Soft robotic finger with variable effective length enabled by an antagonistic constraint mechanism

Abstract

Compared to traditional rigid robotics, soft robotics has attracted increasing attention due to its advantages in compliance, safety, and low cost. As an essential part of soft robotics, the soft robotic gripper also shows its superior while grasping objects with irregular shapes. Recent research has been conducted to improve grasping performance by adjusting the variable effective length (VEL). However, the existing VEL function achieved by mechanisms such as multi-chamber design or tunable stiffness shape memory material requires a complex pneumatic circuit design or a time-consuming phase-changing process. This work proposes a fold-based soft robotic finger with VEL function from 3D printing. It is experimentally tested and modeled by the hyperelastic material property. Mathematic and finite element modeling is conducted to study the bending behaviour of the proposed soft actuator. Most importantly, an antagonistic constraint mechanism is proposed to achieve the VEL, and the experiments demonstrate that better conformity is achieved. Finally, dual-mode grippers are designed and evaluated to demonstrate the advances of VEL on grasping performance.

Full text

Smart Materials and Structures
Smart Mater. Struct. 32 (2023) 055001 (11pp) https://doi.org/10.1088/1361-665X/acc36b
Soft robotic finger with variable effective
length enabled by an antagonistic
constraint mechanism
Xing Wang1,2and Hanwen Kang3,4,∗
1Laboratory of Motion Generation and Analysis, Monash University, Melbourne, Australia
2Robotics and Autonomous Systems Group, Data61, CSIRO, Brisbane, Australia
3College of Engineering, South China Agricultural University, Guangzhou, People’s Republic of China
4Foshan-Zhongke Innovation Institute of Intelligent Agriculture and Robotics, Foshan,
People’s Republic of China
E-mail: hanwen.kang@outlook.com
Received 3 November 2022, revised 31 January 2023
Accepted for publication 10 March 2023
Published 22 March 2023
Abstract
Compared to traditional rigid robotics, soft robotics has attracted increasing attention due to its
advantages in compliance, safety, and low cost. As an essential part of soft robotics, the soft
robotic gripper also shows its superior while grasping objects with irregular shapes. Recent
research has been conducted to improve grasping performance by adjusting the variable
effective length (VEL). However, the existing VEL function achieved by mechanisms such as
multi-chamber design or tunable stiffness shape memory material requires a complex pneumatic
circuit design or a time-consuming phase-changing process. This work proposes a fold-based
soft robotic nger with VEL function from 3D printing. It is experimentally tested and modeled
by the hyperelastic material property. Mathematic and nite element modeling is conducted to
study the bending behaviour of the proposed soft actuator. Most importantly, an antagonistic
constraint mechanism is proposed to achieve the VEL, and the experiments demonstrate that
better conformity is achieved. Finally, dual-mode grippers are designed and evaluated to
demonstrate the advances of VEL on grasping performance.
Keywords: soft actuator, 3D printed actuator, variable effective length, constraint mechanism
(Some gures may appear in colour only in the online journal)
1. Introduction
Soft robotics has been extensively studied recently due to its
inherent advantages of being compliant, robust to impact, ex-
ible, and safe compared to traditional rigid robotics [1]. The
soft bending nger demonstrates excellent applicability in the
∗Author to whom any correspondence should be addressed.
Original Content from this work may be used under the
terms of the Creative Commons Attribution 4.0 licence. Any
further distribution of this work must maintain attribution to the author(s) and
the title of the work, journal citation and DOI.
design and manufacture of the soft prosthetic hands [2], soft
robotic gripper [3,4], soft prosthetic device [5], soft articial
sh [6], rehabilitation device [7], and various types of loco-
motion robots [8,9]. Among these, the soft robotic gripper
shows advantages in grasping objects with different shapes
and weights, which make them excellent candidates for applic-
ations like agricultural harvesting [10,11], industrial food and
fruit processing [12,13], etc. The soft robotic bending actuator
can be classied into three main categories based on its bend-
ing principle: bre-reinforced soft actuator [14], PneuNet [15],
and eccentric actuator [16], corresponding to multi-material
asymmetry, pleated structure asymmetry, and eccentric void
asymmetry principle respectively.
1361-665X/23/055001+11$33.00 Printed in the UK 1 © 2023 The Author(s). Published by IOP Publishing Ltd

