Download as pdf or txt
Download as pdf or txt
You are on page 1of 35

Supplementary Materials for

3D-printed biomimetic artificial muscles using soft actuators that contract


and elongate
Corrado De Pascali et al.

Corresponding author: Corrado De Pascali, corrado.depascali@iit.it; Barbara Mazzolai, barbara.mazzolai@iit.it

Sci. Robot. 7, eabn4155 (2022)


DOI: 10.1126/scirobotics.abn4155

The PDF file includes:

Supplementary Methods
Supplementary Discussions
Figs. S1 to S23
Tables S1 to S3
Legends for movies S1 to S7

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S7
Supplementary Methods

Model
The geometry of the GRACE is obtained by a loft of the cross section along a length 𝐿 which is
half of the total length of the GRACE (𝐿 = 𝐿𝐶𝑈 /2). The loft is performed along the parabolic
curves shown in Fig. S1A: the crest, passing through point Q, the valley, through point S, and the
edge through point P. In the following we will denote such curves by ℒ𝑐 , ℒ𝑣 , and ℒ𝑒
respectively. The lengths of such curves will be denoted by 𝐿𝑐 , 𝐿𝑣 and 𝐿𝑒 . Hereafter, all the
lengths are normalized with respect to 𝐿.
The cross section is characterized by six parameters: the couples of semi-axes (𝑎1 , 𝑏1 ) and
(𝑎2 , 𝑏2 ) and the distances 𝑅1 and 𝑅2 . However, only four of these parameters are independent.
Here we assume {𝑎1 , 𝑎2 , 𝑏2 , 𝑅2 } as independent parameters, while 𝑏1 and 𝑅1 are obtained as
𝜋−𝛼 𝛼
𝑏1 = 𝑅2 cos − 𝑏2 cos
2 2
𝜋−𝛼 𝛼
𝑅1 = 𝑅2 sin + 𝑏2 sin
2 2

Once the number of pleats and the length of the actuator 2𝐿 are assigned, the goal is to find the
optimal values of independent design parameters such that the GRACE satisfies a given
requirement: here the goal is to determine the geometry of the GRACE that best performs equally
in contraction and in extension, i.e., the GRACE that exhibits the maximum 𝜌𝐶 + 𝜌𝐸 under the
condition that 𝜌𝐶 ≈ 𝜌𝐸 , and the geometries for which the maximum contraction and the maximum
extension are achieved. To find such geometries, the design parameters are treated as variables;
for each given set {𝑎1 , 𝑎2 , 𝑏2 , 𝑅2 }, 𝜌𝐶 and 𝜌𝐸 are computed as solutions of an optimization problem,
as explained in the following.

Calculation of the deformed configuration


Fig. S1B shows how the parabolic lines ℒ𝑐 , ℒ𝑒 and ℒ𝑣 are displaced when the GRACE is
subject to the action of a positive pressure. The abscissa of the central cross section shifts from
𝑥𝑚 to 𝑥𝑚∗
; hence, the contraction ratio is given by

𝑥𝑚
𝜌𝐶 = 1 −
𝑥𝑚

Conversely, in case of depressurization, the parabolic lines are displaced below their
original position, in conjunction with the extension of the GRACE; the extension ratio can be
obtained as

𝑥𝑚
𝜌𝐸 = −1
𝑥𝑚

Due to normalization, here and in the following, it is 𝑥𝑚 = 1. The computation of the


deformed configuration of the GRACE is performed by setting an optimization problem.
The objective function depends on the volume enclosed by the GRACE. In case the
GRACE is subject to a positive pressure, its volume increases; conversely, in case of
depressurization the internal volume tends to reach a minimum value. Hence, denoting the
volume by 𝑉, the objective function is
1
𝑓𝐶 (𝑿) =
𝑉

for a GRACE performing contraction, while it is

𝑓𝐸 (𝑿) = 𝑉

for a GRACE performing extension. In both cases, the minimum of the objective function must
be found. Vector 𝑿 contains the variables of the problem:

𝑋 = {𝑦𝑐∗1 , 𝑦𝑐∗2 , 𝑦𝑐∗3 , 𝑦𝑒∗1 , 𝑦𝑒∗2 , 𝑦𝑒∗3 , 𝑦𝑣∗1 , 𝑦𝑣∗2 , 𝑦𝑣∗3 , 𝑥𝑚



}
The first nine variables represent the ordinates of the points in Fig. S1A-B. It is important to
𝑥
notice that the point of coordinates ( 3𝑚 , 𝑦𝑐1 ) it is not necessarily displaced to coordinates

