AA_V6

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/280500040
Design of a novel dexterous robotic gripper for in-hand twisting and

positioning within assembly automation
Article in Assembly Automation · August 2015

DOI: 10.1108/AA-05-2015-046
CITATIONS READS
17 1,515
5 authors, including:
Fei Chen Luca Carbonari

The Chinese University of Hong Kong Università Politecnica delle Marche
90 PUBLICATIONS 1,300 CITATIONS 90 PUBLICATIONS 656 CITATIONS
SEE PROFILE SEE PROFILE
Carlo Canali Mariapaola D'Imperio

Istituto Italiano di Tecnologia Istituto Italiano di Tecnologia
106 PUBLICATIONS 1,701 CITATIONS 38 PUBLICATIONS 159 CITATIONS
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Fei Chen on 14 October 2015.
The user has requested enhancement of the downloaded file.

May 11, 2015 12:44 Emerald/INSTRUCTION FILE AA˙V6
Assembly Automation
⃝
c Emerald Group Publishing Limited
Design of a Novel Dexterous Robotic Gripper for In-Hand Twisting

and Positioning within Assembly Automation
Fei Chen
Department of Advanced Robotics, Istituto Italiano di Tecnologia
via Morego, 30, 16163 Genova, Italy
fei.chen@iit.it
Luca Carbonari
luca.carbonari@iit.it
Carlo Canali
carlo.canali@iit.it
Mariapaola D’Imperio
mariapaola.dimperio@iit.it
Ferdinando Cannella
ferdinando.cannella@iit.it
Received (Day Month Year)

Revised (Day Month Year)
Purpose: The growing demand from traditional high-mix low-volume and new massive
customized manufacturing industry requires the robot with configurability and flexibil-
ity. In electronic manufacturing industry particularly, the design of the robotic hand
with sufficient dexterity and configuration is important for the robot to accomplish the
assembly task reliably and robustly. It is important for the robot to be able to grasp and
manipulate a large number of assembly parts or tools.
Design/Approach/Methodolody: In this research, a novel jaw like gripper with
human-sized anthropomorphic features is designed for online in-hand precise position-
ing and twisting. It retains the simplicity feature of traditional industrial grippers and
dexterity features of dexterous robotic hands.
Findings: The gripper is able to apply suitable gripping force on assembly parts and
performs reliable twisting movement within limited time to meet the industrial require-
1
2 Fei Chen, et al.
ments. Manipulating several cylindrical assembly parts by robot, as an experimental

