See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341758467

3D printed, modularized rigid-flexible integrated soft finger actuators for

anthropomorphic hands

Article in Sensors and Actuators A Physical · May 2020

DOI: 10.1016/j.sna.2020.112090

CITATIONS READS

33 754

5 authors, including:

Ningbin Zhang Lisen Ge

Shanghai Jiao Tong University Shanghai Jiao Tong University
26 PUBLICATIONS 1,070 CITATIONS 11 PUBLICATIONS 635 CITATIONS

SEE PROFILE SEE PROFILE

Haipeng Xu
Shanghai Jiao Tong University
15 PUBLICATIONS 511 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ningbin Zhang on 09 June 2020.

The user has requested enhancement of the downloaded file.

 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx

Contents lists available at ScienceDirect

Sensors & Actuators: A. Physical

F
journal homepage: www.elsevier.com

OO
3D printed, modularized rigid-flexible integrated soft finger actuators for
anthropomorphic hands
Ningbin Zhang a, Lisen Ge a, Haipeng Xu a, Xiangyang Zhu a, b⁎, Guoying Gu a, b⁎
a

PR
Robotics Institute, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
b
State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University, Shanghai 200240, China

ARTICLE INFO ABSTRACT

Article history: Owing to the integrated muscular, ligamentous and skeletal structures and coupled degrees of freedom
Received 8 March 2020 (DoFs), it is a long-term challenge in the field of robotics to design an anthropomorphic hand that mimics the
Received in revised form 8 May 2020

ED
biological structures and dexterous motions of human hands. In this paper, we present pneumatical, multi-ma-
Accepted 19 May 2020
terial 3D-printed, modularized rigid-flexible integrated soft finger actuators (RFiSFAs) that can be directly
Available online xxx
assembled to an anthropomorphic hand. First, we introduce the mechanism of the RFiSFA with a pneumatic
bellow chamber and a joint structure, and investigate the influence of the chamber material and the bellow
Keywords: number on the flexion angles and output forces performances of the RFiSFA. Next, we design and fabricate
Rigid-flexible integrated soft finger actuator a 2-DoF flexion finger with two serial RFiSFAs and a 3-DoF thumb with two serial RFiSFAs and two par-
Pneumatical bellow structure allel RFiSFAs. Then, we perform tests to characterize the motion and force performance of the fingers and
Anthropomorphic hand thumb. Finally, we integrate and assemble an 11-DoF anthropomorphic hand with four flexion fingers and
CT
Multi-material 3D printing one thumb, and experimental results demonstrate the capability of the hand in grasping objects with different
Modular design and fabrication
dimensions, shapes and textures.
© 2020.

1. Introduction therly manually assembled with other flexible or rigid components.

However, both fabrication and assembly process are laboriously inten-
RE

Over countless generations, our hands have evolved integrated sive, and may be not suitable for scenarios requiring compact design
muscular, ligamentous and skeletal structures to perform dexterous and high consistency. To address these issues, three-dimensional (3D)
movements with coupled DoFs, which motivates inventors and sci- printing is utilized to create monolithic soft fingers or hands [18–20]
entists to create various anthropomorphic hands in the field of ro- with single flexible material. However, it is difficult to fully replicate
botics [1]. The success of these anthropomorphic hands, however, the articular motions of the human finger which is induced by large
mostly introduces complex rigid and flexible components by integrat- stiffness variations between biological joints and skeletons. To over-
R

ing numbers of motors, linkages, transmission gears, tendons, connec- come the stiffness limitation due to single material, multiple materi-
tors, electronic boards and controllers, usually with years of devel- als with a wide range of constituent moduli are introduced into the
opment for each prototype [2]. To simplify the structure and achieve 3D printing process for constructing robotic systems or components,
CO

rapid prototyping of anthropomorphic hands, some minimalist and including multilegged robots [21–23], combustion-powered jumpers
useful approaches are developed, such as underactuation [3–5], syner- [24], end effectors [25], soft actuators [2627] and hybrid stiffness-tun-
gies [4,6–8] and soft robotics [5,9–11]. able robots [28]. Recently, the multi-material 3D printing approach
Among these approaches, soft robotics renders the anthropomor- is employed for fabricating hand’s flexible joints and rigid skeletons.
phic hand distinctive features of inherent compliance and impact re- For example, as presented in [29], a soft skeleton hand with passive
sistance with the involvement of soft functional materials, which is joint and skeleton motions is designed and fabricated to perform pi-
partially or difficultly achieved by other approaches. Based on soft ano music. However, there is no active actuation modules in the devel-
UN

robotics, numerous “soft hands” are designed and fabricated rang- opment. To the best of our knowledge, there is rare study to develop
ing from the single-finger grippers to the multi-finger anthropomor- 3D-printed anthropomorphic hands with active and multiple DoFs.
phic hands [12–17]. Most soft hands have been designed with fin- The challenges not only lie in designing bioinspired finger actuators
gers integrating self-contained elastomeric actuators and deformable with rigid-flexible features, but also in decoupling hand DoFs and re-
structures, which are generally manually fabricated by replica mold- constructing them in a confined space.
ing process and its variant approaches. The fabricated fingers are fur In this paper, we propose the multi-material, 3D-printed pneu-
matic actuators with rigid-flexible features, termed as RFiSFAs, to
directly generate an anthropomorphic hand in a modularized manner
⁎
(see movie S1 for an illustration). The RFiSFA has a bioinspired,
Corresponding authors at: Robotics Institute, School of Mechanical Engineering,
Shanghai Jiao Tong University, Shanghai 200240, China.
hybrid structure integrating a pneumatic bellow chamber and joint
Email addresses: [email protected] (X. Zhu); [email protected] (G.
skeletons. We firstly investigate the influence of the chamber mater
Gu)

https://doi.org/10.1016/j.sna.2020.112090
0924-4247/ © 2020.
2 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx

F
OO
PR
Fig. 1. Schematic illustration of the RFiSFA. (a) Structure and actuation mechanism. (b) Dimension parameters. (c) Parameter values.

ED
CT
R RE

Fig. 2. Characterization on the mechanical properties of the printing materials. (a) Customized specimens with different hardness for tensile tests. (b) Mean stress (n = 3) of the speci-
mens plotted as a function of the strain.

ial and the bellow number on the flexion angles and output forces per- The remainder of this paper is organized as follows. Section 2
CO

formances of the RFiSFA. After that, we present the design and fab- demonstrates the design of the RFiSFA. Section 3 presents the modu-
rication of two classes of fingers: 2-DoF flexion fingers with two ser- lar design and fabrication of the multi-DoF fingers based on the RFiS-
ial RFiSFAs and 3-DoF thumb with two serial RFiSFAs and two par- FAs for an anthropomorphic hand. Section 4 characterizes the motion
allel RFiSFAs. Furtherly, we characterize the motion and force per- and force performance of the fingers. This section also presents the
formance of the fingers and thumb based on the prototype testing. Fi- mobility, dexterity and robustness of the fabricated 11-DoF anthropo-
nally, we develop an 11-DoF anthropomorphic hand with one thumb morphic hand. The conclusions and future work are presented in Sec-
and four flexion fingers. The experimental results show that the fab- tion 5.
UN

ricated hand is robust against proper hammer strike and capable of

grasping objects with different dimensions, shapes and textures under 2. Design of the RFiSFA
pneumatic control.
The main contribution of this paper is summarized as follows: 2.1. The structure and the actuation mechanism
(1) We design and fabricate modularized, pneumatically actuated RF- The motion of human hand can be approximately regarded as cou-
iSFAs through multi-material 3D printing; pled motions of multiple joint rotations [30]. Based on this fact, we
(2) We develop an 11-DoF anthropomorphic hand with one thumb design RFiSFA as 1-DoF active joint to perform analogous motions.
and four fingers made of 12 RFiSFAs. As shown in Fig. 1(a), an RFiSFA consists of a hollow, flexible bel-
(3) We evaluate the motion and force performance of the fingers and low-type chamber and two rigid joint levers connected by a flexi
the dexterity of the hand.
 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx 3

is constrained by the joint structure, thus generating the flexion angle

α and the output torque M. The flexion angle and the output torque of
the RFiSFA are not only related to the supplied air pressure, but also
to materials and structural dimensions [31]. In general, the selections
of the used materials depend on the 3D printing system, which will
be illustrated in Section 2.2. The dimension parameters and their val-

F
ues are shown in Fig. 1(b) and (c), respectively. Note that the length
of RFiSFA is calculated by Nh3+2l2, where N is the bellow num-

OO
ber. Some other parameters are designed considering the dimensions
of human finger joints, such as the width (2r2 = 16.12 mm) and the
height (l1+c1+2r2 = 18.12 mm) of the RFiSFA. The cross-section of the
chamber is non-uniformly designed (d1=d2 = 1.3 mm, d3 = 0.8 mm) to
reduce stress concentration and enhance working stroke simultane-
ously [21,31]. The inner radius of the chamber (r1 = 3 mm) should be
designed as large as possible for rapid removing of the supporting ma-
terial. In this work, we focus on investigating the effect of the cham-

PR
ber material and the bellow number on the performances of RFiSFA s
while other parameters are set to be constant for simplicity.
Fig. 3. Mean burst pressure (n = 3) and the standard deviation plotted as a function of
the material hardness of one-bellow actuator.
2.2. The selection of the chamber material
ble but inextensible sheet. An air inlet and an end cap are designed
for pneumatic connection and blocking, respectively. Upon pressur- We use the Polyjet 3D printer (J750, Stratasys Ltd., USA) and the
ization, the bellow-type chamber extends along the axial direction and GrabCAD Print software (version 1.28.16.50383) to fabricate the RF-

ED
iSFA. This printing system provides two basic series of materials in-
cluding Vero series (rigid) and Agilus30 (flexible), and numerous in
CT
R RE

Fig. 4. Characterization of the flexion angle of the RFiSFAs with different bellow number. (a) Still images of RFiSFAs with different bellow numbers at 16 kPa. (b) Flexion angle
plotted as a function of the supplied pressure. Error bars represent standard deviation over n = 3 trials.
CO
UN

Fig. 5. Characterization of the blocking force of the RFiSFAs with different bellow number. (a) Experimental setup. (b) Blocking force plotted as a function of the supplied pressure.
Error bars represent standard deviation over n = 3 trials.
4 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx

F
OO
Fig. 6. Design spaces of the maximum flexion angle and the chamber length. (a) Selection of the first RFiSFA. (b) Selection of the second RFiSFA.