𝑥𝑚
( , 𝑦𝑐∗1 ), and the same holds for the other points with abscissa 𝑥𝑚 /3 or 2𝑥𝑚 /3.
3
The search is performed under geometrical constraints. Under the assumption of
inextensibility of the membrane, we require that the lines ℒ𝑐 , ℒ𝑒 and ℒ𝑣 maintain their length
constant in any configuration of the GRACE:

𝐿∗𝑐 = 𝐿𝑐
𝐿∗𝑒 = 𝐿𝑒
𝐿∗𝑣 = 𝐿𝑣
All lengths are computed approximating the parabolic curves with the broken lines
represented in Fig. S16B. The volume 𝑉 is computed considering the blue solid represented in
Fig. S1C.
Besides the set of equality constraints listed above, inequality constraints must be
introduced in the model. In any deformed configuration, it must be:

𝑦𝑐∗2 − 𝑦𝑐∗3 < 0


𝑦𝑐∗1 − 𝑦𝑐∗2 < 0
𝑦𝑐∗2 − 2𝑦𝑐∗1 < 0
𝑦𝑐∗3 − 2𝑦𝑐∗2 + 𝑦𝑐∗1 < 0

The first couple of inequalities are imposed to guarantee that ℒ𝑐∗ increases monotonically,
while the other two ensure that its slope decreases from the extremity of the GRACE to the
central cross section. Analogous expressions hold for ℒ𝑣∗ and ℒ𝑒∗ .
Another set of constraints is needed to define the relation between the curves. In fact, in any
configuration ℒ𝑣∗ remains the closest to the longitudinal axis of the GRACE, while ℒ𝑐∗ the
furthest; ℒ𝑒∗ lies between the two. Hence, we impose:

𝑦𝑣∗𝑖 − 𝑦𝑒∗𝑖 < 0


𝑦𝑒∗𝑖 − 𝑦𝑐∗𝑖 < 0
for 𝑖 = 1,2,3.
A final set of constraints is assigned to keep into account the inextensibility of the elliptical
arcs in the cross section. It has been observed that defining such constraint in the form of
equality leads to poor results in the computation of the deformed configuration. To relax the
constraint, it is introduced in the form of inequality as follows:

√𝑎12 + 𝑏12 < 𝐿∗𝑒𝑙𝑙 𝑐 < 𝐿𝑒𝑙𝑙𝑐

√𝑎22 + 𝑏22 < 𝐿∗𝑒𝑙𝑙 𝑣 < 𝐿𝑒𝑙𝑙𝑣

where the subscript 𝑒𝑙𝑙 refers to the elliptical arcs in the cross section. Moreover, the parabolic
profiles are built such that their derivative at the abscissa 𝑥𝑚 is slightly greater than zero.
Obviously, the model is valid for all the feasible geometries of a GRACE: the chosen design
parameters must be such that the corresponding central cross section is a non-self-intersecting
line. To guarantee that the problem is physically consistent, the following conditions must be
satisfied. To conform to the definition of valley and crest, it must be
𝑅1 + 𝑎1 > 𝑅2 − 𝑎2
that can be rewritten as
𝜋−𝛼 𝛼
𝑎1 > 𝑅2 (1 − sin ) − 𝑎2 − 𝑏2 sin
2 2
To avoid self-intersection in the radial direction, it must be
𝑅2 > 𝑎2
while to avoid self-intersection between pleats in the circumferential direction it must be
𝛼
√𝑅22 − 𝑎22 ≥ 𝑏2 cot
2
In these conditions, the number of pleats appears by the angle 𝛼.

Parametric analysis and selected GRACEs


The implemented model allows us to investigate how the design parameters influence the
contraction and extension ratios. We have solved the optimization problem to find the deformed
configuration both in contraction and in elongation for several GRACEs, by varying the design
parameters in the following intervals:
𝑎1 ∈ [0.1, 0.9]
𝑎2 ∈ [0.01, 0.19]
𝑏2 ∈ [0.04, 0.24]
𝑅2 ∈ [0.4, 0.9]

For our study, we have selected 18, 14, 10 and 16 equally spaced values from the mentioned
intervals, respectively, hence computing the contraction and elongation for 40,320 GRACEs. The
model is implemented in MATLAB using the built-in function fmincon in the Optimization
Toolbox to find the deformed configuration for each considered geometry.
Since the parameters space is 4-dimensional, the results presented in the following are
displayed treating two parameters as variables, while keeping constant the other two. The results
are obtained for GRACEs with 𝑅 = 0.2, being 𝑅 the radius of the extremity of the GRACE,
denoted in Fig. S1.

