SOFT ROBOTICS

Volume 4, Number 4, 2017

ª Mary Ann Liebert, Inc.
DOI: 10.1089/soro.2016.0081

ORIGINAL ARTICLE

Shape Memory Alloy-Based Soft Gripper

with Variable Stiffness for Compliant and Effective Grasping
Wei Wang1,2 and Sung-Hoon Ahn1,2
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

Abstract

Soft pneumatic actuators and motor-based mechanisms being concomitant with the cumbersome appendages
have many challenges to making the independent robotic system with compact and lightweight configuration.
Meanwhile, shape memory actuators have shown a promising alternative solution in many engineering ap-
plications ranging from artificial muscle to aerospace industry. However, one of the main limitations of such
systems is their inherent softness resulting in a small actuation force, which prevents them from more effective
applications. This issue can be solved by combining shape memory actuators and the mechanism of stiffness
modulation. As a first, this study describes a shape memory alloy-based soft gripper composed of three identical
fingers with variable stiffness for adaptive grasping in low stiffness state and effective holding in high stiffness
state. Each finger with two hinges is fabricated through integrating soft composite actuator with stiffness
changeable material where each hinge can approximately achieve a 55-fold changeable stiffness independently.
Besides, each finger with two hinges can actively achieve multiple postures by both selectively changing the
stiffness of hinges and actuating the relevant SMA wire. Based on these principles, the gripper is applicable for
grasping objects with deformable shapes and varying shapes with a large range of weight where its maximum
grasping force is increased to *10 times through integrating with the stiffness changeable mechanism. The
final demonstration shows that the finger with desired shape-retained configurations enables the gripper to
successfully pick up a frustum-shaped object.

Keywords: shape memory alloy, stiffness modulation, soft fingered gripper

Introduction gas or vacuum and motor-based cable-driven mechanisms

being concomitant with the cumbersome appendages have

R igid manipulators made of cumbersome mechanical

components and sensors with finite degrees of freedom
can produce accurate movements through kinematic joints in
limitations to making compact and lightweight autonomous
robotic systems. DEAs are an attraction option for robotic
application, but the relatively high-inducing voltage and small
well-defined environments. Excluding massive structures actuation force limits their usability in human environments.
and sophisticated controls, recent advances in soft robotics Meanwhile, SMA-based soft actuators with compact configu-
exploit the flexibility and compliance of materials to create ration have shown the advantages of being able to signifi-
manipulators exhibiting soft and adaptive interactions with cantly reduce the size, weight and system complexity, and
unpredictable environments.1 Soft grippers with dexterous being easy to fabricate by rapid manufacturing techniques.13–16
grasping performance have been developed based on differ- Besides, their structures largely composed of elastomeric
ent actuation techniques such as soft pneumatic actuators,2–5 polymers can be easily integrated with other adaptive func-
granular jamming,6,7 cable-driven mechanisms,8 dielectric tional components.17,18
elastomer actuators (DEAs),9,10 and shape memory alloy SMA-based gripper and hand had been developed by
(SMA) actuators.11,12 Among these approaches, soft pneu- embedding thin SMA wires eccentrically in a polymeric
matic actuators and granular jamming powered by compressed matrix with flexible reinforcements.11,12 However, the main

1
Department of Mechanical and Aerospace Engineering, Seoul National University, Seoul, Korea.
2
Institute of Advanced Machines and Design, Seoul National University, Seoul, Korea.

379
 380 WANG AND AHN

limitations of these designs are their inherent softness and Design and Fabrication of Proposed Gripper
small actuation force, which prevent them from grasping ob- Materials
jects with large weight. To improve the grasping force, thin
SMA wires grouped mechanically in parallel as SMA bundle The basic design used in this work is that of a self-activated
was used to enable the robotic finger to achieve higher actua- soft finger actuator possessing both soft morphing and shape
tion force.19 SMA strip had also been used to design robotic retention capabilities with two locally bendable hinge sections
fingers with larger grasping force, which is because SMA strip embedded with stiffness changeable materials. The schematic
with a much larger cross-sectional area than SMA wire could of the finger actuator and all its components are described in
generate greater output force.20,21 However, all the SMA-based Figure 1A. The finger actuator mainly combines smart materi-
fingers need to maintain the desired deformation with constant als, a flexible polymeric matrix and embedded stiff components.
energy input. In addition, all the fingers can only switch their There are two different types of smart material, SMA and
configuration between a straight state and a bending state and shape memory polymer (SMP), which are used for com-
they cannot adjust the posture of the finger to change the ac- pletely different purposes. SMA is only used as tendon ac-
tuation behavior of the manipulator for specific grasping tasks. tuator to generate bending deformations12 and SMP is only
From another perspective, soft robotics technologies of used as the stiffness changeable material to adjust the stiff-
variable compliance and deformability being able to be ness of the finger hinges.29–31 The transformation temperature
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