PR
min. The specimens of different materials are shown in Fig. 2(a) and
their stress-strain curves are shown in Fig. 2(b). The results indicate
that an increase of material hardness results in an increase of the max-
imum stress (also the Young’s module), while a decrease of the max-
imum strain. Therefore, we should make a tradeoff between the max-
imum stress and the stroke. As an example, we design and fabricate a
series of one-bellow actuators without joint structures to measure their

ED
burst pressures (Fig. 3). We may mention that all the dimension para-
meters of a one-bellow chamber are in accordance with those corre-
sponding parts of the RFiSFA. In the test, we connect the actuators to
the air source controlled by the air regulator (ITV2050−312BL, SMC
Inc., Japan) and inflate the actuators with 2 kPa pressure increments
until they burst. The recorded burst pressures are shown in Fig. 3. The
results indicate that with the increase of material hardness, the actu-
CT
ator burst pressure increases. Particularly, the burst pressures of 60A
Fig. 7. The flexion angle and the blocking force of the selected RFiSFAs plotted as a
function of time. and 70A actuators are almost the same (22.9 kPa vs. 23.1 kPa). How-
ever, the standard deviation (SD) of 60A actuator is smaller than that
terpolated materials which are termed as Agilus30-Vero digital mate- of the 70A (1.1 kPa vs. 2.9 kPa). Therefore, we select 60A material in
rials. Their fundamental mechanical properties can be found in [32]. this work to fabricate the chamber and set a pressure limit of 16 kPa
To select a proper material for the chamber, we firstly perform uni- for the RFiSFA.
RE

axial tensile tests to investigate the mechanical properties of five print-

ing materials in the used 3D printer, including pure Agilus30 (hard-
2.3. The selection of the bellow number
ness = 30A) and Agilus30-Vero digital materials (hardness = 40A,
50A, 60A, 70A). In this test, the ZwickRoell Materials Testing Sys-
In the following, we will investigate the bellow number on the
tem (Model Z0020) is used with a tensile speed of 10 mm/
influences of the flexion angle, the blocking force, and the chamber
length of RFiSFAs.
R
CO
UN

Fig. 8. Modularized design of fingers. (a) The 2-DoF flexion finger. (b) The 3-DoF thumb.
 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx 5

F
OO
PR
Fig. 9. Design, fabrication and assembly of the anthropomorphic hand. (a) Schematic illustration of an anthropomorphic hand integrating three classes of components: active joints
generated with RFiSFAs (green), flexible fingertips and passive joints with pure flexible materials (yellow), and the rigid skeletons (grey). (b) Prototype of the fabricated anthropo-
morphic hand and assembly connections. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

ED
Fig. 10. Still image of the flexion motion of a 2-DoF finger.
CT
R RE
CO

Fig. 11. Characterization of the flexion angles. (a) Flexion angles of MCP joint and fingertip. (b) Relative error of the flexion angles.

To this end, we fabricate a series of RFiSFAs with different bellow sults in an increase of the flexion angle of each RFiSFA. The max-
numbers (i.e. 6, 7, 8, 9,10, and 11). The fabrication steps are similar to imum flexion angle is 47.3 ± 0.82° for the 6-bellow RFiSFA and
those of the one-bellow actuators. The only difference is that a flexible 99.0 ± 4.24° for the 11-bellow RFiSFA under the supplied pressure of
but inextensible sheet is required to connect two joint levers. In this 16 kPa.
work, we attach a tailored non-woven fabric (0.2 mm in thickness) to Next, we measure the blocking force of the RFiSFA with differ-
UN

the joint levers with the glue (A401, Ausbond Ltd., China) to realize ent bellow number. Fig. 5(a) shows the measurement setup (taking the
reliable joint structures. 11-bellow RFiSFA as an example). The proximal end of RFiSFA is
Then, we measure the flexion angle of each RFiSFA under sup- fixed on a platform and the tip is regulated in contact with the load cell
plied pressures. As shown in Fig. 4(a), the RFiSFA is mounted on a (ME K3D120, ME-Messsysteme Ltd., Germany). Upon pressuriza-
base and can flex freely. We attach 2 markers on the proximal lever tion, the motion of the RFiSFA is blocked by another constraint plat-
and 2 markers on the distal lever, respectively. The flexion angles form and the blocking force is recorded. The results in Fig. 5(b) show
are calculated based on the relative positions of the markers, which that an increase of bellow number results in an increase of blocking
is captured by two cameras (6D, Canon Ltd., USA). The Stereo cam- force. The maximum blocking forces are 0.33 ± 0.01 N for the 6-bel-
era calibration tool in Matlab 2018a is used for camera calibration. low RFiSFA and 0.44 ± 0.01 N for the 11-bellow RFiSFA under the
The results in Fig. 4(b) show that an increase of the bellow number re supplied pressure of 16 kPa.
6 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx

F
OO
PR
Fig. 12. Blocking force test setup for (a) the 2-DoF flexion finger and (b) the circumduction unit of thumb. (c) The blocking forces under increasing supplied air pressures.