The surfaces in the Fig. S2A-B are obtained for GRACEs having 𝑎1 = 0.38 and 𝑏2 = 0.15.
It is possible to see that the parameter 𝑅2 plays a major role in the deformation of the GRACE,
both in contraction and in extension. Large values of 𝑅2 enable the GRACE to perform large
extension, especially for small values of 𝑎2 : the longer ℒ𝑣 , the more the GRACE can extend
under a negative pressure difference applied. As expected, such combination of values is not
suitable to build a GRACE with high contraction ratio: Fig. S2A shows that 𝑥𝑚 ∗
reaches the
minimum (hence, 𝜌𝐶 is maximum) for a relatively low value of 𝑅2 (~0.5) and for the highest
value of 𝑎2 . As already mentioned, those features that promote the contraction of the GRACE
penalize the same GRACE in extension, and vice versa.
While 𝑅2 and 𝑎2 turn out to be the only parameters that influence the extension ratio
significantly, the contraction ratio shows to be affected by 𝑏2 as well. From Fig. S2C it can be
seen that to maximize the contraction ratio, small values of 𝑏2 are preferred if a small 𝑎2 is
adopted; moreover, small values of 𝑅2 enable the GRACE to perform a greater contraction only
for the couples (𝑎2 , 𝑏2 ) for which the surfaces exhibit a plateau, while it plays the opposite effect
outside such plateau.
However, it is important to keep in mind that the four independent geometrical parameters,
plus the number of pleats, are related in a non-trivial way. For a glimpse, the reader can observe
Fig. S3, which collects the results obtained, in case of contraction, for actuators with 4 and 10
pleats, for the same 𝑎1 , different combinations of 𝑅2 and 𝑏2 and varying 𝑎2 . Based on the
obtained results, we cannot conclude that, in general, a number of pleats should be preferred
among others; the same holds for the value of all the geometrical parameters. For this reason, the
current work mainly reports results for the particular case of 6 pleats. When a GRACE must be
selected for a given application, our suggestion is to set a priori the number of pleats, to impose
any set of constraints accounting for limitations introduced by the fabrication process, the
material used, the maximum allowed bulk of the GRACE, etc., and to use the geometrical model
to find the shape that best performs for the specific application. It is important to keep in mind,
moreover, that due to the modelling assumptions, effects introduced by the material strain are not
modelled; moreover, the computation of the deformed GRACE, either in contraction or in
extension, is based on an approximation of the enclosed volume (see Fig. S1C), which is a
source of error as well. Table S1 contains the geometrical parameters for the 6-pleats GRACEs
that best perform in contraction, extension, and in antagonistic mode. Table S2 allows to
compare the results predicted by the model for these three designs with those obtained by FEA.
As expected, the model overestimates the contraction ratio, that is worsened by the membrane
strain, and underestimates the extension ratio. Nevertheless, our model can be used to compare
different geometries and to establish which performs better in contraction/extension, provided
that they will be built with the same material, by the same process, and they are intended to
perform under the same boundary conditions. To the readers who wish to implement a more
performant and accurate model, we suggest computing the volume enclosed by the GRACE as a
volume integral: this would lead to a more accurate evaluation of the objective function,
avoiding unphysical spikes like the one in Fig. S3 for the 4-pleats GRACE with 𝑎1 = 0.665,
𝑏2 = 0.040, and 𝑅2 =0.567. However, based on our experience, such computation would increase
the computational cost of the optimization. Another improvement would consist in modelling the
mechanical strain, implementing, as a matter of fact, a quasi-static structural model, making FEA
unnecessary.

Supplementary Discussions

Role of material and thickness


Since the GRACE operating principle is based on the actuator’s geometry, it is possible to
employ a wide range of materials for their manufacturing. Through samples printed with several
materials and FEA simulations, we analyzed the contribution of the material stiffness to the
GRACE performances (Fig. S4). We observed that the material stiffness does not notably affect
the contraction and elongation ratios but highly influences the output forces. Particularly, stiffer
materials can potentially lead to larger output forces, provided that they can sustain higher input
pressures, while softer materials can operate at lower input pressures but provide smaller output
forces (and even lead to unwanted longitudinal strain, reducing contraction and increasing
elongation). The thickness of the membrane plays a similar role on the exerted forces without
altering the contraction and elongation performances. Indeed, thicker walls can sustain higher
pressures and lead to larger forces, while thinner walls accept limited pressures, resulting in
small forces. Therefore, the GRACEs allow for widely tunable operating ranges and output
performances.

