IEEE Robotics & Automation Magazine - March 2017 - 52

Figure 5. The Pisa/IIT SoftFoot is a passive deformable foot
designed to adapt to uncertainties and walk on uneven terrain.

provides the foot the abilities to store energy, absorb impacts,
adapt, and stabilize the body [17].
Add-Ons
One of the main advantages of using RP is related to the possibility of building and operating a whole robotic system in a
few hours. For this reason, another relevant part of NMMI
consists in the group of add-ons specifically designed by users
for one or more tasks. The add-ons database includes a series
of elements that can be combined with Qbmoves and soft
EEs, thanks to the standardized interconnection system. In
this database, it is possible to find all the ancillaries that allow
for the creation of soft robotic interacting systems of arbitrary
complexity. All of them have to be designed to be mechanically interconnected with the other components. Figure 6 shows
some examples.
Interconnection Layer
The interconnection layer provides a standard mechanical
interfacing system, a highly customizable family of electronic

(a)

(b)

boards, and a software layer that lets users freely combine,
customize, and interface with NMMI modules. On the
mechanical side, flanges allow the intuitive assembly of
NMMI modules to form different kinematic chains. This is
possible by properly using the flanges of different shapes, as
shown in Figure 7. Each Qbmove is housed in a box of cubic
shape. This regular shape allows for the interconnection of the
actuator with other modules in any direction, thanks to the
grooves provided on each edge. We fixed the cube edge
lengths low, i.e., 66 mm, with the idea of further increasing
the manageability of the modules. Similar interfaces are also
embedded in the soft EEs.
With the idea of making all devices self-contained,
NMMI includes custom-made electronic boards. They are
presented in Figure 8 and are used in all the modules, with
the exception of the Qbmove Maker, which uses an Arduino
board. Thanks to such electronics, multiple Qbmove units
can be connected in series, using different identifiers with a
daisy-chain topology. In Figure 8, the circuit uses a Cypress
PSoC 3, communicating with the external personal computer through a micro universal serial bus port. An input/output set of ports is included in the boards and is used to
retrieve data from magnetic encoders and to supply power
to the dc motors. Current-sensing circuits are included and
used to enable the user to estimate the power consumption,
which, in turn, can be used to estimate the torque/stiffness
behavior of the device.
NMMI also comprehends a set of software libraries. A first
layer is developed in C and allows the setting of basic parameters, reading sensors, and writing commands to NMMI modules. The second layer is aimed at parsing first-layer functions
to the most commonly used robotics software, such as matrix
laboratory (MATLAB)/Simulink (www.mathworks.com) and
Robot Operating System (ROS) (www.ros.org). Dynamic
simulators for the Qbmove and soft grippers for both MATLAB/Simulink and ROS are also provided.
User Experiences
NMMI is a project in constant evolution. Novel setups and
components are continuously thought up and added to
the platform through its web portal [22]. We believe that this

(c)

Figure 6. Some examples of the add-on pool: (a) a one-shell head that holds a 3-D camera actuated by two Qbmoves implementing
roll-and-pitch neck DoF, (b) a pincer consisting of two rigid parts actuated by a Qbmove, providing a variable-stiffness pinch grasp,
and (c) a paw designed to realize a hexaped and to support it in the lifting and walking phases.

IEEE rOBOTIcS & aUTOMaTION MaGaZINE

March 2017

http://www.mathworks.com http://www.ros.org

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2017