ED
CT
R RE

Fig. 13. Experimental demonstration of the anthropomorphic hand. (a) Independent DoFs for fingers. (b) Compliant grasping for objects with different dimensions, shapes and tex-
tures. (c) Robustness test.

Table 1 The above experimental results demonstrate that both the flexion
CO

Comparison of the proposed hand with representative 3D printed, pneumatic soft hands, angle and the blocking force become larger with the increase of the
grippers or fingers.
bellow number. Due to the limited space in an anthropomorphic hand,
Active Printing Printing a tradeoff between the maximum flexion angle/maximum blocking
DoF(s) method material(s) force and the chamber length should be made. In the current work, we
This work 11 Multi- Agilus30 series and
should design two kinds of RFiSFAs for the anthropomorphic hand.
material Vero series The first RFiSFA is used to drive the metacarpophalangeal (MCP)
Three-finger soft gripper 9 Multi- Tango Black + and joint, the proximal interphalangeal (PIP) joint or the joint for thumb
UN

[22] material Vero Clear adduction whose maximum flexion angles should be within 60° to 90°
Soft robotic hand [18] 8 Single TPU
[30]. The second RFiSFA is used to drive the joint for thumb circum-
material
High-force soft gripper [20] 4 Single NinjaFlex duction whose maximum flexion angle should be larger than 90° [30].
material To select proper bellow numbers for these two RFiSFAs, we plot the
Antagonistic fluidic finger 2 Single EP resin maximum flexion angle and the chamber length as a function of the
[19] material bellow number (Fig. 6). The design spaces of the maximum flexion
Two-finger soft gripper 2 Multi- Elastomer and
[21] material polyethylene glycol
angle and the chamber length are colored with grey and red, respec-
Anthropomorphic soft 0 Multi- Tango Black and Agile tively. Therefore, we select the bellow number for the first RFiSFA
skeleton hand [29] material White (Fig. 6(a)) and second RFiSFA (Fig. 6(b)) as 7 bellows and 10 bel-
lows, respectively.
 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx 7

Finally, to validate the performance of chamber, we characterize struct the biological structure and motion of human hand in an anthro-
the flexion angle and the blocking force for the selected RFiSFAs un- pomorphic form.
der the constant supplied pressure of 16 kPa (total 12 h with a step of The overall printing time is about 5 h. In the post process, we man-
an hour). As shown in Fig. 7, the flexion angles of both RFiSFAs in- ually remove the support materials coating the inner and external sur-
crease in the first several hours (0–7 t h hour for the 7-bellow one and face of the printed models mildly to avoid scratch. Next, we regulate
the water pressure of the WaterJet cleaning station for deeper support

F
0–9 t h hour for the 10-bellow one) and tend to be stable later. The to-
tal flexion angle increments are 9.9 % and 6.0 % for the 7-bellow and material removal. The fabrics are then attached to the joint levers as
the 10-bellow RFiSFAs, respectively. On the other side, there are no described in Section 2.3 and air tubes are attached to the air inlets to

OO
obvious variation trends for the blocking force of both RFiSFAs. The generate airtight fingers. Finally, we use threaded connections to as-
slight fluctuations measured are 0.017 N and 0.020 N for the 7-bellow semble the fingers to the palm skeleton within a few minutes (Fig.
and the 10-bellow RFiSFAs, respectively. 9(b)).

3. Modular design of the anthropomorphic hand 4. Evaluation of the motion and force performance

3.1. Multi-DoF fingers based on RFiSFAs 4.1. Characterization of the flexion angle

PR
Although the human hand is reported to have more than 20 DoFs, In this section, we test the flexion angles of the 2-DoF fingers over
completely imitating the actuation and transmission system of human increasing supplied pressures. Since the 3-DoF thumb has different ro-
hand may introduce difficulties in design, fabrication and control [30]. tation axes, its mobility is not quantitatively characterized and com-
In this section, we decouple the mobility of human hand into 11 DoFs pared with other fingers in this work. As shown in Fig. 10, we attach
and employ the modularized RFiSFAs to construct the 2-DoF flexion 6 markers on the finger base, the levers of MCP joint, and the levers
finger and the 3-DoF thumb. Note that index, middle, ring and little of PIP joint, respectively. For the 2-DoF finger, the DIP joint and fin-
fingers share the same finger type, i.e., the 2-DoF flexion finger. gertip are passive which can be regarded as a rigid body in the proce-

ED
Fig. 8(a) illustrates the 2-DoF flexion finger consisting of two seri- dure. Therefore, the positions of markers on the distal end of the PIP
ally linked 7-bellow RFiSFAs. These two RFiSFAs are responsible for joint lever can also be used to calculate the flexion angle of the fin-
flexion motions at MCP joint and PIP joint, respectively. MCP joint gertip. The experiments are performed for three times, and the results
and PIP joint are underactuated with an internal air channel and hence are averaged. As shown in Fig. 11(a), the maximum flexion angles of
only one air inlet is reserved for tube connection. The distal interpha- MCP joint and fingertip reach 66° and 130° at the maximum tolerable
langeal (DIP) joint is passive and integrated with the fingertip, which air pressure (16 kPa). The error bar indicates the standard deviation of
is designed as an interactive flexible interface with the environment. the flexion angle. The curves are approximately linear, which is con-
CT
Fig. 8(b) illustrates the 3-DoF thumb consisting of two parallel ducive to position control of fingers. Furthermore, the slope of finger-
10-bellow RFiSFAs and two serial 7-bellow RFiSFAs. It performs cir- tip fitting curve (8.31) is almost two times of that of MCP joint (4.41),
cumduction, adduction and flexion motions from the proximal end to which shows good coherence of the two serially linked RFiSFAs. To
the distal end. Firstly, two parallel 10-bellow RFiSFAs form the cir- furtherly explore the consistency between the four fingers, we calcu-
cumduction unit which is fixed on the palm skeleton with an incidence late the relative error variations of the flexion angles (Fig. 11(b)). The
angle of 30°. The circumduction unit flex around the palm axis to help relative error at a certain supplied pressure is calculated by e = Δθ/θ0
RE