achieved by stiffness modulation can provide new solutions of the SMP is the glass transition temperature (Tg). Below its
for manipulator design to fulfill the mentioned shortcomings. Tg, SMP is stiff and in its high modulus state referred to as the
Inspired by biological systems such as octopus arm and glassy state, corresponding to the finger hinge in the high
kangaroo tail, stiffness modulation mechanisms had been stiffness state; above its Tg, SMP markedly softens and be-
developed based on different principles.7,22–25 Among these comes a low modulus state referred to as the rubbery state,
principles, various studies demonstrated that variable stiff- corresponding to the finger hinge in the low stiffness state. In
ness structures using embedded phase changeable materials this work, the SMA wire (55 wt% Ni, 45 wt% Ti, Flexinol,
can undergo a significant stiffness reduction by heating, and Dynalloy) with austenite start temperature (As) of 68C hav-
this type of concept can be easily integrated with other types ing a diameter of 0.15 mm was used due to its large force-
of structures with compact configurations.26–28 Hence, this weight ratio, negligible volume, and large life cycles.32 There
study aims to apply phase changeable material-based stiff- are many thermally activated phase changeable materials such
ness modulation mechanism to the design and fabrication of as shape memory materials, thermoplastics, and fusible al-
SMA-based soft manipulator with the capabilities of both loys. SMP (SMP Technologies) with its Tg of 55C was se-
compliant and safe contacts to environments in the low lected due to its high viscosity in rubbery state to prevent the
stiffness state and transmitting high force for effective in- possible material leaking, its significant change of elastic
teractions in the high stiffness state. modulus by two to three orders of magnitude under different
Our previous work has demonstrated an SMA-based soft phases, its nontoxicity, its Tg falls within the temperature
robotic gripper composed of three fingers exhibiting a good range of the polydimethylsiloxane (PDMS) matrix and is
passively adaptive capability for grasping objects with a large lower than the As of the SMA; moreover, SMP structures can
range of dimensions, however, the inherent softness of the be easily fabricated by rapid fabrication techniques such as
finger hinge hinders the gripper from effectively grasping three-dimensional (3D) printing and molding method.29,33,34
objects with a larger weight.12 This work aims to apply the Ni-chrome (Ni-Cr, 80 wt% Ni, 20 wt% Cr) wire with a
principle of stiffness modulation to the same gripper to en- diameter of 0.15 mm is used to heat up the SMP structure to
able it to be capable of performing effective and stable reach its Tg to change the SMP from the glassy state to rub-
grasping for objects with various shapes and a large range of bery state.20 PDMS (Sylgard 184, Dow Corning) is used as
weight. This was accomplished by embedding variable the polymeric matrix of all finger actuators and acrylonitrile
stiffness materials into the soft hinges of the robotic finger to butadiene styrene (ABS) is used for manufacturing the em-
enable it to be capable of both soft morphing in the low bedded stiff components and mold of all actuators by 3D
stiffness state and shape retention in the high stiffness state. printing through a fused deposition modeling (FDM) ma-
Based on this, the fingers with low stiffness can easily deform chine. In addition, a thin and soft Ecoflex overlay is cast on
and adapt to the shape of an object to grasp, then after altering the surface of the arc-shaped end of the finger actuator to
the stiffness, the fingers with high stiffness can keep the mimic the human finger pad, which is the fleshy mass on the
adaptively deformed shape to effectively lift up the objects. palmar aspect of the extremity of the finger (Fig. 1B). Ecoflex
In addition, the inner side of the fingertip was covered with a 00-30 (SmoothOn) with its inherent properties of low hard-
thin layer of soft elastic materials to mimic the functions of ness (much softer than PDMS) and high elasticity was se-
the human finger pad. The soft elastic cover enables the de- lected to achieve two main purposes.35,36 One is to make the
formation of the contacting surface to achieve a larger contact robotic finger pad highly deformable to obtain a good contact
area and a large friction force thus improving grasp stability. with the grasping objects, the other one is to improve the
Each robotic finger with two hinges can actively achieve coefficient of friction (COF) to obtain a larger grasping force.
different postures by both selectively changing the stiffness All the used materials are commercially available and the
of the hinges and actuating the relevant SMA wire. This main properties of the materials are shown in Table 1.
characteristic of the finger with multiple possible postures
can always enable the gripper to use the finger pad to grasp
Design and fabrication
objects. As a final demonstration, the robotic fingers with
desired shape-retained posture enable the gripper to suc- The robotic finger is designed to have a rectangular shape
cessfully pick up a frustum-shaped object. with one arc-shaped end referred to as fingertip, where the
 SMA-BASED SOFT GRIPPER WITH VARIABLE STIFFNESS 381
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

FIG. 1. Soft robotic finger and assembled gripper. (A) Schematic of the soft finger and all its components. The inset shows
the cross section of the finger indicating the relative positions of SMA and SMP. (B) The reference of human fingers and the
inset shows the schematic of a human fingertip. (C) The top view of the fabricated soft finger structure with all the
components, scale bar is 10 mm. (C) The actual appearance of the gripper without actuation, scale bar is 40 mm. SMA,
shape memory alloy; SMP, shape memory polymer. Color images available online at www.liebertpub.com/soro

