of his or her body schema. According
to one experiment, artefacts such as additional (rather than prosthetic) artificial
limbs, connected via brain-computer interfaces, can be used roughly as naturally as
natural limbs. Most notably, a robotic exoskeleton was used by a paralysed person
to accomplish the symbolic kick-off at the 2014 FIFA World Cup Brazil, as a demonstration of the accomplishments of the “Walk Again Project” (Lin et al. 2014).