palm opposition. The parallel arrangement also increases the stiffness , where Δθ is the standard deviation of the flexion angle and θ0 is the
of the whole thumb. Secondly, the adduction unit is a 7-bellow RF- average flexion angle of the four fingers. The results show that both
iSFA that compels the thumb to bend inward, which is responsible for relative errors for the MCP joint and fingertip are less than 10 % at
lateral pinch. Its rotational axis is perpendicular to that of the circum- supplied pressures from 4 kPa to 16 kPa. At 0 kPa to 4 kPa, the relative
duction unit. In our design, the active circumduction unit and adduc- errors may exceed 10 % (to 20.2 % and 13.2 % at 2 kPa for the MCP
tion unit approximatively replicate the function of the human’s car- joint and fingertip, respectively). But the absolute errors at 2 kPa reach
R

pometacarpal (CMC) joint with non-intersecting motion axes. Thirdly, only 1.41° and 2.16° for the two joints, which is negligible and can be
the flexion unit is designed with a 7-bellow RFiSFA that acts as the attributed to measuring errors. Therefore, the results show that the fab-
MCP joint. It has an inclination angle of 10° with that of the adduction rication process ensures good motion accuracy between the fingers.
unit to enhance the grasp capability. Finally, the interphalangeal (IP)
CO

joint is passive and integrated with the thumb tip. Three air inlets are 4.2. Characterization of the blocking force
reserved for tube connection. Note that we design anchor holes on the
bases of the 2-DoF flexion finger and the 3-DoF thumb for the quick Since all the fingers are all composed of modularized 7-bellow and
assembly. Similarly, corresponding anchor holes are reserved on the 10-bellow RFiSFAs, we choose the 2-DoF finger and the circumduc-
palm skeleton. tion unit of thumb for the blocking force tests without loss of general-
ity. Fig. 12(a) and (b) show the setups of blocking force measurement.
3.2. Anthropomorphic hand Fig. 12(c) shows the maximum blocking forces of two test specimens
UN

reach 0.27 N and 0.86 N at 16 kPa, respectively. The relationship of

As shown in Fig. 9(a), we develop the anthropomorphic hand blocking force and supplied pressure are nearly linear for both of the
mechanism based on three classes of components: active joints gener- two test specimens. The slope of the fitting curve of the circumduction
ated with RFiSFAs (green), flexible fingertips and passive joints with unit of thumb is 3.24 times of that of the index finger, which also ver-
pure flexible materials (yellow), and the rigid skeletons (grey). The ify the stiffness improvement of the parallel arrangement during actu-
designed 11-DoF anthropomorphic hand integrating one thumb and ation.
four flexion fingers is driven by total 12 RFiSFAs. We also design
customized architectures for quick assembly of fingers and the base 4.3. Performance validation of the anthropomorphic hand
skeleton, which provides improved replaceability compared to other
3D-printed soft hands [18]. Through this way, we simplify and recon The overall dimension of the fabricated hand prototype is
155 × 92 × 46 mm, which is comparable to that of a female hand.
8 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx

The hand mechanism weighs only 138 g due to the lightweight and jects with larger mass. We believe that the development in mechanical
modularized design. The hand prototype is attached to a pneumatic properties (tensile strength, tear resistance, fatigue life) of the print-
system which is composed of an air source (2–950 W-50, Outstand- able flexible material will make the printable RFiSFA for more practi-
ing Inc., China), pressure regulators (AR2000, Delixi Inc., China) and cal application. Secondly, the joint “ligament” that we expect is flexi-
manometers (82100, AZ Inc., China). Three channels are connected to ble but inextensible. However, there is no such material in our Polyjet

F
thumb and four channels to other four fingers. Coordinated finger mo- system for the current version and we use non-woven fabrics instead
tions can be achieved through sequential pneumatic control. Fig. 13(a) as a compromise. Thirdly, the cost of the multi-material Polyjet print-
and Movie S1 show the independent motion for thumb circumduction, ing is still high. Despite of these challenges that we encounter, this

