IEEE Robotics & Automation Magazine - December 2020 - 52

and the relation between the bending angle and the overpressure in the air chambers for downward and upward bending,
can be found in Figure 6. Because of the hyperflexibility, the
direction of the gravity field has an influence on the bending
characteristic. This highlights the importance of adding feedback to the controller of these soft actuators. The model-free
feedforward controller used in this article is insufficient
when these finger actuators are used in industrial and commercial applications because the response characteristic of
the actuators changes when orientated differently.
Feedback control requires the integration of one or more
sensors in the actuator. To avoid losing the desired flexible characteristic of the soft actuator, the integrated sensor will have to
be soft, as well. In general, in many soft robots, feedback control
is required, hence the recent development of a new field in
robotics: " soft sensors. " In future work, soft sensors and feedback controllers will be added to the SH soft actuators. Five
identical actuators were manufactured out of DPBMFT5000-r0.5. By placing them in one soft hand assembly, we
demonstrated the BSPAs' usability for social soft robotics applications (Figure 7). The five fingers can be separately controlled,
which enables the hand to perform simple gestures that can be
employed in social soft robots to express emotions.
Validation of the Healing Ability in the Soft Hand
The healing ability of the BSPAs was demonstrated by
applying macroscopic cuts all the way through the soft

Figure 7. The five BSPAs can function together in a soft hand,
illustrating the potential of autonomously healable DA materials
for social soft robotic applications. The fingers can be controlled
separately, which enables the hand to make simple gestures
(a video is available at https://youtu.be/ejbIcpHqplU).

IEEE ROBOTICS & AUTOMATION MAGAZINE

DECEMBER 2020

membranes at different locations on the actuator (Figure 8).
For these tests, a clean blade was used. The first cut (with
a length of 12 mm, extending all the way through the material) was made in the thick bottom sheet, perpendicular to
the longitudinal axis of the noninflated actuator [Figure 8(a)]. When the blade was taken out, the elastic
response of the DA material pushed the cut surfaces back
together. The actuator was left untouched for only 30 s,
after which it was inflated. After 30 s of healing at room
temperature, the actuator was airtight and could resume its
activities. The airtightness was confirmed when the actuator
was submerged in water. No air bubbles escaped through the
membrane during actuation across the full bending range.
From the previous material tests [Figure 4(a)], it is known
that the cut is far from fully healed after only 30 s. However,
during actuation, the freshly healed cut perpendicular to the
longitudinal axis is compressed, increasing contact. Thus, the
instantaneous adhesion that relies on secondary interactions
and the very low number of covalent DA bonds is sufficient to
keep the actuator airtight.
The actuator is fully recovered when the cut is healed and
remains completely airtight in the actuation range of the actuator, while the actuator performance is recovered. As a result,
for this particular damage, the actuator is fully healed after
only 30 s. This illustrates very well the dependence of the
required duration of the healing process on the size of the
damage and, even more, on the location of the injury. After
only 30 s, in this case, the actuator can be reused. During further operation, the interfacial strength of the " scar " increases
progressively because the number of interfacial DA bonds will
gradually grow. Eventually, all microscopic cavities at the scar
will be completely sealed, and the strength at the location of
the scar will be recovered.
In a second test [Figure 8(b)], a cut with the same size was
made in the thick bottom sheet but along the longitudinal
axis of the actuator. During inflation, the stresses on this cut
are larger, and 30 s of healing time was not sufficient to make
the actuator airtight again. For this cut, which had the same
dimensions but another orientation, a healing time of 2 h at
room temperature was required. In the first test, after 30 s, we
can say only that there is enough adhesion to keep the cut
closed, as we do not know whether DA bonds already play an
important role. In the case of the cut along the longitudinal
axis of the actuator, the stresses on the scar during actuation
are higher, and it is clear that covalent bonds are needed to
keep the actuator airtight (physical adhesion is not enough).
After 2 h, enough interfacial DA bonds were formed to keep
the part airtight.
A third cut [Figure 8(c)] was made in the thinner top membrane of one of the rectangular cells. In this membrane, the
stresses during actuation are higher than the ones in the bottom
layer. This translates to a longer healing time of 16 h because
many more DA bonds must be formed across the cut surfaces
to ensure sufficient interfacial strength to keep the actuator airtight. However, even this damage could be healed without the
need of a heat stimulus and in a relatively short time of 16 h.

https://youtu.be/ejbIcpHqplU

IEEE Robotics & Automation Magazine - December 2020

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2020