Content uploaded by Zhiguo Li
Author content
All content in this area was uploaded by Zhiguo Li on Jun 26, 2020
Content may be subject to copyright.
J Bionic Eng 17 (2020) 1–14 Journal of Bionic Engineering
DOI: https://doi.org/10.1007/s42235-020-0068-6 http://www.springer.com/journal/42235
*Corresponding author: Zhiguo Li
E-mail: lizhiguo0821@163.com
The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
Zhiguo Li1,2*, Zhongliang Hou1,2, Yuxiao Mao1, Yan Shang1, Lukasz Kuta3
1. College of Mechanical and Electronic Engineering, Northwest A&F University, Yangling 712100, China
2. Key Laboratory of Agricultural Internet of Things, Ministry of Agriculture and Rural Affairs, Yangling 712100, China
3. Institute of Environmental Protection and Development, Wroclaw University of Environmental and Life Sciences, Wroclaw 50-375, Poland
Abstract
Bionic inspiration from human thumb and index finger was the drive to design a high-performance two-finger dexterous hand. The
size of each phalanx and the motion range of each joint in the human thumb and index finger were summarized, and the features of three
grasping patterns were described in detail. Subsequently, a two-finger dexterous bionic hand with 6 Degrees of Freedom (DoFs) was
developed. Both the mechanical thumb and index finger were made up of three rigid phalanx links and three mechanical rotation joints.
Some grasp-release tests validated that the bionic hand can perform three grasping patterns: power grasp, precision pinch and lateral pinch.
The grasping success rates were high under the following cases: (1) when power grasping was used to grasp a ring with external diameter
20 mm – 140 mm, a cylinder with mass < 500 g, or objects with cylinder, sphere or ellipsoid shape; (2) when the precision pinch was used
to grasp thin or small objects; (3) when the lateral pinch was used to grasp low length-to-width ratio of objects. The work provided a
method for developing two-finger bionic hand with three grasping patterns, and further revealed the linkage between the difference in
finger structure and size and the hand manipulation dexterity.
Keywords: dexterous bionic hand, mechanical finger, power grasp, precision pinch, lateral pinch
Copyright © Jilin University 2020.
1 Introduction
Recently, there is a growing demand for robots
working in unstructured environments, such as kitchen
working and fresh fruit harvesting[1,2]. However, it is a
big challenge for a robotic hand to grasp some unpre-
dictable objects by far[3,4] because the shape, size, posi-
tion and posture information of these objects are un-
known in advance and the robotic hand cannot pick up
the same object from the same place. If the target object
includes some agricultural products, serious mechanical
damage may occur during manipulation[5–7]. Therefore,
designing robotic hands that can grasp and manipulate
objects with anything approaching human levels of
dexterity is first on the to-do list for robotics[8], and some
typical dexterous bionic hands, such as the Anatomi-
cally-Correct Testbed (ACT) hand, Zhejiang University
of Technology (ZJUT) hand, Yale hand and in-Harvard
and Yale (i-HY) hand, were developed[9]. However, most
of these five-finger bionic hands included many me-
chanical sub-structures and joints with many Degrees of
Freedom (DoFs), so their stable and dexterous grasp
control depends on complex motion planning algo-
rithms[10,11]. The human thumb and index finger can
easily manipulate many daily-use items, so more and
more researchers pay attention to the structure of the
human thumb and index finger and the development of
two-finger dexterous bionic hands.
Previous research in this topic can be divided into
two categories. One category is related to understanding
human grasp taxonomy. Feix et al.[12] divided the human
five-finger grasp type into 33 patterns and the thumb and
index finger can only perform 3 grasp patterns, namely
power grasp, precision pinch and lateral pinch.
Chen et al.[13] proposed that the thumb-index finger’s
precision-pinch was more likely to grasp objects with an
equivalent diameter < 30 mm and a mass < 29 g, while
the thumb-index finger’s power-grasp was more likely to
grasp the objects with an equivalent diameter ≥ 30 mm
and a mass ≥ 129 g; by the cooperation of thumb and
index finger, the maximum grasping mass and diameters
of 3 years – 27 years participants ranged from 690 g to
Journal of Bionic Engineering (2020) Vol.17 No.*
2
9856.9 g and 70 mm to 170 mm, respectively. The other
category focuses on the design of two-finger bionic
robotic hands, especially in agricultural robots. Zhao et
al.[14] developed a clip-like two-finger end-effector for
apple picking where each finger is like a curved spoon.
Odhner et al.[15] proposed an underactuated two-finger
hand that can successfully pick up some small and thin
objects, such as key, coin, U-disk, battery, screw-driver
and pen, by flip- and -pinch motion. Ciocarlie et al.[16]
proposed a versatile single-actuator two-finger gripper
that can perform enveloping, parallel and fingertip
grasps and its grasping capability depended on the finger
link size. Rojas et al.[17] presented an underactuated
two-finger gripper to extend the capabilities of a tradi-
tional two-finger gripper by adding an elastic pivot joint
between two fingers, and its grasping manipulations do
not need know the size, shape, or other particulars of the
grasped object in advance. Ma and Dollar[18] proposed a
T42 two-finger bionic hand consisting of two, directly
opposing, underactuated fingers that can passively
conform to different geometries through power-grasps
or perform precision-grasps on smaller objects. Honar-
pardaz et al.[19] proposed a two-finger robotic gripper
that can stable pinch some cylindrical workpieces.
Heidari et al.[20] developed a self-adaptive, reconfigura-
ble two-finger robotic hand for space operations through
mechanical compliance; this design increased the capa-
bility of grasping large objects and decreased the contact
grasping forces. Zhang et al.[21] proposed a two-finger
end-effector for grasping tomato fruits. Levesque et
al.[22] proposed a two-fingered gripper for picking some
thin objects (e.g., card, cable, eraser) lying on a flat
surface by scooping grasp method, but the bin, key ring,
pack of cheese nips (flat) and cutting board cannot be
successfully grasped. Mu et al.[23] designed a clip-like
two-finger end-effector for kiwifruit picking and both
fingers had a 32.5 mm radius of curvature.