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 1. Tendon arrangement in human nger [24].
Pioneering research has been conducted on the develop-
ment of dexterous soft robotic grippers, which aims to improve
the performance of the gripper in terms of the allowable
grasping size, grasping force, etc. Park et al [17] proposed a
hybrid PneuNet gripper for improved force and speed grasp-
ing application. Zhou et al [18] designed a three-segment soft
gripper to grasp objects of more sizes. Afterwards, he pro-
posed another 13 degree of freedom (DoF) soft hand for dex-
terous grasping [19]. In addition, Wang et al proposed the uid
robotic arm with three ber reinforced hydraulic chambers
to achieve 3 DoF bending motion [20]. However, the multi-
segment/DoF design typically requires multi-active channels,
tubing, and multi-valve pneumatic systems to control indi-
vidual segments for a desired effective length, making the
pneumatic control system signicantly complex and bulky.
Besides, the variable effective length (VEL) is also explored
for the soft nger with one whole segment to achieve better
conformity of the objects, allowing better grasping in various
shapes and weights. Hao et al [21] proposed a gripper that can
achieve VEL by selectively softening shape memory polymer
(SMP) sections via a exible heater. Even the softening can be
realized within 0.6 s; the cooling of SMP takes up to 14 s. The
exact timing issue limits the broader application of an SMP-
enabled VEL soft actuator designed in [22,23].
This study introduces a soft folded-based bending actuator
with a VEL function enabled by an antagonistic constraint
mechanism (ACM). The VEL is bio-inspired by the tendon
arrangement in the human nger as shown in gure 1, where
different tendons are xed on different phalanges. The tendons
can be either actively tightened or passively xed. Different
tendon arrangement helps the nger achieve variable effective
bending length, which benets the grasping of various objects.
This VEL in the proposed soft actuator is also enabled by
the selectively placed tendon constraint on the top side of the
soft actuator, which allows a design without a time-consuming
reprint process. There are some researchers that implemented
the hybrid pneumatic and tendon mechanism in soft robot-
ics. For example, Li et al presented a pre-charged pneumat-
ics actuator, which is actuated by pre-charged air pressure and
retracted by tendons [25]. Its bending angle changes when the
tendons are pulled or released. Shiva et al proposed a soft
manipulator that was actuated by pneumatic pressure and ten-
dons. Tendons were connected to the distal ends of the robot
section and run along the outer sleeve allowing the sleeve to
bend when the corresponding tendon was pulled [26]. This
work is different from existing work because the tendon is not
used as an actuation unit but it can be customized to a certain
length and integrated with the soft robotic nger or it can be
constrained by the motor to maintain its length.
In addition, the stress–strain curve of NinjaFlex is experi-
mentally determined due to its nonlinear property as a hypere-
lastic material. Experiments are also conducted on the bending
motion of the soft nger with VEL enabled by tendon con-
straint. The main contributions of this research work are as
follows:
•Mathematically modeling with the hyperelastic property of
NinjaFlex is derived for the fold-based soft actuator to study
the bending angle under various input pressures.
•VEL is achieved with fold-based design by an ACM.
•A two-mode gripper is designed and tested to be capable
of grasping and holding objects with various weights and
shapes.
2. Design and manufacture
The soft actuator was designed in Solidworks (Dassault Sys-
tems Solidworks Corp.), as shown in gure 2. The computer
aided design (CAD) model was then sliced in PrusaSlicer. The
key printing settings that affect the airtight of the soft nger
were sourced from our previous work [27,28], then tuned and
tested to t this manufacture. After testing, a low-cost fused
deposition manufacturing (FDM) 3D printer, Prusa Mk3s, was
utilized to print the proposed soft nger with a commercially
available lament, NinjaFlex.
The soft nger was integrated with the 3D-printed rigid
mounts, as shown in gure 2. They are three holes within the
3D printed mount, where the bourdon tubes were slid in for
the tendon to sit. These bourdon tubes also reduce the poten-
tial friction between the rigid mount and the tendon during
the actuation process. A commercially available stiff tendon
(26.4 kg diameter−1:0.342 mm) was slid into the tendon guide
to provide constraints. One end was xed on the soft nger by
making a node and gluing on the soft actuator, and the other
end could be a free end or connected to the actuation mechan-
ism, as shown in gure 3. The section that is constrained by the
stiff tendon (red line) would remain in a straight conguration
under actuation while the rest of the nger bends. The cross-
section view of the design with detailed parameters is shown
in gure 4. The detailed parameters and values are also shown
in table 1.
The soft nger is expected to bend uniformly under ina-
tion pressure due to the fold-based geometry design. In addi-
tion, the tendon (in yellow) as shown in gure 2, can be select-
ively xed on the top side of the robotic nger where there is no
strain-limiting layer. The other end of the tendon is controlled
by motors. A candidate of the xed point on the soft nger is
illustrated and highlighted in the red dot in gure 2. This xed
point can be tuned manually to vary the number of segments
2