overall dimensions are 120 mm in length, 20 mm in width, distance of dSMA of 1 mm for bidirectional bending actuation.
and 3 mm in thickness (Fig. 1A). Each finger contains two Two SMP structures are placed symmetrically in each hinge
hinges where the length of each hinge segment is 15 mm and section in parallel to the SMA wires and the relative positions
the distance between the two hinges is 20 mm. The hinge of SMA and SMP are shown in the schematic of the cross
being close to the fingertip is referred to as hinge-1 and the section of the actuator in the inset of Figure 1A. The dimen-
other one is referred to as hinge-2. sions of each SMP structure are 15 mm in length, 5 mm in
The proposed finger actuator has two SMA wires, referred width, and 1.5 mm in thickness with a 4 mm latch inserted in
to as SMA-1 and SMA-2, embedded in the PDMS matrix the stiff components at each end to fix the position of the SMP
along the x-axis close to the upper and lower surfaces. The two structure and to avoid the small interspace between the SMP
SMA wires are prestrained and placed symmetrically about and the stiff component.
the neutral plane of the actuator with a fixed off-centered At ambient temperatures, the SMP is in the glassy state and
the hinge is stiff and inflexible being able to resist external
Table 1. Main Material Properties forces, while with increased temperature by heating the SMP
through Ni-Cr wires exceeding its Tg, the SMP will change to
Parameters Values rubbery state and the hinge will become soft and flexible. In
addition, the stiffness of each hinge of the finger can be in-
SMA martensitic elastic modulus (EM) 28 GPa dependently changed by activating the relevant Ni-Cr wires.
SMA austenitic elastic modulus (EA) 75 GPa With hinge in the low stiffness state, applying an electric
SMA martensitic start temp. (Ms) 52C
SMA martensitic finish temp. (Mf) 42C current to one of the two embedded SMA wire causes its
SMA austenite start temp. (As) 68C temperature to increase through Joule heating and, when it
SMA austenite finish temp. (Af) 78C reaches its Af, the actuated SMA wire will start to contract in
SMA resistance per meter 55 O the matrix to produce a bending moment resulting in a large
SMP glassy elastic modulus (EG) 2.15 GPa bending deformation concentrated on the hinge sections in
SMP rubbery elastic modulus (ER) 10 MPa low stiffness.11 Actuating either SMA-1 or SMA-2 will result
SMP glass transition temp. (Tg) 55C in either an upward or downward bending deformation de-
PDMS temperature range 200C pending on which SMA wire is actuated. Then, electric
PDMS specific gravity 1.03 current continues to be applied to the actuated SMA wire to
PDMS elastic modulus (EPDMS) 1.84 MPa retain the deformation until the softened SMP changes back
PDMS hardness (Shore A) 46
Ecoflex hardness (Shore A) 30 from the rubbery state to the glassy state. Afterwards, the
Ni-Cr resistance per meter 61.68 O deformed postures of the robotic finger will be maintained
with a high stiffness. Once the SMP structures were softened
SMP, shape memory polymer. again, the deformed finger will be restored to a straight
 382 WANG AND AHN

configuration, by means of the elastic recovery force of the Results and Discussion
PDMS matrix. Variable stiffness performance of robotic finger
The manufacturing process of each robotic finger actuator
is started with building an ABS overall mold by FDM with The schematic of modulus variation of SMP and SMA
the small holes for positioning of the SMA wire within the against a temperature scale is illustrated in Figure 2A. The
actuator structure. The SMP structures are fabricated by 3D hinge of the finger is capable of varying its stiffness by
printing and then are intertwined with the Ni-Cr wire for heating and free cooling the embedded SMP structures. To
heating. After that, the SMP structures are positioned in determine the heating and cooling time, an electric current of
the 3D-printed stiff components, and then, the components 0.3 A (0.98 W for each SMP structure) was applied to the Ni-
are placed in the overall mold. The SMA wire is placed in the Cr wire for heating the SMP to a temperature above its Tg to
mold through its holes, prestrained, and held in place with the change it from glassy state to rubbery state resulting in the
help of bolts attached to a fixture. Lastly, PDMS is mixed significant stiffness decrease of the finger structure. Then,
with a base-to-catalyst components ratio of 10:1, poured into after free cooling of the heated SMP, the hinge turns back to
the mold and cured for 8 h at 55C, which is below the As of the high stiffness state. A thermal camera (CX-320U, Han-
the SMA wires. After curing of the polymer, the whole mac) at ambient temperature (27C) is used to detect the
structure can be obtained after removing the overall external temperature change of the positions of SMP and SMA (point
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

mold. The last step will be casting a thin Ecoflex layer with a A and point B, Fig. 1C) in real-time and the result is shown in
thickness of 1 mm, through a small ABS mold without the Figure 2B. As can be seen from the results, the temperature of
bottom, on the surface of the arc-shaped end of the finger SMP increases fast and it takes around 30 s to achieve its
actuator. The fabricated finger actuator with all its compo- glass-transition temperature of 55C. During this study, the
nents is illustrated in Figure 1C. Then, three identical fingers SMP is heated up to 60C using around 35 s to confirm the
are mounted in a triangular pattern to a metacarpal plate whole SMP structure is changed to rubbery phase, and a
with a fixed length of 25 mm for each finger so that a pair of cooling time of 240 s is needed for the SMP changing back to
fingers locates in opposite to a third finger at an angle of 60 its glassy phase. It is notable that during the whole heating
to mimic the human thumb, index finger, and middle finger process of the SMP, the temperature of point B never exceeds
(Fig. 1D). the As of SMA.