In summary, significant progress has been achieved
in comprehending the anatomy and grasping taxonomy
of human hand and the grasping capability and stability
of thumb and index finger, and in developing the
two-finger robotic hands. However, information is very
limited about a two-finger dexterous bionic hand which
can perform three grasping patterns, namely, power
grasp, precision pinch and lateral pinch. Most of the
thumb and index finger in each existing robotic hand is a
symmetric mechanical structure and has a same size, so
a basic structural difference relative to human thumb and
index finger exists. Whatever, the existing two-finger
robotic hands still have an obvious difference in grasp-
ing dexterity compared with the human thumb and index
finger system. Dexterous manipulation is one of the
primary goals in robotics[1]. Human thumb and index
finger can manipulate a wide range of daily-use items by
applying different grasping patterns (e.g., power grasp,
precision pinch and lateral pinch), which might be due to
that the thumb and index finger is an asymmetric struc-
ture and has an obvious difference in geometric size. The
bio-inspiration from the human thumb and index finger
was the drive to design a new two-finger dexterous
bionic hand for further improving the grasping capabil-
ity of robotic hands. Therefore, the objective of this
research was to provide a method for the development of
two-finger dexterous bionic hand-Northwest A&F
University (NWAFU) hand and further investigate the
linkage between the difference in finger structure and
size and the hand manipulation dexterity.
2 Bionic concepts and integrated design
2.1 Measurement of human thumb and index finger
dimensions
After human evolution over a period of several
million years, the modern human hands can stably grasp
different kinds of objects using up to 33 grasping pat-
terns[12,24], which were attributed to the phalanx size, the
joint DoF and the natural properties of skin. The phalanx
dimensions and joint motion range of the thumb and
index finger of some participants aged from 5 years to
70 years old were previously measured based on the
National Standard GB/T 5703-2010 “Basic human body
measurements for technical design”[25] and listed in
Table 1. There are some common acronyms used in
human hand anatomy, such as, DIP-distal interphalan-
geal joint, PIP-proximal interphalangeal joint, MCP-
metacarpal phalangeal joint, and BC-basal carpometa-
carpal joint. In Fig. 1, the proximal lengths, Lt1 and Li1,
of thumb and index finger refer to the distance from the
proximal flexion crease to the middle flexion crease of
the corresponding finger, whereas the medial length, Li2,
of the index finger refers to the distance from the middle
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
3
Ta bl e 1 Phalanx dimension and joint motion range of the human hand thumb and index finger
Parameters Thumb Index finger
Dimension[25]
Proximal length (mm) 30.1 ± 5.2 26.9 ± 3.5
Medial length (mm) – 21.6 ± 2.9
Distal length (mm) 27.8 ± 5.4 21.3 ± 3.3
Proximal breadth (mm) 22.6 ± 3.1 20.0 ± 2.5
Distal breadth (mm) – 17.0 ± 2.2
Pad breadth (mm) 19.9 ± 2.9 15.9 ± 4.3
Proximal thickness (mm) 20.4 ± 2.6 19.3 ± 2.5
Distal thickness (mm) – 15.2 ± 2.1
Range of motion[26–29]
DIP: Extension/Flexion − −10˚ – 80˚
PIP: Extension/Flexion −15˚ – 80˚ −10˚ – 100˚
MCP: Extension/Flexion −10˚ – 55˚ −10˚ – 90˚
MCP: Adduction/Abduction unknown 0˚ – 60˚
BC: Extension/Flexion 0˚ – 45˚ −
BC: Adduction/Abduction 0˚ – 60˚ −
Fig. 1 Phalanx dimensions of human thumb and index finger. Li1, Li2 and Li3 are the proximal, medial and distal lengths of the index finger,
respectively; Bi1, Bi2 and Bi3 are the proximal, distal and pad breadths of the index finger, respectively; Lt1 and Lt3 are the proximal and
distal lengths of the thumb, respectively; Bt1 and Bt3 are the proximal and pad breadths of the thumb, respectively; Ti1 and Ti2 are the
proximal and distal thicknesses of the index finger, respectively; Tt1 is the proximal thickness of the thumb; DIP, PIP, MCP and BC are the
distal interphalangeal joint, proximal interphalangeal joint, metacarpal phalangeal joint and basal carpometacarpal joint, respectively.
flexion crease to the distal flexion crease of the index
finger, and the distal lengths, Lt3 and Li3, of the thumb and
index finger refer to the distance from the distal flexion
crease to the tip of the corresponding finger. The proximal
breadths, Bt1 and Bi1, of the thumb and index finger refer
to the distance from the most lateral point to the most
medial point of the corresponding finger PIP joint. The
distal breadth, Bi2, of the index finger refers to the dis-
tance from the most lateral point to the most medial point
of the corresponding finger DIP joint, and the pad
breadths, Bt3 and Bi3, of thumb and index finger refer to
the breadth of the corresponding finger pad at the middle
of the distal phalanx. The proximal and distal thicknesses,
Ti1 and Ti2, of the index finger refer to the thickness at the
PIP and DIP joint of the index finger, respectively. The
proximal thickness, Tt1, of the thumb refers to the
thickness at the PIP of the thumb.
2.2 Three grasping patterns of human thumb and
index finger
The thumb includes distal, proximal and metacar-
pal phalanxes (Fig. 1), and its PIP joint has 1 DoF for
Journal of Bionic Engineering (2020) Vol.17 No.*
4
extension/flexion motion and its MCP and BC joints
have 2 DoFs for extension/flexion and adduc-
tion/abduction motion. The index finger includes distal,
medial, proximal and metacarpal phalanxes; and its DIP
and PIP joints have 1 DoF for extension/flexion motion
and its MCP joint has 2 DoFs for extension/flexion and
adduction/abduction motion[30]. The normal range of
motion of each joint in the thumb and index finger de-
scribed by Gutierrez et al.[26], Jaworski and Karpiński[27]
and Hammert et al.[28] is presented in Table 1. There are
three typical thumb-index finger grasping patterns,
shown in Fig. 2. According to the grasp taxonomy pro-
posed by Feix et al.[12], the power-grasp is defined as the
posture in which the index finger, including at least two
phalanxes, and thumb wrap around an object; in this
grasp both MCP and BC joints in the thumb perform an
abduction motion before holding the object and the
thumb MCP joint performs an extension at the end of the
motion[31,32] and is more or less flexed at the DIP, PIP
and MCP joints in the index finger and at the PIP joint in
thumb[33,34]. The precision-pinch is defined as the post-
ure in which the fingertips of the thumb and index finger
squeeze the opposite sides of the held object, without
contacting other finger surfaces; this grasp is abducted at
both MCP and BC joints in the thumb and the MCP joint
in the index finger[31], and more or less flexed at the DIP,
PIP and MCP joints in the index finger and at the PIP
joint in the thumb[35,36]. The lateral pinch is defined as
the posture in which the thumb applies force to a small
and flat object to hold it against the lateral aspect of the
index finger[37]; this grasp is abducted at MCP joint in
the thumb, and more or less flexed at the all joints in the
thumb and index finger[31,32].
