A novel cluster-tube self-adaptive robot hand
Fu et al. Robot. Biomim.
A novel cluster-tube self-adaptive robot hand
Hong Fu 0
Haokun Yang 0
Weishu Song 0
Wenzeng Zhang 0
0 Department of Mechanical Engineering, Tsinghua University , Beijing 100084 , China
This paper proposes a novel cluster-tube self-adaptive robot hand (CTSA Hand). The CTSA Hand consists of a base, a motor, a transmission mechanism, multiple elastic tendons, and a group of sliding-tube assemblies. Each sliding-tube assembly is composed of a sliding tube, a guide rod, two springs and a hinge. When the hand grasping an object, the object pushes some sliding tubes to different positions according to the surface shape of the object, the motor pulls the tendons tight to cluster tubes. The CTSA Hand can realize self-adaptive grasping of objects of different sizes and shapes. The CTSA Hand can grasp multiple objects simultaneously because the grasping of the hand acts as many grippers in different directions and heights. The grasping forces of the hand are adjusted by a closed-loop control system with potentiometer. Experimental results show that the CTSA Hand has the features of highly self-adaption and large grasping forces when grasping various objects.
Robot hand; Underactuated mechanism; Self-adaption; Cluster-tube grasping
Robot hands have a wide range of uses in the field of
robotics. They are used to connect object with robot
temporarily and can release the object at the
appropriate time. It is one of the important interaction terminals
between robots and their outside world.
There are many robot hands mimic structure and
configuration of human hands, designed to have multiple
fingers and multiple joints in their fingers, for example, the
Utah/MIT Dexterous Hand , the DLR/HIT Hand II ,
the Gifu Hand II  and the Gifu Hand III , the Shadow
Hand , etc. However, these dexterous hands are
expensive with complex control and sensing system [6–8].
Since 1970s, a large number of underactuated hands
have been developed to highly decrease the difficulty in
real-time control of robot hands. These underactuated
hands can be easy to control. Some grippers are designed
to be self-adaptive, which do not need to know the shapes
and sizes of the objects in advance, and do not need
sensors to detect positions of objects while grasping. The
selfadaptive performance for objects of different shapes and
sizes enables the gripper grasp a wide range of objects.
Underactuated hands developed are mainly divided
into two kinds: (
) multi-fingered underactuated hands,
such as the underactuated prosthetic hand , the FRH-4
Hand , the tendon-mechanism hand , the GR2
Gripper , and the SDM Hand . (
underactuated hands without obvious finger, for example, the
universal gripper with a spherical structure [14, 15], the
adaptive gripper  by FESTO company, and the SSA
Gripper . These hands are well fitted to the target
objects in grasping. However, the two kinds of traditional
underactuated hands mentioned above can only grasp
one object each time and are sensitive to orientations of
In 1985, Peter B. Scott proposed a general gripper,
called Omnigripper , as shown in Fig. 1. Many of the
independent telescopic rods of the gripper can be freely
telescoped when the object is grasped. When grasping,
this Omnigripper moving toward an object placed on a
support surface like Fig. 2a, then object can squeeze the
telescopic rods to slide to the palm. As there are many
telescopic rods, different telescopic rods have varying
degrees of sliding to the palm, which have relation to the
shape of the object, as shown in Fig. 2b. After that, the
left and right of the two sets of telescopic rods close and
provide clamping force from both sides of the object, as
shown in Fig. 2c. This kind of self-adaptation can be well
adapted to the size and shape of the object and can
provide a large clamping force.
However, there are still shortcomings:
1. No multi-directional grasping When the device
applies a gripping force to the target object, the
gripping force can only be in the moving direction of
two sets of rods, resulting in only one-dimensional
clamping mode, so the adaptive effect is not ideal.
For example, when grasping elongated object whose
direction is the same as that of the left and right
groups of rods, like the object A shown in Fig. 3, the
Omnigripper will fail. What’s more, when grasping
object which is completely covered by a set of rods,
like the object C shown in Fig. 3, the Omnigripper
will fail, too.