FIG. 2. Variable stiffness performance of robotic finger. (A) Schematic of SMP and SMA modulus comparison on a
temperature scale with particular emphasis of their transformation temperatures. (B) Temperature variation by heating the
SMP from the beginning to time t1 and cooling the SMP from time t1 until time t2. The inset shows the thermal image of the
hinge structure during the heating process. (C) Top view of the schematic of the composed structure with the same
configuration of the hinge. The inset shows the schematic and geometric dimension of the cross section of the structure.
(D) Results of the three-point-bending test of the composed structure with embedded SMP in the rubbery state and in the
glassy state. Color images available online at www.liebertpub.com/soro
 SMA-BASED SOFT GRIPPER WITH VARIABLE STIFFNESS 383

With a negligible volume of the embedded SMA wires, the deformed configurations under the influence of self-gravity
bending properties of the hinge are mainly determined by the by varying the stiffness locally.37 However, the multiple
characteristics of PDMS matrix and the embedded SMP deformed configurations of the robotic finger in this research
structures. To evaluate the bending modulus of the hinge can be actively obtained from the actuation of the two ec-
structure, a composed structure having the same configuration centrically embedded SMA wires by sequentially changing
with no SMA wires is fabricated. The schematic of the struc- the stiffness of the selected hinges.38,39 Applying an electric
ture and its geometric parameter of the cross-section area are current of 0.65 A (2.79 W) to one embedded SMA wire
described in Figure 2C. Then, the bending rigidity of the hinge causes increasing of the temperature through Joule heating to
is given by the sum of bending rigidity in each component and reach its As; the eccentrically placed SMA wire starts to
can be calculated by Equation (1) where EH, EPDMS, and ESMP contract in the PDMS matrix to produce a bending moment
denote the elastic modulus of the hinge structure, PDMS and resulting in a large bending deformation to the eccentric side
SMP, respectively. IH, IPDMS, and ISMP denote the second concentrated on the low-stiffness hinge segments and the
moment of area of the structures of the hinge, PDMS and SMP. whole process will take 5 s. The mechanical modeling of
hinge actuator for maximum bending deformation by using
EH IH ¼ EPDMS IPDMS þ ESMP ISMP (1) Brinson modeling and Euler beam theory was developed and
described in the previous studies.13,40,41
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

where IH ¼ w1 (2t1 þ t2 )3 =12, IPDMS ¼ w1 t1 3 =6 þ w1 t1 (t1 þ Each hinge actuator is capable of maintaining three con-
t2 )2 =2 þ (w1  2w2 )t2 3 =12, ISMP ¼ w2 t2 3 =6 and IH = IPDMS+ figurations: straight, upward bending, and downward bending.
ISMP. Then, Equation (1) can be rewritten as Equation (2). Then, each robotic finger composed of two hinge actuators is
capable of deforming into a greater amount of shapes of in-
EH ¼ (1  a)EPDMS þ aESMP (2) creased complexity and retaining those shapes. Thus, de-
pending on which hinge is actuated, the finger actuator is
where a ¼ ISMP =IH . With fixed design parameters (Fig. 2C), the capable of achieving nine different configurations (Fig. 3).
bending modulus of the hinge EH was theoretically calculated Figure 3A shows the original straight configuration of the
as 136.10 and 2.35 MPa in high and low stiffness states, which robotic finger. Figure 3B–G show the bending deformation of
correspond to the embedded SMP with glassy and rubbery a single finger actuator under the actuation of one SMA wire
states. With ESMP in glassy state is much larger than EPDMS, it (SMA-1 or SMA-2) with hinge-1, hinge-2, and both of the two
can be seen from Equation (2) that the maximum EH is mainly hinges in low stiffness state, respectively. Figure 3H, I show
determined by ISMP. That is, the bending rigidity of the hinge the opposite deformation of hinge-1 and hinge-2, which is
structure can be easily tailored by modifying the volume, po- achieved by first selectively reducing the stiffness of one
sition, and geometries of the embedded SMP structure. hinge (hinge-1 or hinge-2), which takes 35 s, and then actu-
The bending modulus of the composed structure was also ating one SMA wire (SMA-1 or SMA-2) to bend the hinge in
tested under these two different stiffness states, using a three- the low stiffness state until its stiffness changes back to the
point-bending test according to the standard test method high state (such as one configuration is Figure 3B, C, E, and
(ASTM D 790-03). The test was done using a span-to-depth F), which takes 240 s; after that, the other hinge is softened
ratio of 16:1 with a support span L of 48 mm on a universal and the other SMA wire is actuated to obtain the retained
testing machine (5948 MicroTester, Instron) with a three- opposite bending deformation concentrated on the other
point flexure fixture (2810-400, Instron). The results are hinge, which takes another 275 s. Besides, the robotic finger
shown in Figure 2D, and EH of the structure can be calculated was tested again after continuously switching the stiffness of
using the followed Equation (3) where m is the gradient of the both hinges between the low state and the high state for
load-displacement curve (Fig. 2D). around 8 h (100 cycles), and results showed that the finger still
has a good capability for achieving multiple deformed con-
L3 m figurations.
EH ¼ (3)
4w1 (2t1 þ t2 )3
Soft gripper and its grasping performance
Experimental results show that the hinge structure has a
bending modulus of 128.48 MPa in the high stiffness state and The gripper with three robotic fingers, by means of active
2.31 MPa in the low stiffness state, which corresponds to a ratio soft morphing and stiffness modulation in one compact de-
of *55.62 times. That is, the overall loss of the bending modulus sign, was designed to mimic the human thumb, index finger,
from low temperature to high temperature was around 98%. and middle finger to be able to grip a wide variety of objects.
Compared with the theoretical and experimental results, it can be All the three SMA-1s of the three fingers of the gripper are
seen that the theoretical calculation is able to predict the stiffness connected in series to be actuated simultaneously with in-
of the hinge structure in the two different states. However, there ward deformations for grasping, regardless of the target ob-
are still some differences (mean error 5.6% and 1.7%) between jects. All the three SMA-2s are also connected in series to be
the theoretical and experimental results, which could be due to actuated simultaneously with outward deformations to relax
the fabrication error of the SMP structure and the position error the objects. Either SMA-1s or SMA-2s is actuated using a
of the placement of embedded SMP in the PDMS matrix. uniform electrical current of 0.65 A. Besides, the three hinge-
1s and three hinge-2s of the gripper are connected in series
separately to selectively change their stiffness.
Multiple deformed configurations of robotic finger
Grasping the deformable objects and unpredictable objects
Variable stiffness composite structures have been studied is particularly difficult for hard grippers due to the stringent
in previous research, which can passively achieve different requirement of equilibrium and the high computational cost
 384 WANG AND AHN