2.3 Bio-integrated design of the dexterous bionic
hand
2.3.1 Mechanical structure of the thumb and index
finger
The mechanical thumb and index finger of the
two-finger dexterous bionic hand is shown in Figs. 3a
and 3b, and its design was kept close to the anatomical
structure of the human thumb and index finger. The
mechanical thumb of the bionic hand included a distal
phalanx link (1), a proximal phalanx link (2), a meta-
carpal phalanx link (3), a PIP mechanical joint (4), a
MCP mechanical joint (5) and a BC mechanical joint (6).
Both the PIP and MCP mechanical joints had one rota-
tion DoF for achieving the flexion/extension movement
of the distal and proximal phalanx links of the thumb,
respectively. To let the mechanical thumb laterally rotate
and further perform lateral pinch tasks, a rotation DoF
was added to the BC mechanical joint so that the bionic
hand thumb can laterally rotate around the BC mechan-
ical joint and its motion plane is transformed (Fig. 3b).
Hence, the human hand thumb with 5 DoFs was simpli-
fied into a bionic mechanical thumb with 3 DoFs. The
index finger of the bionic hand included a distal phalanx
link (7), a middle phalanx link (8), a proximal phalanx
link (9), a DIP mechanical joint (10), a PIP mechanical
joint (11) and a MCP mechanical joint (12). During
power grasp, precision pinch and lateral pinch manipu-
lations, the mechanical index finger of the bionic hand
will not perform the adduction/abduction motion,
so its mechanical joints were not considered to have an
Fig. 2 Three typical grasping patterns of human thumb and index finger.
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
5
Fig. 3 Two-finger dextrous bionic hand. (a) Mechanical structure of thumb and index finger, the initial intersection angle between thumb
and index finger was 150˚; (b) 90˚ lateral rotation of mechanical thumb; (c) two-finger bionic hand system; (d) sectional view of the
two-finger bionic hand. 1-distal phalanx link of mechanical thumb; 2-proximal phalanx link of mechanical thumb; 3-metacarpal phalanx
link of mechanical thumb; 4-PIP joint of mechanical thumb; 5-MCP joint of mechanical thumb; 6-BC joint of mechanical thumb; 7-distal
phalanx link of mechanical index finger; 8-middle phalanx link of mechanical index finger; 9-proximal phalanx link of mechanical index
finger; 10-DIP joint of mechanical index finger; 11-PIP joint of mechanical index finger; 12-MCP joint of mechanical index finger; 13, 14,
15-DC motor; 16, 17, 18-tendon rope; 19, 20, 21-winding reel; 22, 23, 24, 25, 26-tension spring; 27-motor box; 28-Arduino UNO-based
controller; 29, 30, 31, 32, 33, 34-FSR 400 film sensor; 35-torsion spring.
adduction/abduction motion DoF during design. To let
the mechanical index finger perform the flex-
ion/extension movement, the DIP, PIP and MCP me-
chanical joints were designed as a rotation joint for
achieving the flexion/extension movement of the distal,
middle and proximal phalanx links, respectively. Ac-
cording to the anthropometric data, the maximum
opening intersection angle between the human thumb
and index finger is about 150˚, so the initial natural in-
tersection angle between the mechanical thumb and
index finger of the bionic hand was designed as 150˚
(Fig. 3a).
The phalanx link dimension and mechanical joint
motion range of the bionic hand thumb and index finger
are presented in Table 2. During bionic design, to sim-
plify the structure of the human thumb for easily auto-
matic control, the BC mechanical joint of the bionic
hand thumb did not include a rotation DoF for achieving
the flexion/extension movement of the metacarpal pha-
lanx link. Because the mechanical thumb of the bionic
hand needs to flex 0˚ – 90˚ toward to the side face of the
mechanical index finger for achieving lateral pinch tasks,
the maximum flexion/extension rotation angle of the
thumb MCP mechanical joint and the maximum lateral
rotation angle of the thumb BC mechanical joint were
increased to 90˚. To let the distal phalanx link of the
mechanical index finger keep a relatively large and pa-
rallel contact area with the mechanical thumb during a
precision pinch, the maximum extension rotation angle
of the DIP mechanical joint in the bionic index finger
was increased to 50˚. Furthermore, if the maximum
extension rotation angle of the thumb PIP mechanical
joint was more than 0˚, the lateral pinch between the
mechanical thumb and index finger would be unstable.
There might be a phenomenon of kinematic interference
between the adjacent phalanx links when a mechanical
finger flexes, hence, in order to ensure that each me-
chanical joint had enough rotation range, the length of
Journal of Bionic Engineering (2020) Vol.17 No.*
6
Ta bl e 2 Phalanx link dimension and mechanical joint motion range of the bionic hand thumb and index finger
Parameters Thumb Index finger
Dimension
Proximal link length (mm) 48 56
Medial link length (mm) – 41
Distal link length (mm) 36 31
Proximal link breadth (mm) 24 20
Distal link breadth (mm) – 20
Pad link breadth (mm) 24 20
Proximal link thickness (mm) 18 18
Distal link thickness (mm) 20 15
Range of motion
Mechanical DIP joint: Extension/Flexion – −50˚ – 90˚
Mechanical PIP joint: Extension/Flexion 0˚ – 90˚ 0˚ – 90˚
Mechanical MCP joint: Extension/Flexion 0˚ – 90˚ 0˚ – 90˚
Mechanical MCP joint: Adduction/Abduction 0 –
Mechanical BC joint: Extension/Flexion – –
Mechanical BC joint: Adduction/Abduction 0˚ – 90˚ –
each phalanx link in the bionic hand was appropriately
extended based on human hand phalanx size. In addition,
for the structural design of the bionic hand, each phalanx
link in the mechanical thumb and the mechanical index
finger was set as the same width, and the distal phalanx
link of the mechanical thumb and index finger was set as
the same thickness. A 1 mm thick silicone pad was
pasted on the internal surface of each phalanx link in the
mechanical thumb and index finger of the bionic hand.