OO
thumb adduction, thumb flexion, and index finger flexion (as an exam- work provides an alternative method to design and fabricate the an-
ple). In the grasping tests, the hand prototype is installed vertically or thropomorphic hands with integrated rigid-flexible structures, which
horizontally on the base. We synchronously control the pressure reg- may be utilized for prosthetic hands in the future. Furthermore, with
ulators for the fingers. As shown in Fig. 13(b), the anthropomorphic the development of 3D printing technologies including material per-
hand is capable of grasping objects with different dimensions, shapes formance and reduced cost, other 3D printed robots, especially bioin-
and textures (Movie S1). The hand’s inherent compliance provides spired ones that contains modularized rigid-flexible component and
adaptive envelop of objects and rigid skeletons provide sufficient sup- multiple DoFs will meet prosperous development in the future.
port for objects weighting below 50 g. The load capability can be fur-

PR
therly improved by developed materials and optimal topological struc-
Declaration of Competing Interest
tures in future works. Finally, the robustness test shows that the hand
is able to withstand the vertical impact of the hammer struck and still
None.
work well without air leakages or damages (Fig. 13(c) and Movie S1),
which attributes to the inherent flexibility of the digital materials and CRediT authorship contribution statement
tight connections between printable components. Ningbin Zhang: Methodology, Investigation, Formal analysis,
In Table 1, we also compare our developed hand with existing
Visualization, Writing - original draft, Writing - review & editing.

ED
3D printed, pneumatic soft hands, grippers or fingers in the litera-
ture, in terms of active DoFs, printing method and printing materi- Lisen Ge: Methodology, Investigation. Haipeng Xu: Methodology,
als. The results indicate that our hand has more active DoFs than oth- Investigation. Xiangyang Zhu: Supervision, Conceptualization,
ers due to the modular and compact design approach. Among the di- Project administration, Funding acquisition, Writing - original draft,
verse printing methods, the multi-material method is popular mainly Writing - review & editing. Guoying Gu: Supervision,
because it introduces multiple materials with a range of mechanical Conceptualization, Project administration, Funding acquisition,
performances. In addition, it is worthy of noting that the flexible bel- Writing - original draft, Writing - review & editing.
CT
low chamber can also be fabricated by the traditional molding method
[15,33]. However, the additionally manual assembly of the joint levers
to a sealed system is required. Acknowledgements

5. Conclusions This work was supported by the National Natural Science Founda-
tion of China (Nos 51620105002 and 91848204).
RE

In this paper, we present pneumatical, multi-material 3D printed,

modularized RFiSFAs that can be directly utilized to build anthropo- Appendix A. Supplementary data
morphic hands. The RFiSFA has a bellow chamber and joint mecha-
nisms with seamless connections. The experiment results show the in- Supplementary material related to this article can be found, in the
fluence of the chamber material and the bellow number on the burst online version, at doi:https://doi.org/10.1016/j.sna.2020.112090.
pressure, the flexion angle, the blocking force and the chamber length
R

of the RFiSFA. Based on the experiment results, we select the 60A

material as the chamber material, and 7-bellow and 11-bellow RFiS- References
FAs for different parts of an anthropomorphic hand. Two classes of
fingers are designed, in which the 3-DoF thumb has two parallel RF- [1] C. Piazza, G. Grioli, M. Catalano, A. Bicchi, A century of robotic hands, Annu.
CO

Rev. Control Robot. Auton. Syst. 2 (2019) 1–32.

iSFAs and two serial RFiSFAs, and the 2-DoF flexion finger has two [2] A. Billard, D. Kragic, Trends and challenges in robot manipulation, Science 364
serial RFiSFAs. For the 2-DoF finger, the flexion angle increases with (6446) (2019), eaat8414.
increasing supplied pressure in a nearly linear manner and the maxi- [3] L. Birglen, T. Laliberté, C. Gosselin, Underactuated Robotic Hands, Springer
mum flexion angle of fingertip reaches 130° at the supplied pressure Tracts in Advanced Robotics, Springer Verlag, Berlin, Heidelberg, 2008.
[4] M. Gabiccini, E. Farnioli, A. Bicchi, Grasp analysis tools for synergistic under-
of 16 kPa. The errors of flexion angles between four 2-DoF flexion fin-
actuated robotic hands, Int. J. Robot. Res. 32 (13) (2013) 1553–1576.
gers are relatively small (less than 2.16° at 0∼4 kPa for the absolute er- [5] A.M. Dollar, R.D. Howe, The highly adaptive SDM hand: design and perfor-
ror and less than 10 % at 4 ∼16 kPa for the relative error). The parallel
UN

mance evaluation, Int. J. Robot. Res. 29 (5) (2010) 585–597.