FIG. 3. Nine different

configurations of one robotic
finger. (A) The original
straight configuration. (B-I)
The deformed configurations.
Insets show the schematic of
the robotic finger where red
and blue blocks indicate the
hinges in the low stiffness
state and in the high stiffness
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

state respectively. Blue and

red lines indicate the SMA
wires are currently under un-
actuation and actuation re-
spectively. Scale bar is 20 mm.
Color images available online
at www.liebertpub.com/soro

of modeling.42 However, soft gripper exploits the compliance universal testing machine to measure the force generated by
and flexibility of materials with a high number of degrees of the gripper on a fixed object during an upward motion. The
freedom showing the capabilities of gripping the objects with test begins by placing the gripper to grab a fixed cylindrical
varying shapes including the deformable objects. Besides, the object with a diameter of 80 mm, as shown in the inset in
gripper can facilely lift up the objects with large weight by Figure 4G. Then, the gripper is moved upward under unin-
maintaining the adaptively deformed shapes with high stiff- terrupted actuation. From grabbing the object to being sep-
ness. The soft gripper is first used to grip a deformable and arated from the object, the vertical displacement of the
lightweight hollow cylinder (2.1 g) with a thickness of the wall gripper is around 32.5 mm and the vertical pulling force
of 0.1 mm. The gripper picks up the object after achieving the generated by the gripper is shown in Figure 4G. During the
equilibrium between the force from the fingers and the rebound pulling process, it is notable that the gripper with low stiff-
elastic force from the deformed structure (Fig. 4A). A similar ness adapts its grasp from a caging grasping to a fingertip
grasping strategy by human fingers is described in Figure 4B. grasping depending on the changing situations, and the
This gripper can be also used for gripping fragile and irregu- pulling force undergoes a smooth increase process and then a
larly shaped objects such as an uncooked egg (51.9 g) shown in smooth decrease process. However, the gripper with high
Figure 4C; because of the small actuation force from the finger stiffness cannot adjust its configuration because of the re-
actuator, the egg was not damaged during manipulation. Re- tained shape obtained once finishing the grasping, and the
ferring to human fingers (Fig. 4D), it can be seen that the pulling force quickly goes up and then drops abruptly after
fingers softly grip the middle lower part of the egg using small the fingers surmounting the middle of the fixed cylinder.
constant force. The third test for robust grasping was con- Results show that the largest pulling forces for the gripper
ducted to lift up a red pepper with a much larger volume and with low stiffness and with high stiffness are 0.6 N and 5.8 N
weight (143.2 g) shown in Figure 4E. The corresponded pos- respectively, corresponding to a ratio of *10 times. From
tures of human fingers in Figure 4F show that the fingers another aspect, if the pulling force is larger, the ability of the
through finger pads applied the force directly to the pepper peel gripper to resist external force is better for being able to lift up
to increase the friction and then to pick up the object. the objects with larger weight.

Maximum grasping force Grasping strategy of frustum-shaped object

To further compare the grasping capability of the grippers While grasping process, the posture of human fingers is not
with low stiffness and high stiffness, an experiment was always immutable. In contrast, they demonstrate different
conducted with the gripper installed to the previously used prepared postures depending on the shape of the targeted
 SMA-BASED SOFT GRIPPER WITH VARIABLE STIFFNESS 385

FIG. 4. Grasping performance. (A–F)

Postures of the gripper and human fin-
gers while gripping different objects.
(G) Pulling force of the gripper from
grasping to releasing the object. In-
set: the schematic of the experimental
setup. Scale bar is 30 mm. Color images
available online at www.liebertpub
.com/soro
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

objects. Keeping the immutable posture of the finger may obliges them to lift up a frustum-shaped rigid object easily
cause invalid grasping. For instance, grasping can be hardly and stably through fingernail grasping (Fig. 5E–H, Supple-
possible to human fingers to grip a frustum-shaped object mentary Movie S1; Supplementary Data are available online
through fingernails (Fig. 5A–D). The same invalid grasping at www.liebertpub.com/soro). Grasping the frustum-shaped
happens to both the soft and the articular fingered grippers object through fingernail is difficult, because the fingertip is
where their predesigned and unchangeable configuration always stiff, is not deformable, and is in point-contact with the

FIG. 5. Invalid grasping of a frustum-shaped object via fingernails. (A–D) The grasping process of human fingers. (E–H)
The corresponding grasping process of the soft robotic gripper. Scale bar is 30 mm. Color images available online at www
.liebertpub.com/soro
 386 WANG AND AHN