2.3.2 Mechatronics system of the two-finger dexterous
bionic hand
The two-finger dexterous bionic hand is shown in
Figs. 3c and 3d, which includes the mechanical thumb
and index finger, driving system, and sensing and con-
trol system. Because the structure and size of a human
thumb and index finger was mimicked in the design of
the mechanical thumb and index finger, the two-finger
bionic hand can also perform three human-like grasping
patterns: power grasp, precision pinch and lateral pinch,
by the cooperative manipulations of the mechanical
thumb and index finger. The driving system included
three DC motors, three tendon ropes, three winding reels,
ten tension springs and one torsion spring. Each tension
spring or torsion spring was to assist the reset of each
corresponding phalanx link.
When the DC motor (13) starts rotating in a
clockwise direction and its output shaft will drive the
winding reel (19) to recycle the tendon rope (16), the
tendon rope becomes tight and generates a tension force.
If the tension force of the tendon rope is larger than that
of the tension springs (22) and (23), the PIP and MCP
mechanical joints will rotate inward and the mechanical
thumb performs a flexion motion accordingly. When the
DC motor (13) starts rotating in a counter-clockwise
direction, the tendon rope becomes slack and the relax-
ation of the tension springs (22, 23) will result in the
outward rotation of the PIP and MCP mechanical joints,
so the mechanical thumb performs an extension motion.
Similarly, when the DC motor (15) starts rotating in a
counter-clockwise direction and its output shaft will
drive the winding reel (21) to recycle the tendon rope
(18), the tendon rope becomes tight and generates a
tension force. If the tension force of the tendon rope is
larger than that of the torsion spring (35), the thumb
metacarpal phalanx link (3) will rotate around the BC
joint in a clockwise direction and the mechanical thumb
performs an adduction motion accordingly. When the
DC motor (15) starts rotating in a clockwise direction
and the tendon rope becomes slack, the relaxation of the
torsion spring (35) will result in the rotation of the thumb
metacarpal phalanx link (3) around the BC joint in a
counter-clockwise direction, so the mechanical thumb
performs an abduction motion. Similarly, when the DC
motor (14) starts rotating in a counter-clockwise direc-
tion and its output shaft will drive the winding reel (20)
to recycle the tendon rope (17), the tendon rope becomes
tight and generates a tension force. If the tension force of
the tendon rope is larger than that of the tension springs
(24), (25) and (26), the DIP, PIP and MCP joints of the
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
7
mechanical index finger will rotate inward and so the
mechanical index finger performs a flexion motion ac-
cordingly. While the DC motor (14) starts rotating in a
counter-clockwise direction, the tendon rope will be-
come slack and the relaxation of the tension springs (24,
25, 26) will result in the outward rotation of the DIP, PIP
and MCP mechanical joints, so the mechanical index
finger performs an extension motion.
The sensing and control system mainly included an
Arduino UNO motion controller (ATMEL Corporation,
USA) (28) and six FSR 400 film force sensors (29–34)
(Interlink Electronics, Inc., USA). Every two force
sensors were parallel pasted on the internal surfaces of
the proximal phalanx link in the mechanical thumb and
the metacarpal and proximal phalanx links in the me-
chanical index finger, respectively. The controller was
mainly used to send commands for driving DC motors
during power grasp, precision pinch and lateral pinch,
and to adjust the power supply mode of the bionic hand
in a power grasp based on the feedback of six force
sensors. Once one of the six force sensors received a
normal contact force signal, the power supply mode of
the bionic hand in a power grasp will be changed im-
mediately.
2.3.3 Working process
The working process of the two-finger dexterous
bionic hand with three grasping patterns was as follows.
Power grasp: The two mechanical fingers are in-
itially in a natural opening state (Fig. 3a). Firstly, the
bionic hand approaches an object and the DC motors (13)
and (14) are started simultaneously. The tendon ropes
(16) and (17) are recycled so that each phalanx link in
the mechanical thumb and index finger force is forced to
make an inward flexion motion around a corresponding
mechanical joint (Fig. 3c). Once one of the phalanx link
surfaces in the mechanical thumb and index finger starts
to touch the object, the force signal obtained by a FSR
400 sensor will be transmitted to the motion controller
and the power-supply mode of the DC motors (13) and
(14) will be changed from a continuous power supply to
an intermittent power supply with a frequency of 20 Hz
by controller commands. The intermittent power supply
mode of the DC motor is similar to the pulsing signal of
human muscle while working, which can force the finger
phalanx links to apply normal force to the object at a
high frequency, so that the object can be stably mani-
pulated using the power grasp pattern. When the grasp-
ing task is finished, the DC motors (13) and (14) are
controlled to reverse rotation and the tendon ropes (16)
and (17) are relaxed, so, with the action of tension
springs, each phalanx link in the mechanical thumb and
index finger extends to release the object.
Precision pinch: The two mechanical fingers are
initially in a natural opening state (Fig. 3a). Firstly, when
the bionic hand approaches an object, the DC motor (13)
is started. The tendon rope (16) is recycled to force the
inward flexion of the distal and proximal phalanx link in
the mechanical thumb (Fig. 3c). When the thumb distal
phalanx link starts to contact the object, the DC motor
(14) is started. The tendon rope (17) is recycled to force
the inward flexion of the distal, middle and proximal
phalanx links in the mechanical index finger (Fig. 3d).
When the distal phalanx link of the mechanical index
finger starts to contact the object, the bionic hand thumb
and index finger will perform a precision pinch task.
After the grasping task is finished, the DC motors (13)
and (14) are driven to reverse rotation, and the tendon
ropes (16) and (17) are relaxed, so, with the action of
tension springs, each phalanx link in the
mechanical thumb and index finger extends to release
the object.