Antagonistic configuration (simulation, stretching tests, wrist)


Employing FEA, we also investigated the potential advantage of the GRACE-A in antagonistic
operation compared to the GRACE-C. We analyzed a configuration with two identical actuators
in series, fixed at the two extremities, and measured the displacement of their interconnection
(Fig. S6) in four different cases: (i) two GRACE-A, one actively contracted (pressurized) and
one passive (Fig. S6A top); (ii) two GRACE-C, one pressurized and one passive (Fig. S6B top);
(iii) two GRACE-A, one pressurized and the other one actively elongated (depressurized) (Fig.
S6A bottom); (iv) two GRACE-C, one pressurized and the other one depressurized (Fig. S6B
bottom). When the two muscles are antithetically actuated (cases (iii) and (iv)) GRACE-A
performs better than GRACE-C in the whole pressure range. When instead one muscle is
elongated passively (cases (i) and (ii)), GRACE-A still performs better than GRACE-C at low
pressures, while GRACE-C becomes favorable at higher pressures (Fig. S6C). Indeed, the
passive elongation resistance offered by the GRACE-Cs is given almost entirely by the material
stretch, whereas for the GRACE-As and GRACE-Es the elongation is facilitated by the folding
of the pleats (Fig. S6D).
We exploited the better performance of the GRACE-As in antagonistic configuration for the
wrist of the 3D-printed pneumatic hand to enable a two degree of freedom rotation (Fig. S21 and
S22).

Characterization tests
The GRACE-C samples pressurized at 22 kPa performed about 25% free contraction, 24.8%
with the 4 N load and 22% with the 8 N load (Fig. 3A, B, C, S7). As a comparison, the GRACE-
A samples pressurized at 18 kPa reached 11.3% free contraction, 16.5% at 4 N and 17.5% with
8 N (Fig. 3C and S9). This can be explained as the load stretches the GRACE-As, slightly
folding the pleats and thus increasing the unfolding range. The GRACE-A samples were also
tested in elongation: depressurized at −8.6 kPa showed 19.8% free elongation, 10.5% at 4 N and
7.3% at 8 N (Fig. 3C and S11). Therefore, the GRACE-As exhibit an actuation range
(contraction + elongation) of 25% to 30%, which is slightly higher than the contraction range of
the GRACE-Cs. The GRACE-E samples depressurized at -8.3 kPa achieved 20.5 % free
elongation, 13% with 4 N load applied while depressurized at -7.4 kPa and 8.3% with 8 N load
applied while depressurized at -6 kPa (Fig. 3C and S13). When loaded, lower de-pressure inputs
were applied to GRACE-E since in these conditions they can undergo an internal collapsing of
the folds; this can be avoided by increasing the thickness of the membrane in order to also
improve the maximum elongation of this designs.
In the isometric tests, we applied and held three different values of stretch (0 mm, 4 mm,
corresponding to 10% of the artificial muscle length, and 8 mm, corresponding to 20% of the
artificial muscle length) to the samples while applying five cycles of pressure and measuring the
output forces. The GRACE-C samples pressurized at 22 kPa exerted around 17 N pulling
blocking force (0 mm stretch), 17.8 N pulling force with 4 mm stretch and 15.6 N with 8 mm
stretch (Fig. 3 D, E, F, S8). The GRACE-A samples pressurized at 18 kPa performed 7.1 N
pulling blocking force, 11.3 N pulling force with 4 mm stretch and 14.2 N with 8 mm stretch
(Fig. 3F and S10). Here, too, we observed that the GRACE-As perform better in contraction
when stretched. Depressurized at −8.3 kPa they achieved 7.5 N pushing blocking force, and
9.5 N and 8.3 N pushing force with 4 mm and 8 mm pre-stretch, respectively (Fig. 3F and S12).
GRACE-E reached 2.1 N pushing blocking force depressurized at -8.4 kPa, 8.4 N pushing force
with 4 mm stretch applied at – 7.8 kPa and 8.7 N with 8 mm stretch applied at -7.9 kPa (Fig. 3F
and S14).
To test the dynamic behavior of the GRACEs, we also applied to a GRACE-C sample of 4 cm
length, triangular inputs with frequency ranging from 0.066 Hz to 1 Hz (Fig. S15). Almost full
deformation cycle can be achieved with frequency equal or lower than 0.1 Hz while higher
frequencies result in smaller contraction amplitude. Nevertheless, the dynamic behavior of the
actuators strongly depends on the material adopted (both on the material response and on the
pressure that it permits), on the volume of the actuator to be filled (thus on the dimensional scale)
and on the pneumatic system adopted (compressor, pipes, pressure regulator, valves).
We carried out cyclic actuation tests consisting of 1000 cycles of free contraction or elongation
(Fig. 3I and S16) by applying the same input pressures used in the isotonic tests (see Materials
and Methods). The GRACE-C reached 10.4 mm contraction (26% of the initial length)
pressurized at 23 kPa for 1000 consecutive cycles. The GRACE-A achieved 4.8 mm contraction
(12% of the initial length) pressurized at 19 kPa and 7.9 mm elongation (19.8% of the initial
length) depressurized at −8.6 kPa for 1000 consecutive cycles. The GRACE-E performed
8.4 mm elongation (21% of the initial length) depressurized at -8.6 kPa for 1000 consecutive
cycles.