FIG. 6. Force analyses of grasping a

frustum-shaped object. (A) Case A: the fin-
gernail grasping. (B) Case B: the finger pad
grasping. Color images available online at
www.liebertpub.com/soro
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

object; while in this case, the bending of the fingers is neces- show that the static and dynamic COFs of PDMS-ABS (4.56
sary not only for providing enough lifting force (friction force), and 2.92) are *1.86 and 1.34 times of those of Ecoflex-ABS
but for maintaining the equilibrium of the applied set of for- (2.45 and 2.18), respectively. With h is 80 and assuming that
ces.43 However, grasping the frustum-shaped object using the grasping forces for these two cases are equal, the maxi-
finger pads provides better performance. The softness of the mum grasping weight that the gripper can pick up using
finger pad increases the contact area with larger friction, in- finger pad grasping is *1.93 times of that using fingernail
creasing the capability of exerting moment around the surface grasping. Moreover, the finger pad grasping with a much
normal at the contact, thus improving grasp stability.44 larger contact area with the object results in a much better
The force analyses of grasping a frustum-shaped object grasping stability than the fingernail grasping.
through fingernail grasping (case A) and finger pad grasping As discussed above, the better strategy for human fingers to
(case B) of the designed gripper are described in Figure 6. For successfully lift up a frustum-shaped object is using finger pad,
case A, the fingernails made of PDMS is in point contact with the pulp of finger, to pinch the upper part of the object, and the
the object; while for case B, the finger pads made of Ecoflex corresponding continuous motions are shown in Figure 7A–C.
is in area contact with the object. Supposing the maximum During the grasping process, the human finger pads are pas-
weight of the object that the gripper can pick up is Ga for case sively deformed for obtaining large contact areas with the
A and Gb for case B, then the force equilibrium for both object, which can provide enough upward friction forces and
successful grasping manners in the vertical direction can be easily maintain the force equilibrium. Another remarkable
calculated using the followed Equations (4) and (5) where h is characteristic of the robotic finger with variable stiffness as
the inclination angle of the frustum object. described is being capable of achieving diverse postures
(Fig. 3) that can endow the gripper with the desired preparative
Ga ¼ 2Fa (la sin h  cos h) (4) configurations to mimic the human fingers to grasp a frustum-
shaped object, which cannot be achieved by using the same
gripper without stiffening mechanisms. The successful
Gb ¼ 2Fb (lb sin h  cos h) (5) grasping process of the gripper is described in Figure 7D–7I
(Supplementary Movie S1). It can be seen that before grasping,
As can be seen from the above equations, the grasping the gripper adjusts its posture by only reducing the stiffness of
capability for the maximum grasping weight is determined by hinge-2 of each finger and deforming hinge-2 inward with
both the grasping force (Fa for case A, Fb for case B) provided uninterrupted actuation of SMA-1s until the deformed shape is
by the robotic finger and the static COF (la for case A, lb for retained. After that, reducing the stiffness of all three hinge-1
case B) of the contact area. The exerted force of the soft finger and actuating all three SMA-2 enable the fingers to deform in
is inherently small (around 10-3 to 10-1 N) due to its soft and reverse direction to open the gripper. Once the gripper ap-
compliant structure and the small bending force generated by proaches the object from above and comes into contact with it,
the SMA wire.12,14,45 Hence, the maximum grasping force is un-actuating SMA-2s and actuating SMA-1s again enables the
mainly determined by the COF between the fingertips and the hinge-1s to bend inward to grab the object using finger pads,
object, which is highly sensitive to surface texture and con- and then to lift it up.
tact area. Experiments were conducted to measure the COF of
PDMS and Ecoflex film structures to the 3D-printed ABS
Conclusions
sheet (the frustum object was made of ABS by the same 3D
printer). These experiments were conducted according to the This study first demonstrated a soft gripper composed of
ASTM D 1894 standard, using the same universal testing three SMA-based robotic fingers with variable stiffness for
machine with a friction fixture (2810-005, Instron). Results adaptive and effective grasping of a variety of objects. The
 SMA-BASED SOFT GRIPPER WITH VARIABLE STIFFNESS 387

FIG. 7. Grasping strategy for a

frustum-shaped object. (A–C) The
continuous motion of human fingers
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

for grasping the frustum-shaped

object. (D–I) The continuous mo-
tion of the soft gripper for grasping
the frustum-shaped object inspired
by the motion of human fingers.
Short blue curves indicate the hin-
ges with high stiffness; short red
curves indicate the hinges with low
stiffness. The arrows show the
bending direction of the hinge under
actuation. Scale bar is 30 mm. Color
images available online at www
.liebertpub.com/soro

robotic finger was fabricated by embedding the 3D-printed the volume, position, and geometries of the embedded SMP
SMP structures in the hinge sections of an SMA-based hinge structure. Afterward, three fingers were assembled to form a
actuator to enable it to be capable of both soft morphing and bioinspired gripper for grasping the deformable objects and
shape retentions in multiple shapes. There are two main objects with varying shapes, where the largest grasping force
characteristics of the robotic finger: one is that it can signif- of the gripper in the high stiffness state is around 10 times of
icantly change the structural stiffness, and the other one is that in the low stiffness state. In the last test, the soft gripper
that it can actively achieve diverse postures by both changing was adapted to grip the frustum-shaped object that was dif-
the stiffness of the desired hinge and deforming the corre- ficult to achieve using a soft-fingered gripper without the
sponding hinge to locally deform the finger structure. The function of stiffness modulation.
bending modulus of the hinge of the robotic finger was tested The proposed gripper based on the integration of different
and the result shows that the bending modulus of the hinge in functional components provides a new solution to the soft
the high stiffness state is 128.48 MPa and that in the low manipulators to make them be more adaptable and effective
stiffness state is 2.31 MPa, which corresponds to a ratio of for different grasping tasks. This study through introducing
*55 times. The time for changing the finger from the high stiffness modulation enables the soft gripper to be adaptable
stiffness state to the low stiffness state was determined as 40 s to grasp objects with a larger range of weight and with more
and that for the inverse process was determined as 240 s. The varying shapes. Besides, the choice of using SMA for ac-
advantage of this composed finger structure is that the bending tuation and SMP for changing stiffness was motivated by
properties of this structure can be easily tailored by modifying the fact that such materials are easy to be designed and
 388 WANG AND AHN