Lateral pinch: The two mechanical fingers are in-
itially in a natural opening state (Fig. 3a). Firstly, when
the bionic hand approaches an object, the DC motor (15)
is started. The tendon rope (18) is recycled so that the
thumb metacarpal phalanx link (3) is forced to rotate
inwardly 90˚ around the BC joint in a clockwise direc-
tion and performs an adduction motion (Fig. 3c), so the
mechanical thumb will be perpendicular to the lateral
surface of the middle phalanx link in the mechanical
index finger. Subsequently, the DC motor (14) starts
rotating. The tendon ropes (17) are recycled to force the
inward flexion of the distal, middle and proximal pha-
lanx links in the mechanical index finger (Fig. 3d); when
the object is located between the bionic hand thumb and
index finger, the DC motor (13) starts rotating. Tendon
rope (16) is recycled to force the inward flexion of the
distal and proximal phalanx links in the mechanical
thumb (Fig. 3d), so the mechanical thumb and index
Journal of Bionic Engineering (2020) Vol.17 No.*
8
finger will perform a lateral pinch task. After the
grasping task is finished, the DC motor (13) is driven to
reverse rotation, and the tendon rope (16) is relaxed, so,
with the action of tension springs, each phalanx link in
the mechanical thumb extends to release the object.
Finally, the DC motors (14) and (15) is driven to rotate in
reverse, the tendon ropes (17) and (18) are relaxed and,
with the respective action of a torsion and tension spring,
the mechanical thumb metacarpal phalanx link (3) ro-
tates outwardly 90˚ around the BC joint in a coun-
ter-clockwise direction for an abduction motion, and
each phalanx link in the mechanical index finger extends
for reset.
3 Materials and methods
3.1 Target objects for power grasping
In consideration that the human hand power
grasping pattern is always used to grasp large and heavy
objects, three experiments were designed to investigate
the grasp capabilities of the two-finger dexterous bionic
hand in a power grasping pattern. The first experiment
was mainly used to explore the graspable dimension
ranges of the bionic hand. 17 rings made of polylactide
were selected as the grasped objects (Fig. 4a). Each ring
had a 1.5 mm wall thickness and 40 mm height. The
external diameters of the rings ranged from 10 mm to
170 mm in 10 mm increments. The second experiment
was mainly used to explore the graspable mass ranges of
the bionic hand. Seven solid stainless-steel cylinders
were selected as the grasped objects (Fig. 4b). The cy-
linder diameter was 40 mm and their masses ranged
from 200 g to 800 g in 100 g increments. The third ex-
periment was mainly used to explore which object
shapes were suitable to be grasped by the bionic hand.
Seven shapes (cuboid, cylinder, triangular prism, trian-
gular pyramid, cone, sphere and ellipsoid) were selected
(Fig. 4c). All were 3D printed objects and were made of
polylactide. The length × width × height of the cuboid
was 60 mm × 60 mm × 90 mm, the diameter × height of
cylinder was 60 mm × 90 mm, the bottom side-length ×
height of the triangular prism and triangular pyramid
was 60 mm × 90 mm, the bottom circle diameter ×
height of cone was 60 mm × 90 mm, the diameter of
sphere was 60 mm, and the equatorial diameter × polar
axis diameter of the ellipsoid was 60 mm × 90 mm.
3.2 Target objects for precision pinch
In consideration that the human hand precision
pinch pattern is always used to grasp small or thin ob-
jects, some common items, such as campus card, key,
Fig. 4 Target objects for validating three grasping patterns. (a) 17 rings for power grasp; (b) seven rigid cylinders for power grasp; (c)
seven different object shapes for power grasp; (d) some common laboratory items for precision pinch; (e) some common laboratory items
for lateral pinch. The pink bar on each item indicates the grasping position.
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
9
steel ruler, U disk, graver, tweezer and steel file and
cerasus humilis, were selected as target objects for ex-
ploring the grasping performance of the bionic hand in a
precision pinch pattern (Fig. 4d).
3.3 Target objects for lateral pinch
In consideration that the human hand lateral pinch
pattern is always used to grasp light and thin objects,
some common items, such as campus card, key, steel
ruler, scissor, inkpad box, battery, adhesive tape and
steel file, were selected as target objects for exploring
the grasping performance of the bionic hand in a lateral
pinch pattern (Fig. 4e). The pink belt indicates the
grasping position on the item surface. This grasp pattern
can bear large torsion torque. For example, a human
always pinches a key using a lateral pinch pattern to
open a lock on a door.
3.4 Experimental method
Before each test, the silicone pad on the internal
surface of each phalanx link and the grasped object
surface should be cleaned and dried. After switching on
the power supply, an experimental participant grasped
and moved the two-finger bionic hand to approach a
target object and then controlled the mechanical thumb
and index finger into a pre-grasp status. In this process,
the plane formed by the mechanical thumb and index
finger was parallel to the desk surface, and the grasping
approaching direction of the mechanical fingers relative
to the target object is presented in Fig. 5. After the object
was stably grasped, it was lifted and held in the air for
20 s. A grasp trial was characterized as a success if the
object was not dropped from the bionic hand during 20 s;
otherwise, it was characterized as a failure. For each
object, the test was repeated 10 times with the same
grasping pattern. Therefore, the success rate can be
calculated as:
Success rate
(success number of grasps / total number of gras
p
s)
100%.
(1)
When a power grasp pattern of the bionic hand was
used, an object would be wrapped from its ½ height of
the external curved surface using the internal surfaces of
mechanical thumb and index finger (Fig. 5a). When a
precision pinch pattern was used, an object would be
grasped from its upper and lower surfaces using the
internal surfaces of the distal phalanx links in the me-
chanical thumb and index finger (Fig. 5b); additionally,
the graver and steel file were grasped from their handles.
When a lateral pinch pattern was used, a thin-board
object would be grasped from its upper and lower sur-
faces using the internal surface of the distal phalanx link
in mechanical thumb and the lateral surface of the
medial phalanx link in mechanical index finger (Fig. 5c).
The insertable depth between mechanical thumb and
index finger was shorter than 4 cm, so the scissors and
steel file were grasped from their flat portion, about 4 cm
from the end. The mass and thickness of the grasped
items were measured to a 0.01 g accuracy using an
electronic balance and a 0.01 mm accuracy using an
electronic digital calliper.
4 Results and discussion
4.1 Power grasping
The experimental results for having the two-finger
bionic hand to grasp different diameter rings, different
mass rigid cylinders and different shaped objects using
Fig. 5 Three grasping patterns. (a) Power grasp; (b) precision pinch; (c) lateral pinch.