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 2. The explored view and assembly of the proposed soft actuator.
Figure 3. Different tendon x points on the soft robotic ngers.
Figure 4. Cross-section view of the design.
Table 1. Values for key parameters.
Parameters L H W tsh1twlf
Values/mm 114 17.8 20 1 2 2 10
that need to be constrained. A VEL can be programmed in this
case.
3. Mathematical modeling
3.1. TPU characterization and material model
The thermoplastic polyurethanes (TPU) characterization is
based on Yap et al [29]. However, even with the same fabric-
ation material Ninjaex, the hyperelastic material model can
be slightly different, which may affect their predicted perform-
ance. In this case, we performed the tensile test, FEM simula-
tion and further experimental tests on our proposed design.
To nd the stress–strain relationship of the hyperelastic
material, NinjaFlex, an uniaxial tensile test was performed on
the dumbbell samples. It needs to be noticed that the bend-
ing motion of the proposed fold-based design is caused by
the expansion of the walls that are printed longitudinally. So
the materials are mainly experiencing tension in the longitud-
inal direction. The dumbbell samples are then printed in the
3

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 5. (a) Dumbbell sample printed in the longitudinal direction
for tensile test, (b) Instron universal machine for tensile test.
Figure 6. The stress–strain curve and Mooney–Rivlin model of
NinjaFlex.
longitudinal direction, as shown in gure 5. The tests were
performed under ISO 37 standards, where the samples were
stretched by 600% at 200 mm min−1using an Instron univer-
sal Tester E3000 (Instron, USA). The average stress–strain
data shown in gure 6were tting to the hyperelastic model
named Mooney–Rivlin models in Abaqus (Simulia, Dassault
Systems, RI) [29], while POLY_N2 in the gure means the
Mooney Rivlin model. The parameters used to plot the curve
are shown in table 2.
3.2. Finite element modeling
With the material property dened, it is feasible to perform
nite element modeling to study the bending motion of the soft
robotic nger. Firstly, the CAD model is saved as a step format
and imported into the simulation software Abaqus. Next, the
material property needs to be dened with the parameters
from the previous TPU modeling. The section can be added to
the model after creating and assigning the material property.
Two surfaces are created named inner and the contact surface,
while the former is selected inside the cavity, and the latter
enables self-contact interaction during bending motion. Solid
Table 2. Mooney Rivlin model for NinjaFlex.
Hyperelastic model Material constant Value
Mooney Rivlin C01 2.898 MPa
C10 −0.0450 MPa
C11 −0.017 MPa
C20 0.00226 MPa
C02 0.183 MPa
Incompressible parameter D 0 MPa−1
tetrahedral quadratic hybrid elements (element type C3D10H)
are used to mesh the soft nger. Various static pressures are
input and applied on the inner surface as the load, and the
ENCASTRE boundary condition is applied at the proximal
end.
3.3. Bending angle
The bend angle that denes how much the actuator curls is
treated as one key index to evaluate its performance [15]. For
the fold-based soft actuator, it can be concluded from the pre-
liminary simulation that the bending motion occurs due to the
elongation of the top wall of the connector. To calculate the
wall expansion, each wall is modeled as a rectangular plate
with four edges clamped [30,31]. The maximum deection
occurs in the centre of the plate, while its value is affected
by the plate thickness, aspect ratio a/b, and proportional to
(a/2)α[32]. The variables aand bare hwand Wlrespect-
ively, as shown in gure 4. To model the expansion of both
vertical and top walls, we adopt the equivalent connector con-
version from Lotani et al [33] while keeping the rest of the
nger dimensions the same. The equivalent connector with the
updated dimensions is as shown in gure 7. The dimensions
for the equivalent connector xe,he, and tecan be expressed
with the actual dimension of the connector lc,h1, and tw[33]:
xe=a
2+lc(1)
he=a
b[h1+(a
2)α](2)
te=tw+tc
2(3)
Another assumption is that all the energy is assumed to store
at a distance of hee from the bottom layer, which simplies the
structure from gures 7(b) and (c):
Vtt =2vs+vt+vb(4)
Vtt,vs,vt,vbare the total volume, volume at two side, top side
and bottom side, respectively:
hee =2vs∗hs+vt∗ht+vb∗hb
Vtt
.(5)
The total potential energy can be expressed as the sum of the
work done by air and the strain energy stored in the walls:
W=Wair +Wstr.(6)
4