Photoelasticity
The photopolymer used to fabricate the GRACEs (Formlabs Flexible 80A resin) is a photoelastic
material meaning the optical properties are related to the local stress distribution. Therefore,
photoelastic experiments can be conducted to visualize the stress distribution on the artificial
muscles or the structure in which they are embedded, resulting from the pressurization or
depressurization.
The color changes can be related to the actuation stresses because the SLA process produces
outcomes that are virtually free of residual stresses and therefore do not show a color map in their
initial unactuated state.
Supplementary Figures and Tables

Fig. S1. Geometrical model (A) Sketch of cross section of GRACE (right) and parabolic profiles passing through
points Q, P and S (left). (B) Parabolic profiles are displaced from their starting position (solid black lines) to their
final position (dashed blue lines). The length of the GRACE decreases/increases as the GRACE is pressurized/depressurized
and 𝑥𝑚∗
denotes the final distance between the central cross section and the closure of the GRACE. (C)
Approximation of the volume enclosed by the GRACE.
Fig. S2. Model outcomes Distance 𝑥𝑚 ∗
between the central cross section and the closure of the GRACE for different
geometries and actuation modes: the surfaces in (A) and (B) represent the case of contraction and elongation,
respectively, for GRACEs having 𝑎1 = 0.38 and 𝑏2 = 0.15, for variable 𝑎2 and 𝑅2 . (C) Surfaces obtained in case of
contraction for two different values of 𝑅2 , and for variable 𝑎2 and 𝑏2 .
Fig. S3. Comparison of contracting GRACEs Distance 𝑥𝑚 ∗
reached in case of contraction for GRACEs with 4
(top) and 10 (bottom) pleats. The results are displayed for varying 𝑎2 , and obtained for different combinations of 𝑅2
and 𝑏2 . For a deeper comparison, the results on the left and on the right refer to GRACEs with two different values
of 𝑎1 (0.382 and 0.665, respectively). While a trend is observed for the 4-pleats GRACEs considered here, the same
is not found for 10-pleats GRACEs: it is not straightforward to determine the effect of the single design parameter
for the general case.
Fig. S4. GRACEs materials (A) Several GRACEs printed through SLA and FDM with different materials. (B)
Pulling forces exerted by the GRACE-C of 4 cm length resulted from FEM simulations for different materials
properties (constant young modulus).
Fig. S5. GRACEs strain distribution results from FEM simulation (A) Strain distribution for the GRACE-C
during free contraction. (B) Strain distribution for the GRACE-E during free elongation. (C) Strain distribution for
the GRACE-A during free contraction. (D) Strain distribution for the GRACE-A during free elongation.
Fig. S6. GRACE muscles comparison in antagonistic configuration (A) FEM simulation for two GRACE-A in
antagonistic configuration: one contracting by pressurization and the other passively elongating (top); one
contracting by pressurization and the other elongating by depressurization (bottom). In both cases, the two extremes
are fixed while the interconnection can freely translate. (B) FEM simulation for two GRACE-C in antagonistic
configuration: one contracting by pressurization and the other passively elongating (top); one contracting by
pressurization and the other elongating by depressurization (bottom). In both cases, the two extremes are fixed while
the interconnection can freely translate. (C) Displacement of the interconnection– pressure (overpressure of the contracting
actuator) curves resulted from the FEM simulation of the four antagonistic cases. (D) Tensile passive stiffness curves resulted
from the tensile tests for the different GRACE samples of 4 cm length, while being not actuated.
Fig. S7. GRACE-C contraction isotonic characterization (A) Isotonic test with no load applied: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and displacement outputs during the isotonic test
with no load applied for five samples. (C) Actuation and release curves for the isotonic test with no load applied,
mean over the five cycles and the five samples. (D) Isotonic test with 4 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (E) Pressure inputs and displacement outputs during the
isotonic test with 4 N load for five samples. (F) Actuation and release curves for the isotonic test with 4 N load,
mean over the five cycles and the five samples. (G) Isotonic test with 8 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (H) Pressure inputs and displacement outputs during the
isotonic test with 8 N load for five samples. (I) Actuation and release curves for the isotonic test with 8 N load,
mean over the five cycles and the five samples.
Fig. S8. GRACE-C contraction isometric characterization (A) Isometric test with no pre-stretch: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and force outputs during the isometric test with no
pre-stretch for five samples. (C) Actuation and release curves for the isometric test with no pre-stretch, mean over
the five cycles and the five samples. (D) Isometric test with 4 mm pre-stretch: actuator with stretch applied at rest