fabricated in a compact and simple design. Moreover, dif- 10. Shian S, Bertoldi K, Clarke DR. Dielectric elastomer based
ferent types of actuation and stiffness changeable principle ‘‘grippers’’ for soft robotics. Adv Mater 2015;27:6814–
can be also applied to the soft gripper. Besides, the signif- 6819.
icant limitation of the soft gripper in this work is the long 11. Furuya Y, Shimada H. Shape memory actuators for robotic
time for its changing from the low stiffness state back to applications. Mater Des 1991;12:21–28.
the high stiffness state due to the slow heat dissipation of 12. Wang W, Rodrigue H, Kim HI, Han MW, Ahn SH. Soft
the heated SMP. Future research will focus on optimizing composite hinge actuator and application to compliant ro-
the design of the finger structure to shorten the process time botic gripper. Compos B Eng 2016;98:397–405.
for stiffness modulation, and applying different stiffness 13. Wang W, Rodrigue H, Ahn SH. Deployable soft composite
changeable principles. structures. Sci Rep 2016;6:20869.
14. Song SH, Kim MS, Rodrigue H, Lee JY, Shim JE, Kim
MC, et al. Turtle mimetic soft robot with two swimming
Acknowledgments gaits. Bioinspiration Biomimetics 2016;11:036010.
This work was supported by the Industrial Strategic Tech- 15. Shim EJ, Quan Y, Wang W, Rodrigue H, Song SH, Ahn
nology Development Program (10049258) funded by the Min- SH. A smart soft actuator using a single shape memory
istry of Knowledge Economy (MKE), Korea and the National alloy for twisting actuation. Smart Mater Struct 2015;24:
Research Foundation of Korea (NRF) grants funded by the 125033.
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

Korea government (MEST) (NRF-2015R1A2A1A13027910) 16. Rodrigue H, Wang W, Han MW, Kim JY, Ahn SH. An
and (MSIP) (NRF-2016R1A5A1938472). The authors thank overview of shape memory alloy-coupled actuators and
the reviewers of this article for their insightful feedback. Finally, robots. Soft Rob 2017;4:3–15.
17. Wang W, Kim NG, Rodrigue H, Ahn SH. Modular as-
Wang Wei wants to thank Miss Li Jingxuan for the input on the
sembly of soft deployable structures and robots. Mater
experiment demonstration, and moreover, for the invaluable Horiz 2017;4:367–376.
support over the years. 18. Wang W, Li C, Rodrigue H, Yuan F, Han MW, Cho M,
et al. Kirigami/origami-based soft deployable reflector
Author Contributions for optical beam steering. Adv Funct Mater 2017;27:
S.H.A. and W.W. conceived the idea. S.H.A. supervised 1604214.
the research. W.W. designed the study, performed the ex- 19. Simone F, York A, Seelecke S. Design and fabrication of a
periments, analyzed the data, and wrote the article, and all three-finger prosthetic hand using SMA muscle wires. SPIE
Smart Structures and Materials+Nondestructive Evaluation
authors commented on the article.
and Health Monitoring 2015;94290T.
20. She Y, Li C, Cleary J, Su HJ. Design and fabrication of a
Author Disclosure Statement soft robotic hand with embedded actuators and sensors. J
No competing financial interests exist. Mech Rob 2015;7:021007.
21. Engeberg ED, Dilibal S, Vatani M, Choi JW, Lavery J.
References Anthropomorphic finger antagonistically actuated by SMA
plates. Bioinspiration Biomimetics 2015;10:056002.
1. Hughes J, Culha U, Giardina F, Guenther F, Rosendo A, 22. Wall V, Deimel R, Brock O. Selective stiffening of soft
Lida F. Soft manipulators and grippers: a review. Front Rob actuators based on jamming. International Conference on
AI 2016;3:69. Robotics and Automation (ICRA), Seattle, WA, 2015,
2. Galloway KC, Becker KP, Phillips B, Kirby J, Licht S, pp. 252–257.
Tchernov D, et al. Soft robotic grippers for biological 23. Kim YJ, Cheng S, Kim S, Iagnemma K. A novel layer
sampling on deep reefs. Soft Rob 2016;3:23–33. jamming mechanism with tunable stiffness capability for
3. Ilievski F, Mazzeo AD, Shepherd RF, Chen X, Whitesides GM. minimally invasive surgery. IEEE Trans Rob 2013;29:
Soft robotics for chemists. Angew Chem 2011;123:1930–1935. 1031–1042.
4. Deimel R, Brock O. A novel type of compliant and un- 24. Wei Y, Chen Y, Ren T, Chen Q, Yan C, Yang Y, et al. A novel,
deractuated robotic hand for dexterous grasping. Int J Rob variable stiffness robotic gripper based on integrated soft ac-
Res 2016;35:161–185. tuating and particle jamming. Soft Rob 2016;3:134–143.
5. Zhao H, O’Brien K, Li S, Shepherd RF. Optoelectronically 25. Manti M, Cacucciolo V, Cianchetti M, Stiffening in soft
innervated soft prosthetic hand via stretchable optical robotics: a review of the state of the art. IEEE Rob Autom
waveguides. Sci Rob 2016;1:eaai7529. Mag 2016;23:93–106.
6. Brown E, Rodenberg N, Amend J, Mozeikac A, Steltzc E, 26. Schubert BE, Floreano D. Variable stiffness material based
Zakind MR, et al. Universal robotic gripper based on the on rigid low-melting-point-alloy microstructures embedded
jamming of granular material. Proc Natl Acad Sci 2010;107: in soft poly (dimethylsiloxane) (PDMS). RSC Adv 2013;
18809–18814. 3:24671–24679.
7. Amend J, Lipson H. The jamhand: dexterous manipulation 27. Wang W, Rodrigue H, Ahn SH. Smart soft composite ac-
with minimal actuation. Soft Rob 2017;4:70–80. tuator with shape retention capability using embedded
8. Manti M, Hassan T, Passetti G, D’Elia N, Laschi C, Cian- fusible alloy structures. Compos B Eng 2015;78:507–514.
chetti M. A bioinspired soft robotic gripper for adaptable and 28. Cheng NG, Gopinath A, Wang L, Iagnemma K, Hosoi AE.
effective grasping. Soft Rob 2015;2:107–116. Thermally tunable, self-healing composites for soft robotic
9. Shintake J, Rosset S, Schubert B, Floreano D, Shea H. applications. Macromol Mater Eng 2014;299:1279–1284.
Versatile soft grippers with intrinsic electroadhesion based 29. Yang Y, Chen Y, Li Y, Chen ZQ, Wei Y. Bioinspired
on multifunctional polymer actuators. Adv Mater 2016;28: robotic fingers based on pneumatic actuator and 3D print-
231–238. ing of smart material. Soft Rob 2017;4:147–162.
 SMA-BASED SOFT GRIPPER WITH VARIABLE STIFFNESS 389