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 7. The cross-section view of the soft actuator with (a) the dimensions for the equivalent connector, (b) side view of the equivalent
connector, (c) energy storage, (d) bending of one segment.
The potential energy caused by ination pressure is:
Wair =ˆMp(ϕ)dϕ. (7)
The constant moment Mpcaused by the constant pressure is
assumed to act on the soft nger with a distance of d:
Mp=PAd (8)
d=1
2(H1−hee).(9)
Upon ination, each equivalent connector forms a curved
shape with a radius of rand angle of ϕ, as shown in gure 7(d):
ϕ=xe
r=xe
hee
(λ−1).(10)
By combining equations (7)–(10), we have the work done by
the input pressure as follows:
Wair =PA ·(H1−hee)
2
·
xe
hee
(λ−1).(11)
The strain energy stored in the walls can be expressed as:
Wstr =ˆEstrdVstr (12)
Estr =
N
∑
p,q=0
Cpq(I1−3)p(I2−3)q+
M
∑
m=1
1
Dm
(J−1)2m.(13)
We assume that the NinjaFlex is incompressible, so J
equals to 1 and the second component is zero. By combing
equations (12) and (13), the strain energy stored in the walls
can be expressed as:
Wstr = [C10(I1−3) + C01 (I2−3) + C11(I1−3)(I2−3)
+C20(I1−3)2+C02 (I2−3)2]·Vtt.
While I1,I2are the rst and second deviatoric strain invariants,
and they can be calculated with equations:
I1=λ2
1+λ2
2+λ2
3(14)
I2=λ2
1λ2
2+λ2
2λ2
3+λ2
3λ2
1.(15)
The material is incompressible, so λ1·λ2·λ3=1, λ1=λ,
λ2=λ−1,λ3=1. When the soft nger bends to a certain
angle, it is in equilibrium state, so the gradient of the potential
energy W.R.T λequals to zero, which can be written as:
∂Wair
∂λ +∂Wstr
∂λ =0 (16)
∂Wair
∂λ =PA ·(H1−he)
2
·
xe
hee
(17)
∂Wstr
∂λ =2(C10 +C01 )·(λ−λ−3) + 4(C20 +C11 +C02)
·(λ−λ−3)·(λ2+λ−2−2).(18)
By combining equations (1)–(5) and (16)–(18), the bending
angle ϕfor one connector can be solved. The overall bending
for the soft nger is the sum of all connectors, then we have:
θ=n·ϕ. (19)
From equation (19), the bending angle is proportional to the
number of connectors/folds. This means that a VEL can be
potentially achieved by varying the value n, which is the num-
ber of connected segments/folds. Our proposed mechanism
selects different constrain points on the soft nger to vary the
number of segments/folds to tune the effective length.
5