(left) and after applying overpressure (right). (E) Pressure inputs and force outputs during the isometric test with 4
mm pre-stretch for five samples. (F) Actuation and release curves for the isometric test with 4 mm pre-stretch, mean
over the five cycles and the five samples. (G) Isometric test with 8 mm pre-stretch: actuator with stretch applied at
rest (left) and after applying overpressure (right). (H) Pressure inputs and force outputs during the isometric test with
8 mm pre-stretch for five samples. (I) Actuation and release curves for the isometric test with 8 mm pre-stretch,
mean over the five cycles and the five samples.
Fig. S9. GRACE-A contraction isotonic characterization (A) Isotonic test with no load applied: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and displacement outputs during the isotonic test
with no load applied for five samples. (C) Actuation and release curves for the isotonic test with no load applied,
mean over the five cycles and the five samples. (D) Isotonic test with 4 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (E) Pressure inputs and displacement outputs during the
isotonic test with 4 N load for five samples. (F) Actuation and release curves for the isotonic test with 4 N load,
mean over the five cycles and the five samples. (G) Isotonic test with 8 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (H) Pressure inputs and displacement outputs during the
isotonic test with 8 N load for five samples. (I) Actuation and release curves for the isotonic test with 8 N load,
mean over the five cycles and the five samples.
Fig. S10. GRACE-A contraction isometric characterization (A) Isometric test with no pre-stretch: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and force outputs during the isometric test with no
pre-stretch for five samples. (C) Actuation and release curves for the isometric test with no pre-stretch, mean over
the five cycles and the five samples. (D) Isometric test with 4 mm pre-stretch: actuator with stretch applied at rest
(left) and after applying overpressure (right). (E) Pressure inputs and force outputs during the isometric test with 4
mm pre-stretch for five samples. (F) Actuation and release curves for the isometric test with 4 mm pre-stretch, mean
over the five cycles and the five samples. (G) Isometric test with 8 mm pre-stretch: actuator with stretch applied at
rest (left) and after applying overpressure (right). (H) Pressure inputs and force outputs during the isometric test with
8 mm pre-stretch for five samples. (I) Actuation and release curves for the isometric test with 8 mm pre-stretch,
mean over the five cycles and the five samples.
Fig. S11. GRACE-A elongation isotonic characterization (A) Isotonic test with no load applied: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and displacement outputs during the isotonic test
with no load applied for five samples. (C) Actuation and release curves for the isotonic test with no load applied,
mean over the five cycles and the five samples. (D) Isotonic test with 4 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (E) Pressure inputs and displacement outputs during the
isotonic test with 4 N load for five samples. (F) Actuation and release curves for the isotonic test with 4 N load,
mean over the five cycles and the five samples. (G) Isotonic test with 8 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (H) Pressure inputs and displacement outputs during the
isotonic test with 8 N load for five samples. (I) Actuation and release curves for the isotonic test with 8 N load,
mean over the five cycles and the five samples.
Fig. S12. GRACE-A elongation isometric characterization (A) Isometric test with no pre-stretch: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and force outputs during the isometric test with no
pre-stretch for five samples. (C) Actuation and release curves for the isometric test with no pre-stretch, mean over
the five cycles and the five samples. (D) Isometric test with 4 mm pre-stretch: actuator with stretch applied at rest
(left) and after applying overpressure (right). (E) Pressure inputs and force outputs during the isometric test with 4
mm pre-stretch for five samples. (F) Actuation and release curves for the isometric test with 4 mm pre-stretch, mean
over the five cycles and the five samples. (G) Isometric test with 8 mm pre-stretch: actuator with stretch applied at
rest (left) and after applying overpressure (right). (H) Pressure inputs and force outputs during the isometric test with
8 mm pre-stretch for five samples. (I) Actuation and release curves for the isometric test with 8 mm pre-stretch,
mean over the five cycles and the five samples.
Fig. S13. GRACE-E elongation isotonic characterization (A) Isotonic test with no load applied: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and displacement outputs during the isotonic test
with no load applied for five samples. (C) Actuation and release curves for the isotonic test with no load applied,