2. High energy consumption The device has two sets of
rods, requiring two mutually moving movable
supports (or movement bases), which needs to drive
two heavy movement bases to grasp, bringing high
Design of the CTSA Hand
In order to overcome certain deficiency of current ones,
this paper proposes a new type of robot hand named
cluster-tube self-adaptive robot hand (CTSA Hand), as
shown in Fig. 4.
The CTSA Hand is used for grasping objects, capable
of self-adapting the sizes and shapes of objects, realizing
multi-directional grasping with the following features:
1. Capable of providing grasping forces in multiple
directions, making it possible for CTSR Hand to
grasp objects with different sizes, shapes and
2. Capable of grasping multiple objects simultaneously
because the grasping of the hand acts as many
grippers in different directions and heights.
3. Simple structure and low energy consumption.
The following sections describe the composition and
working principle of CTSA Hand.
The inner structure of CTSA Hand is shown in Fig. 5. It
can be divided into three main functional
parts—slidingtube assemblies, transmission system and control system.
The moving direction of cluster-rod
Sliding-tube assemblies are the operating mechanism
of CTSA Hand. As shown in Fig. 6, each sliding-tube
assembly is composed of a sliding tube, a guide rod,
a first spring member, a second spring member and
a hinge. Each sliding tube is connected to the guide
rod, capable of sliding along the guide rods, as shown
in Fig. 7a. Each guide rod is hinged to the base with
shaft, capable of swinging in radial direction, as shown
in Fig. 7b. Each first spring member is placed between
sliding tube and guide rod, and each second spring
member is placed on the hinge connected guide rod
and base, making sliding tubes have the ability of
backing in place.
The function of the transmission system is to transform
the rotation of motor to the gathering of sliding tubes.
The transmission part mainly consists of a driving gear,
a driven gear and elastic ropes. The driving gear is fixed
on the output shaft of motor, and the driven gear is set on
the base by bearings. The elastic ropes are fixed between
the driven gears and the base: one side of the elastic rope
is fixed on the base, while the other side is fixed on the
driven gear by passing the sliding tubes. To ensure all
sliding tubes are stressed uniformly, three groups of
elastic ropes are set in CTSA Hand, the layout of which is
shown in Fig. 8.
The control system adopted Arduino environment, using
an angle sensor to monitor driving gear, consequently
provide feedback to perform driving or not. The specific
principles are as follows: the shaft of the angle sensor is
fixed to the driving gear, output real-time angular
velocity of the driving gear. When angular velocity deduces to
a certain value, which means the CTSA Hand has
provided enough grasping force, then the control system
terminates the PWM waves input to terminates driving the
motor. The control circuit of the Arduino control part is
shown in Fig. 9.
2nd spring member
1st spring member
Connector of base
Connector of rod
The working principle of CTSA Hand
The working process of the CTSA Hand is shown in
Fig. 10. The CTSA Hand approaches target object placed
on support surface with the assistance of
mechanical arm, and press on the target object, meanwhile the
cluster-tube adapt to the object’s shape vertically, as
shown in Fig. 10a, b. When the vertical self-adaption is
finished, the motor drives the driven gear by the
driving gear, causing elastic ropes tense up to gather sliding
tubes to central point, as shown in Fig. 10c, d. While the
tubes gather, they provide grabbing force in every
direction in horizontal plane, resulting in horizontal
selfadapting. An angle sensor is adopted to give feedback of
the angular velocity of the driving gear, thereby
terminates driving the motor at proper time automatically and
then realize grabbing. The grasping forces of the hand
are adjusted by a closed-loop control system with angle
Analysis and design of gears
For compact structure, we chose the ordinary gear train
as the transmission mechanism, and in addition, the
ordinary gear train was composed of two gears.
