IEEE Consumer Electronics Magazine - July 2017 - 31

robots and its capability for localizing and attending to the preferred object. In one experiment, the robot is tested with a single
object, while in the second experiment two objects are shown.
The results show that in 20 runs of the network with different
parameters, 14 resulted in the correct object being selected.
This model has been further extended to add a languagelearning component. Taking inspiration from a neuroanatomical
model of visuomotor cell assemblies, the attention model was
extended by adding an additional layer for the auditory cortex.
This new layer receives as input speech signals, which can, for
example, correspond to the name of the object or its properties
(e.g., color). The extended architecture has been tested for
object naming experiments, with the use of the STDP rule for
the learning of vision-speech association [47].
The neurorobotics approach, and specifically the latest developments in the integration of the robot platforms with neuromorphic systems, offers a novel approach to the design of fully
autonomous systems. In addition to the design of robots capable
of learning to use brain-like control systems, other advantages of
systems like SpiNNaker are, for example, efficiency in energy
consumption, which is a key factor in the design of fully autonomous robots.

BCI INTERFACING ROBOTICS
The previous sections lead us to a short discussion and clarification of the results obtained through the connection of BCI
with robotic systems, especially in assistive robotics [49]. The
recent reviews [50], [51] already indicated the existence of
products that interface BCI systems with robotic environments, especially in the field of entertainment (e.g., Neural
Impulse Actuator, Star Wars Force Training, and Mind Flex).
One of the early works in this regard has been permeated with
the objective of mechanically rebuilding prehensile capacity
through an analysis of the EEG frequencies that are implied in
grasping activities. The primary motor cortex, PFC, sensorymotor cortex, and visual-motor cortex are involved in grasping
movements, but the neural characterization at cortical EEG
activities is poorly coded in these areas. Therefore, the decoding of the kinematics of grasping has aroused great interest,
especially for the construction of a BCI that could extrapolate
the intention of movement and that could interface and control
in such fashion an external device, especially humanoid robotics systems [52].
In this way, humanoid robots can become an active user's
biofeedback. However, such a process is not simple because
the kinematic space, given by the joint angle and the speed of
the movement synergies, is encrypted by potential cortical
fields. For example, in Agashe's work (described in the section
"Neuroprosthetics and BCI"), the subjects had to grasp various
objects, during a BCI recording, with their natural right hand
and through a robotic glove that specifically caught the trajectories by 18 joint angles.
Other recent works, like Chae's research, describe humanoid robots actuating by BCI [53]. These studies developed
humanoid robotics systems that are able to be guided through
mind activity (EEG rhythms). These systems provided low-

These innovative studies
demonstrate the possibility that a
subject can control a robot by
synchronizing mentally in a
human-machine interaction system.
level motion commands (e.g., right, left, and forward) by combining the classification of three motor imagery conditions
(i.e., right hand motion, left hand motion, and foot motion). A
similar study on a hybrid robotics system [54] describes a lowcost interface that allows users to control navigation and space
investigation. This process is actuated by a humanoid robot
that is able to recognize the chosen object by following BCI
signals based on SSVEP and ERD/ERS.
It remains to be verified whether such a hybrid approach
can be adapted to persons with severe disabilities who are in a
condition to imagine the motor gesture. In particular, the latest
research suggests that the imagery and motor training could
restore the control of neural areas that are able to perform prehensile tasks [55] and required to interface robotics prosthesis
[56]. According to this prospective, Chella's research [57] is
dedicated to patients with ALS and others neuromuscular diseases. In this interesting work, Chella et al. developed a museum robotics guide interfaced with a BCI system. These
innovative studies demonstrate the possibility that a subject
can control a robot by synchronizing mentally in a human-
machine interaction system. Such results highlight that hybrid
interfaces are highly suitable for people with physical disabilities, e.g., from the motor control of a wheelchair to the control
of a humanoid robotic system to assistive robotic systems,
which are also useful in domotics solutions.

CONCLUSION
This review, in addition to introducing the latest technologies
developed within the fields of human-robot interaction and
neuroprosthetics, shows that highly interdisciplinary
approaches, such as those related to perceptron studies, computational neuroscience, robotics, and cognitive science, converge toward the building and remodulation of a functional
process. In this human-robot interaction, of particular relevance is the concept of proxemic space [58], which is the personal space that people keep around themselves and which
becomes a cognitive and behavioral extension of the subject.
Here, the function of space and the use of proxemics through
technology [59] become an extension of the user. So, the
users via such interaction technologies can enhance their
effector system, adapting it to the central system person. This
happens without the need to ever define an end to remodeling
and learning feedback between the human system and the
effector or robotic system.
Furthermore, this interaction among cognitive neuroscience,
engineering, robotic systems, and neuroprosthetics becomes
july 2017

IEEE Consumer Electronics Magazine

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - July 2017