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 8. Set up for bending angle test.
Figure 9. Bending angle of the fold-based soft nger (a) 20 kPa experiment, (b) 20 kPa simulation, (c) 50kPa experiment, (d) 50 kPa
simulation, (e) 100 kPa experiment, (f) 100 kPa simulation.
4. Experimental results and discussion
4.1. Test of bending angle
Soft bending actuators have various bending states under dif-
ferent inputs. To nd out the relationship between the input
pressure and the bending angle of this 3D printed fold-based
design, the test of bending angle is set up as shown in gure 8.
The soft nger is xed at the proximal end and placed hori-
zontally on a grid board. This placement is to eliminate the
effect of gravity on the in-plane bending motion. Various pres-
sure inputs are regulated by a pressure regulator (SMC Cor-
poration, Australia) and then applied to the soft nger with a
step of 10 kPa. An red-green-blue (RGB) camera (Realsense
D455, Intel) is xed on the tripod to capture the image at a cer-
tain bending state, which will be further processed to obtain
the value of the bending angle. The experimental data and the
FEM result under certain pressures can be seen in gure 9. The
mathematical result is obtained from equation (19). Besides,
a detailed comparison of the experimental results, FEM, and
the mathematical modeling are also shown in gure 10. It can
be seen that the mathematical modeling matches the exper-
imental results well under lower pressures. While the dif-
ferences are slighter and larger under higher pressure values
(more than 110 kPa). The geometry parameter αis ne-tuned
based on the results to get the best t mathematical model.
The maximum error between the prediction and the experi-
ment result is 6.3%. While the FEM demonstrates higher dif-
ferences between the experimental results, especially under
higher inputs. This might be caused by the relatively softer
material property dened by the TPU uniaxial test.
4.2. Variable effective length
This work proposes an ACM to enable a VEL. It is designed to
constrain the tendon guide side, limiting the bending motion
of a certain number of folds/segments. The fundamental idea
is to selectively choose the xed point of the tendon. With
the antagonistic force from the longitudinal direction of the
tendon, certain predened segments/folds can be constrained,
preventing the bending motion. Simulation is also conducted
6

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 10. Comparison of bending angle data between experiments,
modeling, and FEM.
Figure 11. VEL of the soft nger with ACM, the green line
indicates the constrained segments.
by adding additional constraints to allow various bending
patterns, as shown in gure 11. The ACM is realized by tendon
constraint, as highlighted in the green line in gure 11. The
proposed ACM can achieve VEL and potentially conform to
objects of different contours. To investigate shape conform-
ity, several samples with various curvatures are utilized as the
test samples. The effective length of the soft nger is selected
based on the curvature of the samples. As shown in gure 12,
the full effective length (FEL) shows the best t to a constant
curvature sample, while the ACM soft nger conforms to the
other irregular samples well based on the area of contact. This
is due to the advances of ACM, which extends the soft nger
from a single bending prole to versatile bending shapes.
4.3. Gripper design and grasping tests
The ACM mechanism enables the soft robotic nger with
VEL property, with which a certain effective length can be
tuned for a specic object. To achieve this, we design a soft
robotic gripper with two ngers facing each other. The detailed
Figure 12. Conformity test of the VEL soft actuator.
design for the gripper is shown in gure 13. Both actuators are
actuated with positive pneumatic pressure, while the ACM is
enabled by tendon-fasten. To be specic, one end of the ten-
don is xed by a step motor (Bipolar, NEMA 17) once the
shorter effective length of the actuator is required.The gripper
is designed to have two modes: the full-bending mode and the
tip-bending mode. While the tip-bending mode is achieved by
presetting the xed point on the third last fold on the soft n-
ger. This xed point can be adjusted or added quickly for a
tunable effective length before the grasping tasks.
We mounted the gripper on a commercially-available
industry robotic arm, UR5, to perform grasping tasks. The soft
gripper rst demonstrates its grasping capacity by gripping
objects with various shapes by both pinch and power grasp, as
shown in gure 14. The gripper successfully picked up objects
with pinch and power grasps, weighing from 22 g to 435 g.
To further investigate the effect of VEL on the soft nger,
the tendon position is initially locked by the step motor. As
a result, only unconstrained segments are free to bend. The
comparison between the full bending and the bending with
different effective lengths is shown in gure 15. It can be
seen from the grasping patterns that the contact area can be
increased signicantly when the soft nger switches to differ-
ent effective lengths for a specic object. The increase in the
contact area can benet the envelope grasping and be poten-
tially utilized to receive more contact information when the
sensing mechanism is integrated. The benet of the VEL on
the grasping force is evaluated by a pull-out force test, as
shown in gure 16. Cylinders with different dimensions are
grasped and held by the soft gripper in two separate modes.
The weights are hanging to provide the gravity force to pull
the cylinder out of the gripper. The critical pull-out force is
utilized as an index to evaluate the grasping force. A detailed
comparison between modes when grasping cylinders of differ-
ent dimensions is shown in gure 17(a). From the results, a lar-
ger force is required to pull the cylinder out with the increase of
input pressure or the diameter of the cylinder. This is because a
7