case in this paper, is studied to evaluate its performance. The effectiveness of proposed
gripper design and mechanical analysis is proved by the simulation and experimental
results.
Originality/Value: The main originality of this research is that a novel jaw gripper
with human-sized anthropomorphic features is designed to be suitable for precise in-hand
posture transitions, such as twisting and re-positioning. With this gripper, the robotic
system will be sufficiently flexible to deal with various assembly tasks.
Keywords: Robotic Gripper; In-Hand Manipulation
1. Introduction
During the past tens of year, the development in manufacturing industry has been
divided into three main streams: Massive Production, Medium Production and S-
mall production. Small production, usually fully using robots, cannot keep in step
with the growing social demanding for High-Mix, Low-Volume manufacturing. For
the manufacturing fully with human workers, the labor cost is also increasing rapid-
ly. When addressing this problem, it is important to build more flexible systems to
improve the dexterity and reconfigurability of current manufacturing system by
considering combining and coordinating the above two respects, considering their
trade-off and improving the overall assembly effectiveness and efficiency [G. Micha-
los, et al. (2010)]. From our research, if we can provide the robot with sufficient
flexibility in reconfiguration, it can help the factory boosting the product quality
and meanwhile reducing the cost [F. Chen, et al. (2011)].
Currently in robotic assembly cell for flexible and reconfigurable manufacturing
within medium or small production, there still remain many unsolved issues for
robots. Robot cannot work efficiently in dealing with assembly parts with compli-
cated shapes. They often rely on external sensor systems to help with the assembly
work [F. Chen, et al. (2012)]. However human workers are skilled in performing
complicated tasks using their hands. It is essential that the conventional robotic
manipulator is able to perform complicated manipulation like human worker does.
During the past years of development in robotic manipulators, nevertheless, the ba-
sic appearance of the manipulator has seldom changed, though it is even difficult on
the other hand to modify the basic architecture of robots. Therefore, people tend
to improve the capability of robots by designing various functional end-effectors
(hands) for robots.
For industrial applications, the robotic hands must be inexpensive, compact,
low weight and robust. It also must be capable of performing simple grasping
and manipulation tasks, such as precision in-hand translation, shift or rotation,
and sufficiently general to manipulate different objects and tools. From previous
researches, grippers are categorized in four groups: Impactive, Ingressive, Astrictive
and Contigutive [G. Monkman, et al. (2007)]. The majority of industrial grippers
are either impactive or astrictive. Due to present, most robotic hands in both in-
dustrial and academic worlds are either few actuators or pneumatic powered jaws,
Design of a Novel Dexterous Robotic Gripper for In-Hand Twisting and Positioning within Assembly Automation 3
Fig. 1. Assembly scenario in this research under EU-AUTORECON project. One of its objectives
is to design a dexterous robotic hand which can grasp the assembly part and manipulate it in-hand
to a certain posture so as to fit it with the packaging slot. For example, there are four possible
postures when object (a) lies on the conveyer belt. The (a-1) posture fits the slot perfectly, so
robot just needs to pick it up and place it to the slot. However, for object (b) and (c), none of the
possible postures on the conveyer meet the slot requirements. It is easier for the robot gripper to
twist posture (c-1) or (c-2) into (c-0), but it is more difficult to carry out the same twisting for
object (b) from (b-1), (b-2) or (b-3) into (b-0) due to the fact that the contour of its cross section
has discontinuous transformation. Since the mechanisms of these objects are multi-DOF cases,
there are infinite ways to transit from one configuration to another configuration. It is therefore
necessary to constrain the DOF by choosing appropriate contact points. In this particular case,
there are two solutions – the pair of gripping points A and B. In both cases, robot will grasp the
object and rotate it around y-axis. However, the packaging slot is usually surrounded with wall
structures, it is hard for the robot to put the object into the slot by grasping the B points.
or simplified human hand shape liked multi-fingered hands. To this purpose, much
work have been done to develop such functional robotic end-effectors in the past 25
years. A robotic gripper could be like a jaw [F. Chen, et al. (2012)][A. Pettersson,
et al. (2011)] with two fingers, three fingers [M. Wagner, et al. (2009)], four fingers
[S. Ragunathan and L. Karunamoorthy (2008)] or even same with human’s hand
[Y. Yokokohji, et al. (2009)].
In industry, parallel manipulator [G. Gogu (2008)] has been already successfully
implemented in industry. They can perform and demonstrate pick and place opera-
tions in really fast speed with high reliability [I. Bonev (2001)]. Meanwhile, it is much
cheaper and simpler than serial manipulators. It is widely adopted in high speed,
high-accuracy positioning within only limited workspace, such as food packing and
electronic parts placing. Another novel universal grippers is developed, which is
able to pick up unfamiliar objects of widely varying shape and surface properties
[E. Brown, et al. (2010)]. Different with traditional grippers, the individual fingers
4 Fei Chen, et al.
are replaced by a single mass of granular material that, when pressed onto a target
object, flows around it and conforms to its shape. Upon application of a vacuum
the granular material contracts and hardens quickly to pinch and hold the object