Journal of Bionic Engineering (2020) Vol.17 No.*
10
Ta bl e 3 Three kinds of power grasping tests
Tests Target object Parameters Total number of grasps Number of success grasps Success rate
First kind of tests Different diame-
ter rings (mm)
10 10 0 0%
20 10 10 100%
30 10 10 100%
40 10 10 100%
50 10 10 100%
60 10 10 100%
70 10 10 100%
80 10 9 90%
90 10 10 100%
100 10 10 100%
110 10 10 100%
120 10 10 100%
130 10 10 100%
140 10 9 90%
150 10 0 0%
160 10 0 0%
170 10 0 0%
Second kind of tests Different mass
cylinders (g)
200 10 10 100%
300 10 10 100%
400 10 10 100%
500 10 8 80%
600 10 0 0%
700 10 2 20%
800 10 2 20%
Third kind of tests Different shape
objects
Cuboid 10 1 10%
Cylinder 10 10 100%
Triangular prism 10 0 0%
Sphere 10 10 100%
Cone 10 0 0%
Triangular pyramid 10 0 0%
Ellipsoid 10 8 80%
the power grasp pattern are listed in Table 3. Without
considering the object mass, the graspable ring diameter
of the bionic hand ranged from 20 mm to 140 mm using
the power grasp pattern. Although there was one failure
case when the bionic hand grasped the rings with 80 mm
and 140 mm diameter, the 90% grasping success rate
revealed that there was a highly stable grasping proba-
bility. Objects with a diameter < 20 mm could not be
stably grasped using the power grasping pattern because
when all the mechanical joints in the bionic thumb and
index finger inwardly flexed for grasping objects, the
contact of the distal phalanx links in the mechanical
thumb and index finger restricted the further flexion
motions of the MCP mechanical joint in bionic thumb
and the PIP and MCP mechanical joints in bionic index
finger, so the bionic hand had a grasping dead zone in the
power grasping pattern. Objects with a diameter > 150 mm
could not be stably grasped using the power grasping
pattern because the diameter of the object exceeded the
grasping span of the bionic hand.
For the Φ 40 mm rigid cylinders, the maximum
graspable mass of the bionic hand was 500 g using the
power grasp pattern. There is no possible way to stably
grasp an object with a mass > 500 g because the friction
force between the mechanical finger and cylinder was
not enough to overcome the weight of the cylinder. Al-
though there were two successful grasp cases when
grasping cylinders with masses of 700 g and 800 g, the
low success rate cannot ensure reliable application in
practice.
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
11
The success rate is high when the two-finger bionic
hand grasped the cylinder, sphere and ellipsoid objects,
while the success rate was near zero when the two-finger
bionic hand grasped the cuboid, triangular prism, cone
and triangular pyramid objects. This result indicated that
the bionic hand is more suitable to grasp curved surface
objects than angled objects. When an angled object was
grasped, the contact between mechanical finger surface
and object was always poor and the FSR 400 sensors
pasted on some phalanx link surfaces had difficulty in
getting force signals for changing the power supply
mode of the DC motors, so the grasp force of the bionic
hand was small.
The power grasp is a main function of many ex-
isting two-finger robotic hands and can be compliant
envelope a relative wide variety of objects arbitrarily
placed in a space. In the literature, it is showed that a
two-finger hand developed by Heidari et al.[20] can
power grasp a wide variety of objects with maximum
contact force 4.4 N – 4.8 N and a versatile sin-
gle-actuator two-finger gripper proposed by Ciocarlie et
al.[16] can apply the maximum grasping force 37.5 N
under enveloping manipulation (power grasp). The
grasping capabilities (e.g. graspable size and mass) of
each two-finger robotic hand mainly depend on the link
size of robotic fingers and the power supply. The human
thumb and index finger are an asymmetric structure and
have an obvious difference in geometric size (Fig. 1), so
its grasping advantages include (1) the grasping range is
increased[20]; (2) the needed contact force is decreased[20];
(3) an in-hand manipulation with motion at contact can
be performed[17]; and (4) a thin object can be flipped up
to acquire a pinch grasp[15]. Most of the thumb and index
finger in each existing robotic hand are a symmetric
mechanical structure, so it will affect their grasping
capabilities. In this study, the new two-finger dexterous
bionic hand was designed based on the anatomical fea-
tures of human fingers, so its grasp capabilities were
improved greatly.
4.2 Precision pinch
The results of the precision pinch tests are listed in
Table 4. The success rate is 100% when the two-finger
bionic hand precision pinched the key, steel ruler, cam-
pus card, tweezer and cerasus humilis, which show that
the bionic hand is suitable to grasp thin or small objects
by precision pinch. The success rates were about 50%
and 70% when the bionic hand precision pinched the
handles of the graver and steel file, which illustrated that
the bionic hand had a partial ability to grasp small cy-
linders using precision pinch but the small cylinder was
not the most suitable objects for the bionic hand using a
precision pinch. The essential differences for precision
pinching thin plates (such as the campus card and key)
and small cylindrical objects (such as the handles of the
graver and steel file) are the distance between mechan-
ical thumb and index finger and the contact area between
mechanical finger and object. When the contact surface
between mechanical finger and target object is small and
the distance between mechanical thumb and index finger
is large, the grasping line between two mechanical fin-
gers is not easy to be in the friction cone, so the grasped
object is susceptible to rotate around the grasping line
and the precision pinch will be unstable[38]. Similarly,
some two-finger robotic hands proposed by Odhner et
al.[15], Ma and Dollar[18] and Honarpardaz et al.[19] can
also perform the precision pinch task, but their two fin-
gertip surfaces in the mechanical thumb and index finger
cannot keep parallel during pinch. In this study, the new
developed two-finger dexterous bionic hand is easy to
stable precision-pinch the common items because the
internal surfaces of two distal phalanx links can keep
parallel pinch and the pinch reliability is much im-
proved.