on the side
gathered to grab
Modification of module and teeth number
There is no special requirement for the train wheel drive,
so the spur gear is used. Let z1 and d1 be the teeth
number and standard pitch circle diameter of driving gear,
z2 and d2 be the teeth number and standard pitch
circle diameter of driven gear, m and a be the module and
center distance of spur gears. Their relationship is given
by the following formula:
d1 = mz1,
d2 = mz2,
a = (d1 + d2)/2.
a = 70.5 mm.
d2 < D − 2δ = 94 mm
According to the size of the motor and the base, we get
the center distance of the gears of the two gears:
In order to compact the structure and the maintenance
of the gears, the driven gear needs to be enclosed inside
the base, so d2 needs to meet the condition:
where D is the peripheral diameter of the base and δ is
the wall thickness of the base.
To meet the strength requirements, we set the module
of 1.5, then according to (
), one can get the
z1 = 34, z2 = 60, d1 = 51 mm, d2 = 90 mm.
This design result meets the mechanical design
Transmission ratio of the gears
Let i be the transmission ratio of the gears, according
one can get that the gears transmission ratio is 1.77, in
line with mechanical design requirements.
Grasping force analysis
Let T0 and T1 be the torque of the motor and the output
torque of driven gear, respectively. Their relationship is
T1 = T0 × i × η
where η is efficiency of the ordinary gear train. Let Fe be
the tensile force of the elastic rope acting on a sliding
tube, Ntube be the number of tubes, Ntendon be the
number of tendons, then
× Fe × d2 = 1.5Fe × d2.
We simplify the force of one sliding tube as shown in
Fig. 11. Let θ be the swing angle of a siding rod, Fs be the
squeeze force of the target object acting on a sliding tube
from the side, and Mo be the torque provided by a second
spring member, then
Fs′ × lOB × cos θ + Mo = Fe × lOA × cos θ
where lOA and lOB are the distance from acting point of
elastic rope and acting point of the target object to the
hinge point O, respectively. Let Fr′ be the resistance of the
target object acting on a sliding tube in the vertical
direction, k be the elastic coefficient of the first spring
member, and x be the distance of a sliding tube slides, then
Fr′ = k ×
Fs = 2iηlOAT0 −
) to (
), one can get the following conclusion: The
squeeze force of a sliding tube acting on the target object is
Then, let m be the number of tubes squeezing object, n
be the number of tubes touching the object from above, µ
is the coefficient of friction between the target object and
the surface material of sliding tubes, and F be the
grasping force the CTSA Hand provided, then
F = µ
µF si −
) to (
), one can get
In order to achieve a large grasping force, one can
choose a motor which can provide large torque, reduce
the driven gear diameter, increase the transmission ratio
of the gear train, increase lOA and reduce lOB, choose a
surface material to increase the coefficient of friction
between the sliding tube and the target object, reduce the
elastic coefficient of the first spring member and the
second spring member.
Experiments of verifying adaptation
The grasping experiment is conducted to verify the
adaptation and stability of CTSA Hand. Figure 12 shows
CTSA Hand grab a variety of objects with different
shapes and sizes well. The experiment shows:
1. The CTSA Hand can adapt to the shape and size of
different objects well.
2. The CTSA Hand has the features of highly
self-adaption through the vertical and horizontal adaption.
3. The grasping reliability of the CTSA Hand is high.
Experiments of gripping performance
This experiment is conducted to explore how the size
of the object affects the grasping force of CTSA Hand.
We used 3D printer to produce cuboids with different
sizes. The experimental idea is to measure the maximum
grasping force that CTSA Hand can provide for different
cuboids and then explore how the size of cuboids affects
the grasping force. So we designed the experimental steps
1. Tie the rope to a cuboid of a particular size.
2. Place the cuboid with the rope on the table and drive
CTSA Hand to grab it.
3. Use a dynamometer to pull the rope tied to the
cuboid slowly until the cube leaves CTSA Hand.
4. Record the peak of the dynamometer in the process.
In this experiment, the peak of the dynamometer is the
maximum grasping force that CTSA Hand can provide
for the certain cube.