30. Hager MD, Bode S, Weber C, et al. Shape memory poly- constant material functions and redefined martensite internal
mers: past, present and future developments. Prog Polym variable. J Intell Mater Syst Struct 1993;4:229–242.
Sci 2015;49:3–33. 41. Elahinia MH, Ahmadian M. An enhanced SMA phenom-
31. Sun L, Huang WM, Ding Z, Wang CC, Purnawali H, Tang enological model: I. The shortcomings of the existing
C. Stimulus-responsive shape memory materials: a review. models. Smart Mater Struct 2005;14:1297.
Mater Des 2012;33:577–640. 42. Jia YB, Guo F, Lin H. Grasping deformable planar objects:
32. Huang W. On the selection of shape memory alloys for squeeze, stick/slip analysis, and energy-based optimalities.
actuators. Mater Des 2002;23:11–19. Int J Rob Res 2014;33:866–897.
33. Kim JS, Lee DY, Koh JS, Jung GP, Cho KJ. Component 43. Ghafoor A, Dai JS, Duffy J. Stiffness modeling of the soft-
assembly with shape memory polymer fastener for micro- finger contact in robotic grasping. J Mech Des 2004;126:
robots. Smart Mater Struct 2013;23:015011. 646–656.
34. Azra C, Plummer CJG, Manson JAE. Isothermal recovery 44. Murakami K, Hasegawa T. Tactile sensing of edge direc-
rates in shape memory polyurethanes. Smart Mater Struct tion of an object with a soft fingertip contact. International
2011;20:082002. Conference on Robotics and Automation (ICRA), Barce-
35. Case JC, White EL, Kramer RK. Soft material character- lona, Spain, 2005, pp. 2571–2577.
ization for robotic applications. Soft Robot 2015;2:80–87. 45. Kim HS, Lee JY, Chu WS, Ahn SH. Design and fabrication
36. Johnston ID, McCluskey DK, Tan CKL, Tracey MC. Me- of soft morphing ray propulsor: undulator and oscillator.
Downloaded by Seoul National Univ. Medical College from online.liebertpub.com at 12/20/17. For personal use only.

chanical characterization of bulk Sylgard 184 for micro- Soft Rob 2017;4:49–60.
fluidics and microengineering. J Micromech Microeng
2014;24:035017.
37. McEvoy MA, Correll N. Thermoplastic variable stiffness
composites with embedded, networked sensing, actuation, Address correspondence to:
and control. J Compos Mater 2015;49:1799–1808. Sung-Hoon Ahn
38. Wang W, Lee JY, Rodrigue H, Song SH, Chu WS, Ahn SH. Department of Mechanical and Aerospace Engineering
Locomotion of inchworm-inspired robot made of smart soft Seoul National University
composite (SSC). Bioinspiration Biomimetics 2014;9:046006. Gwanak-ro 566
39. Rodrigue H, Wang W, Bhandari B, Han MW, Ahn SH. SMA- Gwanak-gu
based smart soft composite structure capable of multiple Seoul 151-742
modes of actuation. Compos B Eng 2015;82:152–158. Korea
40. Brinson LC. One-dimensional constitutive behavior of shape
memory alloys: thermomechanical derivation with non- E-mail: [email protected]