4.3 Lateral pinch
The results of the lateral pinch tests are listed in
Table 4. The success rate was 100% when the two-finger
bionic hand laterally pinched the campus card, key, in-
kpad box, battery and adhesive tape, while the success
rate was lower than 30% when the two-finger bionic
hand laterally pinched the steel ruler, scissors and steel
file. This indicated that the lateral pinch pattern of the
bionic hand was always suitable to grasp some objects
with a relative low ratio of length to width. Using the
lateral pinch pattern, the two-finger bionic hand always
pinched an object from one end by the thumb distal
phalanx link and the lateral surface of the middle
phalanx link in the mechanical index finger. Because the
length of the mechanical thumb is limit, the insertable
Journal of Bionic Engineering (2020) Vol.17 No.*
12
Ta bl e 4 Results of the precision pinch and lateral pinch
Grasp pattern Target object Total number of
grasps
Success number
of grasps
Success
rate
Item Mass (g) l × w × t (mm)
Precision pinch
Key 2.93 1.68 10 10 100%
Steel ruler 10.32 0.42 10 10 100%
U disk 11.8 8.58 10 9 90%
Campus card 6.20 0.90 10 10 100%
Graver 11.43 12.20 10 7 70%
Tweezer 14.86 1.69 10 10 100%
Steel file 25.02 6.41 10 5 50%
Cerasus humilis 4.99 21.19 10 10 100%
Lateral pinch
Campus card 6.20 0.90 10 10 100%
Key 2.93 1.68 10 10 100%
Steel ruler 10.32 0.42 10 1 10%
Scissor 51.60 2.83 10 3 30%
Inkpad box 20.71 14.32 10 10 100%
Battery 34.99 16.99 10 10 100%
Adhesive tape 11.81 11.76 10 10 100%
Steel file 25.02 1.94 10 0 0%
depth of an object under the mechanical thumb was
always shorter than 4 cm. When an object is laterally
pinched and then rotated 90˚ around its long axis using
the mechanical thumb and index finger, a stable lateral
pinch task required that the friction torque between the
mechanical finger and object can overcome the rotation
torque formed by the objects’ gravity. If the laterally
pinched object is too long and its gravity center will be
far away from the grasping line between two mechanical
fingers, it will be difficult to overcome the rotation tor-
que formed by the object gravity, so the lateral pinch task
will fail easily. In the current literature it is not found that
these two-finger robotic hands can perform the lateral
pinch task, so no comparisons in the lateral pinch capa-
bility are possible.
5 Conclusion
Human thumb and index finger can manipulate a
wide range of daily-use items by applying different
grasping patterns, which might be due to that the thumb
and index finger are an asymmetric structure and have an
obvious difference in geometric size. Hence, according
to the human hand anatomical structure and the anth-
ropomorphic design concept, a two-finger dexterous
bionic hand with 6 DoFs was developed for performing
three grasping patterns: power grasp, precision pinch
and lateral pinch. The mechanical thumb and index
finger in the bionic hand were mainly made up of three
phalanx links and three mechanical joints. All the me-
chanical joints had only a rotational DoF. A series of
grasp-release tests validated that the cooperative mani-
pulation of the mechanical thumb and index finger in the
bionic hand can achieve three grasping patterns. The
grasping success rates were high under the following
cases: (1) when the power grasping was used to grasp a
ring with external diameter 20 mm – 140 mm, a cylinder
with mass < 500 g, or objects with cylinder, sphere or
ellipsoid shape; (2) when the precision pinch was used to
grasp thin or small objects; and (3) when the lateral
pinch was used to grasp low length-to-width ratio of
objects. The work contributed to provide a method for
the development of two-finger dexterous bionic hand.
These experimental results provided substantial evi-
dence for the original assumptions and gave an expla-
nation for the linkage between the difference in finger
structure and size and the hand manipulation dexterity.
Acknowledgment
This work was supported by a European Marie
Curie International Incoming Fellowship (326847 and
912847), a Special Foundation for Talents of Northwest
A&F University (Z111021801), and two Key Research
and Development Plans of Shaanxi Province
(2019NY-172 and 2018030).
Li et al.: The Development of a Two-finger Dexterous Bionic Hand with Three
Grasping Patterns–NWAFU Hand
13
References
[1] Billard A, Kragic D. Trends and challenges in robot mani-
pulation. Science, 2019, 364, eaat8414.
[2] Arad B, Balendonck J, Barth R, Ben-shalar O, Edan Y,
Hellstrom T, Hemming J, Kurtser P, Ringdahl O, Tielen T,
Tuijl B. Developemnt of a sweet pepper harveting robot.
Journal of Field Robotics, 2020, 37,
https://doi.org/10.1002/rob.21937.
[3] Hao Y F, Gong Z Y, Xie Z X, Guan S Y, Yang X B, Wang T
M, Wen L. A soft bionic gripper with variable effective
length. Journal of Bionic Engineering, 2018, 15, 220–235.
[4] Kim S, Kim J, Kim M, Kim S, Park J. Grasping force esti-
mation by sEMG signals and arm posture: tensor decompo-
sition approach. Journal of Bionic Engineering, 2018, 16,
455–467.
[5] Kim S, Kim J, Kim M, Kim S, Park J. Grasping force esti-
mation by sEMG signals and arm posture: Tensor decom-
position approach. Journal of Bionic Engineering, 2018, 16,
455–467.
[6] Liang Z W, Xu L Z, De Baerdemaeker J, Li Y M, Saeys W.
Optimisation of a multi-duct cleaning device for rice com-
bine harvesters utilising CFD and experiments. Biosystems
Engineering, 2020, 190, 25–40
[7] An X, Li Z G, Zude-Sasse, M, Tchuenbou-Magaia, F, Yang Y
G. Characterization of textural failure mechanics of straw-
berry fruit. Journal of Food Engineering, 2020, 282,
110016.
[8] Hodson R. A gripping problem. Nature, 2018, 557, S23.
[9] Piazza C, Grioli G, Catalano M, Bicchi A. A century of
robotic hands. Annual Review of Control, Robotics, and
Autonomous Systems, 2019, 2, 1–32.
[10] Kim E, Lee S, Lee Y. A dexterous robot hand with a
bio-mimetic mechanism. International Journal of Precision
Engineering and Manufacturing, 2011, 12, 227–235.
[11] Odhner L U, Jentoft L P, Claffee M R, Corson N, Tenzer Y,
Ma R R, Buehler M, Kohout R, Howe R D, Dollar A M. A
compliant, underactuated hand for robust manipulation. In-
ternational Journal of Robotics Research, 2014, 33,
736–752.
