IEEE Robotics & Automation Magazine - March 2016 - 21

robot could be able to execute squat
jumps and dynamic gaits, such as running trot, Pronk, and bounding gait.
The third article, "Innovation in Underwater Robots," presents a new control strategy for straight-line path
following for biologically inspired
swimming snake robots. This new algorithm is based on a line-of-sight (LOS)
guidance law and has been experimentally validated through experiments
with the snake-like robot Mamba. The
proposed path-following controller
consists of three main components. A
gait pattern controller is used to produce a sinusoidal motion pattern, which
propels the robot forward. On top of
that, a heading controller steers the
robot toward and subsequently along
the desired path. Finally, an LOS guidance law generates the desired heading
angle to follow the desired path. Because of this control strategy, the robot
Mamba was actually able to follow
straight lines: collected results showed
that the control strategy successfully
steers the robot toward and along the
desired path for both lateral undulation
and eel-like motion patterns. In addition, a second important contribution
of this article was the fact that authors
carried out a back-to-back comparison
between simulation results and the motion of the physical robot during the
experiments, with respect to fluid coefficients identification.
The fourth article, "Correcting for
Changes," approaches the challenge of
developing control architectures for humanoid robots, which are more efficient
than conventional control schemes, in
terms of computational effort and energy consumption. The innovative control
scheme introduced in this article is bioinspired and based on the biological
principle of expected perception (EP).
Standard control architectures based on
the use of internal (forward, environment, or inverse) models rely on a strict
sequentiality between perception and
action: data are gathered from the sensory system, a future scenario is predicted, and an action to execute is planned.

One problem with this approach is that
long-term predictions and action planning are time and resource consuming.
An alternative strategy is to switch off
the sensory feedback when predictions
are comparable with the current observations: this is the bioinspired principle
at the basis of EP-based control architectures. In this article, the iCub robot is
used as a test platform to validate an EPbased control architecture capable to
anticipate the movement of an external
object and move the robot's hand toward a position, where grasping can be
achieved more reliably.
The fifth article, "Torque-StiffnessControlled Dynamic Walking," presents
a general paradigm named torque-stiffness-controlled dynamic walking to analyze the behavior of bipedal robots
endowed with a controllable joint torque
and stiffness. This paradigm is used in
combination with a bioinspired centralpattern generator-based control method
to investigate the effects of joint torque
and joint stiffness on walking performance, with an additional broader goal
to provide insights into the principles of
efficient and adaptive human walking.
The sixth article, "Walking Assistance Using Artificial Primitives" is
framed in the field of wearable robotics.
The challenge here is the development
of a simplified control framework to
provide walking assistance to people affected by lower-limb impairments in
different locomotion tasks (i.e., walking, climbing/descending stairs, and
standing up/sitting down). Indeed,
people who are affected by lower-limb
impairments but who still have the
ability to step (e.g., amputees or poststroke survivors) may benefit from the
assistive action of lightweight wearable
robots that can cooperate with their
weak or impaired articulations. This article introduces a novel bioinspired
control architecture based on artificial
motor primitives for locomotion assistance. The usability and effectiveness of
the presented control architecture was
tested experimentally with healthy subjects on the robot named active pelvis

orthosis, a bilateral hip orthosis endowed with series-elastic actuators.
The last (but not least) article, "Trial
and Error" focuses the attention on the
ability of biological systems to efficiently
reuse gathered experience to improve
their behavioral
policy, thus opOvercoming
timizing the
number of rethe limitations
quired samples
of a traditional
from real environments. This
mechatronics
article explores
design and control
the use of previous experience
paradigms is a key
as environmenissue to promote a
tal local models
to improve the
pervasive adoption
movement poliof effective robotic
cy of a humanoid robot. In
systems in real-life
this article, an
applications.
existing importance-weighted
policy gradients
with a parameter-based exploration algorithm is further elaborated and
applied to two different tasks: 1) a
cart-pole swing-up task in a real-virtual
hybrid environment and 2) a basketballshooting task with a real robot in a real
environment.
All of the presented articles are just
a small set of research initiatives in the
domain of bioinspired robotic systems: indeed, there are many ongoing
initiatives dedicated to bioinspired robotic systems, such as workshops, special issues, conferences, and new
journals. All of them show that bioinspired robotics (not just control strategies) is attracting increasing interest
from different researchers (with different backgrounds from neuroscience to
material science). This increasing interest demonstrates that overcoming
the limitations of a traditional mechatronics design and control paradigms
is a key issue to promote a pervasive
adoption of effective robotic systems
in real-life applications.

march 2016

IEEE ROBOTICS & AUTOMATION MAGAZINE

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2016