In addition, in order to make the collected data more
reasonable, we collect the data as follows: For a cube
with a specific size, we repeat the above experiment steps
three times and then get three data points; the average of
these three data points is the maximum grasping force
that CTSA Hand can provide.
Experimental data processing and analysis
To explore the effect of cross-sectional dimensions of the
cube on grab performance, we make 25 cubes that have
the same height, different cross-sectional dimensions
using 3D printing. Then, we get the different crawl data
of 25 different sizes of cubes through the above method,
and then use the mathematical analysis tool to fit the
experimental data, and finally get the result shown in
Fig. 13. In Fig. 13, the maximum grasping force is
negative when the size of cuboid is too small, which is caused
by mathematical tools. Here, a negative value indicates
that the CTSA Hand cannot succeed in grabbing object.
In addition, to explore the effect of cube height on
grip performance, we created four sets of cubes each
containing 10 cubes with the same cross-sectional
dimensions and different heights. Then, we get the
different crawl data of the four sets of cubes through the above
method and then use the mathematical analysis tool to fit
the experimental data and finally get the result shown in
In Figs. 13 and 14, the unit of maximum gripping force
is Newton, and the unit of height, width and length is
From the results shown in Figs. 13 and 14, we can get
the following conclusions:
1. The maximum grasping force of CTSA Hand is
changing with the cross-sectional size of the target
object and almost no changing with the height of the
2. When the sizes of cuboids are moderate, the gasping
force of the hand is large, and the degree of influence
by the cross-sectional size is small.
3. When the sizes of the cuboids are too small or too
large, the grasping force of the hand is small, even fail
After analysis, the phenomenon that the maximum
grasping force suddenly become small when the
crosssectional sizes of cubes become too small or too large, is
caused by restrictions of d and δ, where d is the
maximum internal radius of cluster-tube and δ is the
minimum distance between tubes, as shown in Fig. 15. The
size d affects CTSA Hand’s grasping force: The larger
the size d, the greater the size of the object that CTSA
Hand can crawl, and the greater the grasping force for
large objects. The size δ affects the robot’s grasping force:
The smaller the size δ, the smaller size of the object that
CTSA Hand can crawl, and the greater the grasping force
for small objects.
The results of experiments
Through the analysis of experimental data, we can get
that CTSA Hand has the advantages of large
grasping force and high stability when grasping a variety of
medium-sized objects. However, there are still some
shortcomings such as small grasping force when grasping
too small or too large objects. To solve the current
problems, we will reduce the size δ and add more sliding-tube
assemblies to increase the size d in our future design.
In order to overcome some shortcomings of the
existing robot hands, this paper proposes a novel cluster-tube
self-adaptive robot hand (CTSA Hand).
The CTSA Hand consists of a base, a motor, a
transmission mechanism, multiple elastic tendons and a group
of sliding-tube assemblies. Each sliding-tube assembly is
composed of a sliding tube, a guide rod, two springs and
a hinge. When the hand grasping an object, the object
pushes some sliding tubes to different positions
according to the surface shape of the object, the motor pulls the
tendons tight to cluster tubes.
The CTSA Hand has the features of highly self-adaption
and large grasping force when grasping various objects.
The CTSA Hand has the advantages of easy control,
simple structure, low cost and high reliability of grasping.
Therefore, it has the value and potential of mass
production and large-scale applications and can be widely used
in the fields of service robotics, industrial production and
HF designed the mechanical structure of CTSA Hand and made a prototype
of CTSA Hand and drafted the manuscript. HY did the theoretical analysis of
CTSA Hand and helped draft the manuscript. WS designed and did the
experiments and did data collection and analysis and helped draft the manuscript.
WZ proposed constructive comments for mechanical design and helped
revise the manuscript. All authors read and approved the final manuscript.
All authors declare that they have no competing interests.