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 13. The detailed design of proposed soft gripper with key components labelled, (a) two-nger gripper (b) three-nger gripper.
Figure 14. Grasping test of (a) rectangular plate, 22 g, (b) wooden block, (202 g), (c) knife, 78g, (d) wooden block (207 g), (e) water bottle
spray (435 g).
8

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 15. Grasping of rectangular plate and sphere with full FEL (a) (c), and constrained effective length (CEL) (b), (d).
Figure 16. Setup for the test of pull-out force: the weights are applied to pull the cylinder out of the gripper at different inputs.
9

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Figure 17. Results for pull-out force test (a) pull-out force with various input pressures for soft nger with FEL, (b) comparison between the
pull-out force between FEL and CEL.
Figure 18. (a) Test setup for three-nger gripper, (b) pull-out force
results for three-nger gripper.
soft nger can generate more grasping force with the increase
of pressure or the contact area. From gure 17(b), varying the
effective length of grasping can secure the grasping motion, as
more force is required to pull the target out of the gripper. An
increase of 13.93% and 10.63% in the pulling force is observed
while grasping cylinders with a diameter of 60 mm and 65 mm,
respectively.
To further evaluate the benets of VEL, a three-nger grip-
per is tested by pulling a test sample out of the gripper in a
lateral direction, as shown in gure 18. The VEL enables the
soft gripper with a larger contact area with the grasped object,
which results in more stable grasping. The pull-out force is
used to evaluate the benets of VEL. There are two blocks
with a height of 74 mm and 84 mm utilized, and each of them
has a coarse and smooth surface to vary the surface condi-
tion. As shown from the results, the pull-out force is smaller
on a smooth surface due to a smaller coefcient of friction.
The pull-out force becomes larger when the VEL is enabled
while pulling the block out of the three-nger gripper for both
blocks. On the smooth surface the pull-out force increases by
36.9 %and 23.6 %respectively for blocks with 74 and 84 mm
lengths.
5. Conclusions
To achieve a broader application of the soft robotic gripper on
grasping objects with various shapes and weights, this work
proposes a VEL soft robotic nger enabled by ACM. The
robotic nger is made from direct 3D printing, simplifying
the manufacturing complexity compared with the multi-step
casting method. The NinjaFlex utilized is also experiment-
ally tested to provide more detailed and appropriate material
properties for the modeling. Energy-based mathematical mod-
eling is proposed to model the bending angle by considering
the hyperelastic property of NinjaFlex. The experimental res-
ults indicate that the proposed model can predict the bending
behaviour within a maximum error of 6.3% under the max-
imum operation pressure (130 kPa). The VEL is also superior
when grasping non-constant curvature samples based on the
area of contact. Both two-nger and three-nger grippers are
nally designed to test the grasping performance, which can
lift objects of various shapes.
There are still limitations to the proposed ACM; for
example, the VEL in the middle segments of the soft nger
currently requires manual adjustment. Even though it is much
more efcient than redesigning and reprinting the soft actu-
ator, the automatically programmed VEL function will still
benet a more intelligent grasping. Future research will be
focused on enabling the automatic switching of the constraint
point on the soft actuator.
Data availability statement
All data that support the ndings of this study are included
within the article (and any supplementary les).
10