[12] Feix T, Romero J, Schmiedmayer H, Dollar A M, Kragic D.
The GRASP taxonomy of human grasp types. IEEE Trans-
actions on Human – Machine Systems, 2016, 46, 66–77.
[13] Chen X J, Li Z G, Wang Y Q, Liu J Z, Zhao D Z. Investiga-
tion on the cooperative capabilities of human thumb and
index finger. Frontiers in Neurorobotics, 2019, 13, 1–8.
[14] Zhao D A, Lv J D, Ji W, Zhang Y, Chen Y. Design and con-
trol of an apple harvesting robot. Biosystems Engineering,
110, 112–122.
[15] Odhner L U, Ma R R, Dollar A M. Open-loop precision
grasping with underactuated hands inspired by a human
manipulation strategy. IEEE Transactions on Automation
Science and Engineering, 2013, 10, 625–633.
[16] Ciocarlie M, Hicks F M, Holmberg R, Hawke J, Schlicht M,
Gee J, Stanford S, Bahadur R. The Velo gripper: A versatile
single-actuator design for enveloping, parallel and fingertip
grasps. International Journal of Robotics Research, 2014,
33, 753–767.
[17] Rojas N, Ma R R, Dollar A M. The GR2 gripper: An unde-
ractuated hand for open-loop in-hand planar manipulation.
IEEE Transactions on Robotics, 2016, 32, 763–770.
[18] Ma R, Dollar A. Yale openhand project: Optimizing
open-source hand designs for ease of fabrication and adop-
tion. IEEE Robotics & Automation Magazine, 2017, 24,
32–40.
[19] Honarpardaz M, Tarkian M, Ölvander J, Feng X. Experi-
mental verification of design automation methods for ro-
botic finger. Robotics and Autonomous Systems, 2017, 94,
89–101.
[20] Heidari H, Pouria M J, Sharifi S, Karami M. Design and
fabrication of robotic gripper for grasping in minimizing
contact force. Advances in Space Research, 2018, 61,
1359–1370.
[21] Zhang B H, Zhou J Z, Meng Y M, Zhang N, Gu B X, Yan Z
H, Idris S. Comparative study of mechanical damage caused
by a two-finger tomato gripper with different robotic
grasping patterns for harvesting robots. Biosystems Engi-
neering, 2018, 171, 245–257.
[22] Lévesque F, Sauvet B, Cardou P, Gosselin C. A
model-based scooping grasp for the autonomous picking of
unknown objects with a two-fingered gripper. Robotics and
Autonomous Systems, 2018, 106, 14–25.
[23] Mu L T, Cui G P, Liu Y D, Cui Y J, Fu L S, Gejima Y.
Design and simulation of an integrated end-effector for
picking kiwifruit by robot. Information Processing in
Agriculture, 2020, 7, 58–71.
[24] Li Z G, Miao F L, Yang Z B, Chai P P, Yang S J. Factors
affecting human hand grasp type in tomato fruit-picking: A
statistical investigation for ergonomic development of har-
vesting robot. Computers and Electronics in Agriculture,
2019, 157, 90–97.
[25] Li Z G, Miao F L, Yang Z B, Wang H Q. An anthropometric
study for the anthropomorphic design of tomato-harvesting
robots. Computers and Electronics in Agriculture, 2019, 163,
Journal of Bionic Engineering (2020) Vol.17 No.*
14
104881.
[26] Gutierrez J, Ortiz E A, Hernandez R P R, Nunez M A. Sys-
tem to measure the range of motion of the joints of the hu-
man hand. Revista de Investigacion Clinica, 2014, 66,
122–130.
[27] Jaworski L, Karpiński R. Biomechanics of the human hand.
Journal of Technology and Exploitation in Mechanical En-
gineering, 2017, 3, 21–33.
[28] Hammert W C, Calfee R P, Bozentka D J, Boyer M I. ASSH
Manual of Hand Surgery, Lippincott Williams & Wilkins,
Philadelphia, USA, 2010, 93–215.
[29] Endo H. Application of robotic manipulability indices to
evaluate thumb performance during smartphone touch op-
erations. Ergonomics, 2015, 58, 736–747.
[30] ElKoura G, Singh K. Handrix: Animating the human hand.
Proceedings of the ACM SIGGRAPH/Eurographics Sym-
posium on Computer Animation Eurographics Association,
San Diego, California, US, 2003, 110–119.
[31] Lin H T, Kuo L C, Liu H Y, Wu W L, Su F C. The
three-dimensional analysis of three thumb joints coordina-
tion in activities of daily living. Clinical Biomechanics, 2011,
26, 371–376.
[32] Halilaj E, Rainbow M J, Got C, Schwartz J B, Moore D C,
Weiss A P C, Ladd A L, Crisco J J. In vivo kinematics of the
thumb carpometacarpal joint during three isometric func-
tional tasks. Clinical Orthopaedics and Related Research,
2014, 472, 1114–1122.
[33] Curran P F, Bagley A M, Sison-Williamson M, James M A.
Three-dimensional functional workspace of thumb prehen-
sion. Clinical Biomechanics, 2019, 63, 63–72.
[34] Bae S, Armstrong T J. A finger motion model for reach and
grasp. International Journal of Industrial Ergonomics, 2011,
41, 79–89.
[35] Yang D P, Zhao J D, Gu Y K, Wang X Q, Li N, Jiang L, Liu
H, Huang H, Zhao D W. An anthropomorphic robot hand
developed based on underactuated mechanism and con-
trolled by EMG signals. Journal of Bionic Engineering,
2009, 6, 255–263.
[36] Feix T, Kivell T, Pouydebat E, Dollar A. Estimating thumb –
index finger precision grip and manipulation potential in
extant and fossil primates. Journal of the Royal Society In-
terface, 2015, 12, 1–12.
[37] Towles J D, He ntz V R, M urray W M. Use of intrinsi c thumb
muscles may help to improve lateral pinch function restored
by tendon transfer. Clinical Biomechanics, 2008, 23,
387–394.
[38] Shapiro A, Rimon E, Shoval S. On the passive force closure
set of planar grasps and fixtures. International Journal of
Robotics Research, 2010, 29, 1435–1454.