thumb circumduction unit has a larger blocking force output and actu- [6] C. Brown, H. Asada, Inter-finger coordination and postural synergies in robot
hands via mechanical implementation of principal components analysis, In: In-
ation stiffness (0.86 N, 55 mN/kPa) than that of the 2-DoF flexion fin-
telligent Robots and Systems (IROS), IEEE/RSJ International Conference on,
ger (0.27 N, 17 mN/kPa). An 11-DoF anthropomorphic hand (dimen- 2007, IEEE/RSJ 2007.
sion: 155 × 92 × 46 mm, weight: 138 g) is directly assembled with RF- [7] M.G. Catalano, G. Grioli, E. Farnioli, A. Serio, C. Piazza, A. Bicchi, Adaptive
iSFA-based fingers. In the grasping experiment, the fabricated hand is synergies for the design and control of the Pisa/IIT SoftHand, Int. J. Robot. Res.
able to grasp objects with different dimensions, shapes and textures. 33 (5) (2014) 768–782.
[8] S. Li, X. Sheng, H. Liu, X. Zhu, Design of a myoelectric prosthetic hand imple-
In the robustness tests, the proposed hand is able to withstand proper menting postural synergy mechanically, Indu. Robot: Int. J. 41 (5) (2014)
impacts of the hammer struck without air leakage or damage. 447–455.
There also remains some potential aspects to improve the RF- [9] C. Laschi, B. Mazzolai, M. Cianchetti, Soft robotics: technologies and systems
iSFA. Firstly, the payload of the RFiSFA is insufficient for ob pushing the boundaries of robot abilities, Sci. Robot. 1 (1) (2016), eaah3690.
[10] J. Shintake, V. Cacucciolo, D. Floreano, H. Shea, Soft robotic grippers, Adv.
Mater. 30 (29) (2018), 1707035.
 Sensors & Actuators: A. Physical xxx (xxxx) xxx-xxx 9

[11] G. Bao, L. Pan, H. Fang, X. Wu, H. Yu, S. Cai, B. Yu, Y. Wan, Academic re- west Jiaotong University, Chengdu, China, in 2013 and 2016, respec-
view and perspectives on robotic exoskeletons, IEEE Trans. Neur. Sys. Reh. 27
(11) (2019) 2294–2304. tively. He is currently a Ph.D. candidate at School of Mechanical En-
[12] E. Brown, N. Rodenberg, J. Amend, A. Mozeika, E. Steltz, M.R. Zakin, H. Lip- gineering, Shanghai Jiao Tong University, Shanghai, China. His re-
son, H.M. Jaeger, Universal robotic gripper based on the jamming of granular
material, Natl. Acad. Sci. U. S. A. 107 (44) (2010) 18809–18814.
search is in the development of soft robotics.
[13] K. Kim, X. Liu, Y. Zhang, Y. Sun, Nanonewton force-controlled manipulation

F
of biological cells using a monolithic MEMS microgripper with two-axis force Haipeng Xu received the B.S. degree (with honors) in mechanical en-
feedback, J. Micromech. Microeng. 18 (5) (2008), 055013.
[14] Z. Xie, A.G. Domel, N. An, C. Green, Z. Gong, T. Wang, E.M. Knubben, J.C. gineering from Northeastern University, Shenyang, China, in 2017.

OO
Weaver, K. Bertoldi, L. Wen, Octopus arm-inspired tapered soft actuators with He is currently working toward the Ph.D. degree in mechanical engi-
suckers for improved grasping, Soft Robot. (2020).
[15] K.C. Galloway, K.P. Becker, B. Phillips, J. Kirby, S. Licht, D. Tchernov, R.J.
neering with Shanghai Jiao Tong University, Shanghai, China. His re-
Wood, D.F. Gruber, Soft robotic grippers for biological sampling on deep reefs, search interests include 3D-printing soft sensors and robots.
Soft Robot. 3 (1) (2016) 23–33.
[16] O.Brock Deimel, A novel type of compliant and underactuated robotic hand for
dexterous grasping, Int. J. Robot. Res. 35 (1-3) (2016) 161–185. Xiangyang Zhu received the B.S. degree from the Department of Au-
[17] J. Zhou, X. Chen, U. Chang, J.T. Lu, C.C.Y. Leung, Y. Chen, Y. Hu, Z. Wang, tomatic Control Engineering, Nanjing Institute of Technology, Nan-
A soft-robotic approach to anthropomorphic robotic hand dexterity, IEEE Access
jing, China, in 1985, the M.Phil. degree in instrumentation engineer-

PR
7 (2019) 101483–101495.
[18] R.B.N. Scharff, E.L. Doubrovski, W.A. Poelman, P.P. Jonker, C.C.L. Wang, ing and the Ph.D. degree in automatic control engineering, both from
J.M.P. Geraedts, Towards Behavior Design of a 3D-Printed Soft Robotic Hand, the Southeast University, Nanjing, China, in 1989 and 1992, respec-
Soft Robotics: Trends, Applications and Challenges, Springer, 2017.
[19] T.J. Wallin, J.H. Pikul, S. Bodkhe, B.N. Peele, B.C. MacMurray, D. Therriault, tively. From 1993–1994, he was a Postdoctoral Research Fellow with
B.W. Mcenerney, R.P. Dillon, E.P. Giannelis, R.F. Shepherd, Click chemistry Huazhong University of Science and Technology, Wuhan, China. He
stereolithography for soft robots that self-heal, J. Mater. Chem. B 5 (31) (2017)
6249–6255. joined the Department of Mechanical Engineering as an Associate
[20] K.Y. Hong, Y.N. Hui, C.H. Yeow, High-force soft printable pneumatics for soft Professor, Southeast University, Dhaka, Bangladesh, in 1995. Since