mean over the five cycles and the five samples. (D) Isotonic test with 4 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (E) Pressure inputs and displacement outputs during the
isotonic test with 4 N load for five samples. (F) Actuation and release curves for the isotonic test with 4 N load,
mean over the five cycles and the five samples. (G) Isotonic test with 8 N stretching load applied: actuator with load
applied at rest (left) and after applying overpressure (right). (H) Pressure inputs and displacement outputs during the
isotonic test with 8 N load for five samples. (I) Actuation and release curves for the isotonic test with 8 N load,
mean over the five cycles and the five samples.
Fig. S14. GRACE-E elongation isometric characterization (A) Isometric test with no pre-stretch: actuator at rest
(left) and after applying overpressure (right). (B) Pressure inputs and force outputs during the isometric test with no
pre-stretch for five samples. (C) Actuation and release curves for the isometric test with no pre-stretch, mean over
the five cycles and the five samples. (D) Isometric test with 4 mm pre-stretch: actuator with stretch applied at rest
(left) and after applying overpressure (right). (E) Pressure inputs and force outputs during the isometric test with 4
mm pre-stretch for five samples. (F) Actuation and release curves for the isometric test with 4 mm pre-stretch, mean
over the five cycles and the five samples. (G) Isometric test with 8 mm pre-stretch: actuator with stretch applied at
rest (left) and after applying overpressure (right). (H) Pressure inputs and force outputs during the isometric test with
8 mm pre-stretch for five samples. (I) Actuation and release curves for the isometric test with 8 mm pre-stretch,
mean over the five cycles and the five samples.
Fig. S15. GRACE-C dynamical response Responses to triangular inputs with frequencies spanning from 0.066 Hz
to 1 Hz for a 4-cm GRACE-C sample made with Flexible 80A (SLA).
Fig. S16. GRACE muscles fatigue life test (A) Contraction fatigue test for the GRACE-A samples with no applied
load over 1000 cycles. (B) Elongation fatigue test for the GRACE-A samples with no applied load over 1000 cycles.
(C) Elongation fatigue test for the GRACE-E samples with no applied load over 1000 cycles.
Fig. S17. GRACE characterization setup Experimental setup employed to characterize the GRACE samples by
tensile, isotonic, isometric and life-cycles tests with different settings
Fig. S18. GRACE automated workflow realization Schematic workflow of the phases composing the automated
process to develop the desired GRACE starting from the application specifications.
Fig. S19. Pneumatic hand manufacturing Pneumatic hand still with supports, right after being directly printed in
one step with SLA technique and Flexible 80A resin through Form 3.
Fig. S20. Hand pneumatic circuit setup Scheme of the pneumatic circuit employed to pressurized and
depressurized the several GRACEs embedded in the pneumatic hand structure. The valves were controlled by an
Arduino Mega board.
Fig. S21. Wrist antagonistic actuation comparison (A) Rotations around the axis perpendicular to the hand palm,
achieved contracting, through pressurization, two GRACE-A, while passively elongating the other two. (B)
Rotations around the axis perpendicular to the hand palm, achieved by contracting, through pressurization, two
GRACE-A, while actively elongating, by depressurization, the other two. (C) Rotations around the axis parallel to
the hand palm, achieved by contracting, through pressurization, two GRACE-A, while passively elongating the
other two. (D) Rotations around the axis parallel to the hand palm, achieved by contracting, through pressurization,
two GRACE-A, while actively elongating, by depressurization, the other two.
Fig. S22. Pneumatic hand wrist movements (A) Rotations of the wrist achieved by contracting, through
pressurization, two adjacent GRACE-As and elongating by depressurization the other two. In the top right corner in
red the contracting actuators and in blue the elongating ones. (B) Rotations of the wrist achieved by contracting,
through pressurization, three GRACE-As and elongating by depressurization the left one (C) Rotations of the wrist
achieved by contracting, through pressurization, one GRACE-A and elongating by depressurization the other three.
(D) Rotations of the wrist achieved by contracting, through pressurization, the two GRACE-As placed on the
diagonal of the base and elongating by depressurization the other two.
Fig. S23. Photoelasticity setup Experimental setup to track the color change related to the stress distribution in the
pneumatic hand, while being pressurized, due to the photoelasticity of the material adopted. It is composed
of an LCD screen, that already has a linear polarizing filter inside, a second filter rotated by 90° with respect to the
transmission orientation of the LCD, and a Nikon camera.
Table S1. Model parameters. Geometrical parameters resulting from the optimization
implemented by the geometrical model for the three main GRACE designs: GRACE-C,
GRACE-E, GRACE-A.