Smart Mater. Struct. 32 (2023) 055001 X Wang and H Kang
Acknowledgment
We acknowledge Professor Chao Chen for his kind support
and professional discussion. We also thank Dr Hongyu Zhou
for his help in the 3D model design and drawing. Dr Xing
Wang would like to thank the funding support from CSIRO’s
CERC program.
ORCID iD
Xing Wang https://orcid.org/0000-0002-8676-7821
References
[1] Wang H, Totaro M and Beccai L 2018 Adv. Sci. 51800541
[2] Liang Z, Cheng J, Zhao Q, Zhao X, Han Z, Chen Y, Ma Y and
Feng X 2019 Adv. Mater. Technol. 41900317
[3] Kang H, Zhou H, Wang X and Chen C 2020 Sensors 20 5670
[4] Wang X, Kang H, Zhou H, Au W and Chen C 2022 Comput.
Electron. Agric. 193 106716
[5] Gu G et al 2021 Nat. Biomed. Eng. 1–10
[6] Liang J, Wang T and Wen L 2011 J. Field Robot. 28 70–79
[7] Chen Y, Le S, Tan Q C, Lau O, Wan F and Song C 2017 A
lobster-inspired robotic glove for hand rehabilitation 2017
IEEE Int. Conf. on Robotics and Automation (ICRA) (IEEE)
pp 4782–7
[8] Tang Y, Chi Y, Sun J, Huang T H, Maghsoudi O H, Spence A,
Zhao J, Su H and Yin J 2020 Sci. Adv. 6eaaz6912
[9] Li Y, Fish F, Chen Y, Ren T and Zhou J 2019 Bioinspir.
Biomim. 14 066008
[10] Zhou H, Xiao J, Kang H, Wang X, Au W and Chen C 2022
Sensors 22 5483
[11] Wang X, Kang H, Zhou H, Au W, Wang M Y and Chen C
2023 Comput. Electron. Agric. 204 107552
[12] Wang Z, Kanegae R and Hirai S 2021 Soft Robot. 8542–54
[13] Wang Z, Makiyama Y and Hirai S 2021 J. Robot.
Mechatronics 33 935–43
[14] Wang Z, Polygerinos P, Overvelde J T, Galloway K C,
Bertoldi K and Walsh C J 2016 IEEE/ASME Trans.
Mechatronics 22 717–27
[15] Alici G, Canty T, Mutlu R, Hu W and Sencadas V 2018 Soft
Robot. 524–35
[16] Yang C, Kang R, Branson D T, Chen L and Dai J S 2019
Mech. Mach. Theory 139 526–41
[17] Park W, Seo S and Bae J 2018 IEEE Robot. Automa. Lett.
465–72
[18] Zhou J, Chen S and Wang Z 2017 IEEE Robot. Autom. Lett.
22287–93
[19] Zhou J, Yi J, Chen X, Liu Z and Wang Z 2018 IEEE Robot.
Autom. Lett. 33379–86
[20] Wang T, Yao S and Zhu S 2022 Smart Mater. Struct. 31 115011
[21] Hao Y, Liu Z, Liu J, Fang X, Fang B, Nie S, Guan Y, Sun F,
Wang T and Wen L 2020 Smart Mater. Struct. 29 035006
[22] Wang W, Yu C Y, Abrego Serrano P A and Ahn S H 2020 Soft
Robot. 7283–91
[23] Santoso J, Skorina E H, Salerno M, de Rivaz S, Paik J and
Onal C D 2019 Smart Mater. Struct. 28 035012
[24] Ikazia Z 2015 (available at: www.ikazia.nl/aandoeningen-
behandelingen/mallet-vinger)
[25] Li Y, Chen Y, Ren T, Li Y and Choi S H 2018 Soft Robot.
5567–75
[26] Shiva A, Stilli A, Noh Y, Faragasso A, De Falco I, Gerboni G,
Cianchetti M, Menciassi A, Althoefer K and
Wurdemann H A 2016 IEEE Robot. Autom. Lett. 1632–7
[27] Wang X, Khara A and Chen C 2020 Bioinspir. Biomim.
15 056017
[28] Wang X, Zhou H, Kang H, Au W and Chen C 2021 Smart
Mater. Struct. 30 125001
[29] Yap H K, Ng H Y and Yeow C H 2016 Soft Robot. 3144–58
[30] Rosalia L, Ang B W K and Yeow R C H 2018 IEEE Robot.
Autom. Lett. 33489–96
[31] Wang J, Fei Y and Pang W 2019 IEEE/ASME Trans.
Mechatronics 24 990–1001
[32] Timoshenko S and Woinowsky-Krieger S 1959 Theory of
Plates and Shells vol 2 (New York: McGraw-hill)
[33] Lotani A, Yi X, Shao Z, Zhao H and Parkestani A N 2021
Extreme Mech. Lett. 44 101215
11