ED
robotic applications, Soft Robot. 3 (3) (2016) 144–158.
[21] R. MacCurdy, R. Katzschmann, Y. Kim, D. Rus, Printable hydraulics: a method
June 2002, he has been with the School of Mechanical Engineering,
for fabricating robots by 3D co-printing solids and liquids, In: Robotics and Au- Shanghai Jiao Tong University, Shanghai, China, where he is cur-
tomation (ICRA), IEEE International Conference on, 2016, IEEE, 2016. rently a Changjiang Chair Professor and the Director of the Robotics
[22] D. Drotman, M. Ishida, S. Jadhav, M.T. Tolley, Application-driven design of
soft, 3-D printed, pneumatic actuators with bellows, IEEE-ASME Trans. Mech. Institute. His current research interests include robotic manipulation
24 (1) (2019) 78–87. planning, human–machine interfacing, and biomechatronics. Dr. Zhu
[23] M.A. Skylar-Scott, J. Mueller, C.W. Visser, J.A. Lewis, Voxelated soft matter
received the National Science Fund for Distinguished Young Scholars
CT
via multimaterial multinozzle 3D printing, Nature 575 (7782) (2019) 330–335.
[24] N.W. Bartlett, M.T. Tolley, J.T. Overvelde, J.C. Weaver, B. Mosadegh, K. in 2005.
Bertoldi, G.M. Whitesides, R.J. Wood, A 3D-printed, functionally graded soft
robot powered by combustion, Science 349 (6244) (2015) 161–165.
[25] K. Kumar, J. Liu, C. Christianson, M. Ali, M.T. Tolley, J. Aizenberg, D.E. Ing- Guoying Gu received the B.E. degree (with honors) in electronic sci-
ber, J.C. Weaver, K. Bertoldi, A biologically inspired, functionally graded end ence and technology, and the Ph.D. degree (with honors) in mecha-
effector for soft robotics applications, Soft Robot. 4 (4) (2017) 317–323.
[26] L.K. Ma, Y. Zhang, Y. Liu, K. Zhou, X. Tong, Computational design and fabri- tronic engineering from Shanghai Jiao Tong University, Shanghai,
RE

cation of soft pneumatic objects with desired deformations, ACM Trans. China, in 2006 and 2012, respectively. Since October 2012, Dr. Gu
Graphic36 (6) (2017) 1–12.
[27] M. Schaffner, J. Faber, L. Pianegonda, P.A. Ruhs, F. Coulter, A.R. Studart, 3D
has worked at Shanghai Jiao Tong University, where he is currently
printing of robotic soft actuators with programmable bioinspired architectures, appointed as a Professor of School of Mechanical Engineering. He
Nat. Commun. 9 (1) (2018) 878. was a Humboldt Postdoc Fellow with University of Oldenburg, Ger-
[28] Y.F. Zhang, N. Zhang, H. Hingorani, N. Ding, D. Wang, C. Yuan, B. Zhang, G.
Gu, Q. Ge, Fast-response, stiffness-tunable soft actuator by hybrid multimaterial many. He was a Visiting Scholar at Massachusetts Institute of Tech-
3D printing, Adv. Funct. Mater. 29 (15) (2019), 1806698. nology, National University of Singapore and Concordia University.
R

[29] J.A.E. Hughes, P. Maiolino, F. Iida, An anthropomorphic soft skeleton hand ex-
ploiting conditional models for piano playing, Sci. Robot. 3 (25) (2018),
His research interests include soft robotics, intelligent wearable sys-
eaau3098. tems, smart materials sensing, actuation and motion control. He is the
[30] P.W. Brand, A.M. Hollister, Clinical Mechanics of the Hand, 3rd ed., Mosby, St.
CO

author or co-author of over 80 publications, which have appeared in

Louis, MO, 1999.
[31] G. Dämmer, S. Gablenz, A. Hildebrandt, Z. Major, PolyJet-printed bellows actu- Science Robotics, Science Advances, IEEE/ASME Trans., Advanced
ators: design, structural optimization, and experimental investigation, Front. Ro- Functional Materials, Soft Robotics, etc., as book chapters and in con-
bot. AI 6 (34) (2019).
[32] https://www.stratasys.com/.
ference proceedings. Dr. Gu is the winner of multiple awards includ-
[33] B. Mosadegh, P. Polygerinos, C. Keplinger, S. Wennstedt, R.F. Shepherd, U. ing Young Changjiang Scholar of the Ministry of Education, National
Gupta, J. Shim, K. Bertoldi, C.J. Walsh, G.M. Whitesides, Pneumatic networks Science Fund for Excellent Young Scholars, The first prize of natural
for soft robotics that actuate rapidly, Adv. Funct. Mater. 24 (15) (2014)
science of Ministry of Education, Best Paper Award of Advanced Ro-
UN

2163–2170.
botics, ICIRA 2016 and IEEE-ICIA 2011. Now he serves as Associ-
Ningbin Zhang received the B.E. degree in mechanical engineering ate Editor of IEEE Transactions on Robotics. He has also served for
and M.S. degree in mechatronic engineering from Zhejiang Sci-Tech several journals as Editorial Board Member, Topic Editor, or Guest
University, Hangzhou, China, in 2012 and 2016, respectively. He is Editor, and several international conferences/symposiums as Chair,
currently a Ph.D. candidate at School of Mechanical Engineering, Co-Chair, Associate Editor or Program Committee Member.
Shanghai Jiao Tong University. His research is in the development of
soft robotics.

Lisen Ge received the B.E. degree in Engineering Structure Analy-

sis and M.S. degree in Vehicle Operation Engineering from South

View publication stats