GRACE-C GRACE-E GRACE-A

parameters Max contraction Max elongation Antagonistic

a1 0.48 0.10 0.24

b1 0.17 0.26 0.29

R1 0.515 0.89 0.76

a2 0.16 0.06 0.09

b2 0.11 0.22 0.13

R2 0.53 0.90 0.80


Table S2. Comparison between GRACE model and FEA. Comparison between contraction
and extension ratios obtained by the GRACE model and FEAs for the three main GRACE
designs.

CONTRACTION EXTENSION

Model FEA Model FEA

GRACE-C 44% 29% - -

GRACE-E - - 18% 21%

GRACE-A 30% 14% 13% 18%


Table S3. Versatility comparison. Pneumatic artificial muscles versatility comparison along
with those of skeletal muscles.

3D-Printed
Skeletal GRACE GRACE (*) Thin Cavatappi
PPAM Silicone
Features Muscle (Flexible 80A) (FlexFill) McKibben Actuators
(11, 15, 31) Actuators
(3, 4) (4 cm size) (4 cm size) (24) (26)
(29)

Stress 350 kPa 26 kPa (-9 kPa) 261 kPa (-77 kPa) 670 kPa ~ 8650 kPa ~ 1 kPa 700 kPa

20%-
Strain 26% (-21%) 26% (-21%) 38% 27% 8.5% ~ 50%
40%

20 kPa 220 kPa


Pressure n/a 400 kPa 550 kPa 6 kPa 2000 kPa
(-10 kPa) (-83 kPa)

Stress/Pressure n/a 1.3 (0.9) 1.2 (0.93) 1.675 ~ 15.73 ~ 0.17 0.35

13 ∙ 10-3 kPa-1 1.2 ∙ 10-3 kPa-1 0.95 ∙ 10-3 0.5 ∙ 10-3 14.2 ∙ 10-3 0.25 ∙ 10-3
Strain/Pressure n/a
(21 ∙ 10-3 kPa-1) (2.5 ∙ 10-3 kPa-1) kPa-1 kPa-1 kPa-1 kPa-1

Bidirectionality n/a n/a n/a n/a

Scalability n/a

Ease of
n/a
manufacturing

Architectures
n/a
feasibility

Table notes:
*These are results from FEA
() Values between bracket are related to active extension.
Bidirectionality: capability to actively and passively extend (in the McKibben design, as for all the
high
traditional PAMS, the passive extension is hindered by the high tensile resistant fibers; nevertheless, the
thin McKibben might be able to extend if designed with the proper braiding angle as for the traditional
McKibben, but it was not demonstrated). medium
Scalability: capability to be fully scaled up/down without compromising the performances (i.e., thin
McKibben and Cavatappi actuators are both easily scalable in diameter while not in length)
Ease of manufacturing: related to the number of manufacturing steps, assembling phases, and required low
tooling
Architectures feasibility: how easily they can be implemented in complex architectures mimicking the
muscular system configurations
Supplementary Movies

Movie S1.
GRACE-C samples (4 cm length) made with several materials and membrane thickness raising
up different weights.

Movie S2.
Contraction of four GRACE-C samples with different scale factor, from 1 cm length to 10 cm,
made of Flexible 80A through SLA.

Movie S3.
Isotonic characterization with no load applied showing the free contraction, while being
pressurized, for GRACE-C and GRACE-A, and the free elongation, while being depressurized,
for GRACE-A and GRACE-E.

Movie S4.
Isometric characterization with no pre-stretch showing the blocking pulling force, while being
pressurized, for GRACE-C and GRACE-A, and the blocking pushing force, while being
depressurized, for GRACE-A and GRACE-E.

Movie S5.
Color change of a GRACE-C sample, related to the stress distribution, subjected to the isometric
test with no pre-stretch, while pressurized.

Movie S6.
Movements achievable by the pneumatic hand by pressurizing the GRACE-Cs related to the
fingers and pressurized or depressurized the GRACE-As in the wrist. On the top left corner, it is
shown a small scheme that helps to recognize which GRACE is pressurized, depressurized, or
not actuated.

Movie S7.
Color change in the pneumatic hand due to the variation of the local stresses, while actuated.

You might also like