Ethics approval and consent to participate
This research was supported by National Natural Science Foundation of China
(No. 51575302) and Natural Science Foundation of Beijing (No. J170005).
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
1. Jacobsen SC , Iversen EK , Knutti DF , et al. Design of the Utah/MIT dextrous hand . In: IEEE transactions on robotics , 1986 . pp. 1520 - 32 .
2. Liu H , et al. Multisensory five-finger dexterous hand: the DLR/HIT Hand II . In: 2008 IEEE/RSJ international conference on intelligent robots and systems, Nice , 2008 . pp. 3692 - 97 .
3. Kawasaki H , Komatsu T , Uchiyama K. Dexterous anthropomorphic robot hand with distributed tactile sensor: Gifu Hand II” . In: IEEE international conference on systems, man and cybernetics (SMC). Tokyo: IEEE Press; 1999 . pp. 782 - 87 .
4. Mouri T , Kawasaki H , Yoshikawa K , Takai J , et al. Anthropomorphic robot hand: Gifu Hand III . In: International conference on control, automation and systems (ICCAS) . Korea: Jeonbuk; 2002 . pp. 1288 - 93 .
5. Shadow Robot Company. The shadow dextrous hand . [Online]. Available: http://www.shadow.org.uk/products/newhand.shtml.
6. Liu H , Wang T , Fan W , Zhao T. Study on the structure and control of a dexterous hand . In: Control automation robotics & vision (ICARCV) , 2010 11th International Conference, Singapore; 2010 . pp. 1589 - 93 .
7. Mattar E. Dexterous robotics hands: ANN based artificial muscles control . In: Computer modelling and simulation (UKSim) , 2011 UkSim 13th International Conference, Cambridge; 2011 . pp. 224 - 29 .
8. Hua W , Xiaodiao H , Jin H . Finger tracking control of underwater dexterous hand based on fuzzy CMAC . In: 2007 Chinese control conference, hunan; 2007 . pp. 121 - 24 .
9. Massa B , Roccella S , Carrozza MC , et al. Design and development of an underactuated prosthetic hand . In: IEEE international conference on robotics and automation (ICRA) ; 2002 . vol. 4 , pp. 3374 - 79 .
10. Che D , Zhang W. GCUA humanoid robotic hand with tendon mechanisms and its upper limb . Int J Soc Robot . 2011 ; 3 : 395 - 404 .
11. Gaiser I , Schulz S , Kargov A , et al. A new anthropomorphic robotic hand . In: IEEE international conference on humanoid robots, Daejeon, Korea; 2008 . pp. 418 - 22 .
12. Rojas N , Ma RR , Dollar AM . The GR2 gripper: an underactuated hand for open-loop in-hand planar manipulation . IEEE Trans Robot . 2016 ; 32 ( 3 ): 763 - 70 .
13. Dollar AM , Howe RD . The SDM hand: a highly adaptive compliant grasper for unstructured environments . Int J Robot Res . 29 ( 5 ): 585 - 97 .
14. Amend JR Jr, Brown E , Rodenberg N , et al. A positive pressure universal gripper based on the jamming of granular material . IEEE Trans Robot . 2012 ; 28 ( 2 ): 341 - 50 .
15. Okatani Y , Nishida T , Tadakuma K. Development of universal robot gripper using MRα fluid . In: 2014 International symposium on advanced intelligent systems (ISIS) ; 2014 . pp. 231 - 35 .
16. FESTO Company. FlexShapeGripper. [Online]. Available:http://v.youku. com/v_show/id_XOTMwMTI2NDU2.html? from=s1.8-1-1.2 , 2016 - 03 -10.
17. Zhu T , Yang H , Zhang W. A spherical self-adaptive gripper with shrinking of an elastic membrane . In: International conference on advanced robotics and mechatronics IEEE ; 2016 . pp. 512 - 17 .
18. Scott PB . The Omnigripper: a form of robot universal gripper . Robotica . 1985 ; 3 : 153 - 8 .