without requiring sensory feedback. However, neither parallel manipulator nor this
universal hand is able to do in-hand dexterous manipulation. When addressing the
dexterous robotic hands, DLR hand [J. Butterfass, et al. (2001)], Barrett hand [W.
Townsend (2000)] and i-HY hand [L. U. Odhner and A. M. Dollar (2011)], which
are all human-sized anthropomorphic robotic hand, are the most representative.
Both DLR hand and Barrett hand have the dexterity to secure target objects of
different sizes, shapes, and orientations. Furthermore, i-HY is specially designed for
passive adaptive grasp. It has also demonstrated a good ability in in-hand object
transition, such as pinch grasp to power grasp gesture transition. However, these
robotic hands cannot be directly applied in industrial automation for its lack of, at
least, speed and precision.
Although various grippers are designed and built for precise picking and plac-
ing, seldom attention has been paid to the in-hand transition problem for potential
industrial requirement, particularly in lots of assembly case within 3C electronic
manufacturing industry. In such industry the jaw like gripper is widely adopted
because the assembly parts are usually in regular shape with rigid structure. How-
ever, there still remains a large number of assembly parts or tools which are in other
shapes, i.e. cylindrical structure, made by either rigid or soft materials [J. Huang,
et al. (2010)](Fig. 2). Considering a typical assembly scenario, if these parts are not
fixed to some fixed positions but just naturally lying on the conveyor (Fig. 2-(e)),
various possible postures will make it difficult for the robot to grasp. The robotic
system has to carefully examine the assembly parts utilizing external sensors, and
then calculate an appropriate posture to pick it. This method is straight but not
effective in time due to the its complexity when the robot search and decide proper
posture or even afterwards adjust its posture by whole body motion. If the gripper
can pick up the object and adjust its posture in-hand, it would be very effective
by saving a huge amount of time. As a consequence, it is essential to design a jaw
like gripper combining the advantages of both conventional industrial gripper and
dexterous robotic hands, which can not only be used for general picking and placing
usage, but also for in-hand manipulation. It also needs to be able to work under
very extreme working conditions, including high velocity and acceleration, little
time consumption but high accuracy requirements, and so on.
On the contrary, it is not easy to achieve perfect in-hand manipulation by
using dexterous robotic hands. Study a typical manufacturing scenario for automatic
packaging that the robot grasps the assembly part from the conveyer belt and put
it to the package slots (Fig. 1). In most cases, the assembly parts are symmetric
shaped like a cylinder or prism, can be simply picked and placed by the robot.
However, when the shape of the assembly part is irregular, only limited number of
its postures will fit the slot (refer to explanation under Fig. 1). If the posture is
not fit, there remains two main solutions, 1) change the posture of the assembly
parts before they are grasped by the manipulator (manipulation), or 2) adjust the
posture of the assembly parts after they are grasped (grasping). If the gripper can
picks the assembly part up and adjusts its posture in-hand when it is moving from
the conveyer to the slot, it will save a huge amount of time. This solution could
be very adaptive and reliable, and it can cope with more situations than the first
solution. However, little research has been done to study the successful in-hand
manipulation to meet critical industrial requirement, such as operation precision
and time cost.
Inspired by human’s hands, lots of dexterous robotic hands have been developed
to perform both grasping and manipulation tasks. To achieve perfect grasping
is always attracting to those who are studying dexterous robotic hands. For decades,
the methods, which serve this objective, have gradually been divided into two cat-
egories, the analytic solution and the data-driven solution [J. Bohg, et al. (2013)].
As the first method to be widely studied, the analytic solution discusses how to
construct a force-closure grasp going with the dynamic requirements by dexterous
robotics hands. It requires that the geometric, kinematic and dynamic features,
which will interfere with the grasping, should be well studied in combination to for-
m a grasp map [M. Richard, et al. (1994)]. For example, in [D. Prattichizzo, et al.
(2012)] a set of springs has been introduced to model the contact so as to study the
dexterity. In [V. Lippiello, et al. (2013)], the grasp map is formulated into a convex
optimization problem, by solving which, the grasping force will be optimized. On
the other hand, Data-driven solution is more like a heuristic method, which is not
a optimal solution, but relatively effective in practical implementation. It helps the
robot to learn from samples to resemble reality. For this point of view, robot can
either learn from human demonstration [D. R. Faria, et al. (2012)], or estimate the
object posture to find out suitable grasping points by means of vision system for
known object [P. Azad, et al. (2007)], familiar object [D. Song, et al. (2011)] and
unkown object [M. Popovic, et al. (2011)]. Nevertheless, it brings the possibility to
apply this solution to meet the industrial requirements.
The main contribution of this research is that the development of a novel jaw
gripper with human-sized anthropomorphic features especially suitable for precise
in-hand posture transitions, such as twisting and re-Positioning (TP). This gripper
is well studied following the analytic solution, in terms of kinematics and dynamics.
The advantages of this grippers, compared with traditional robotic gripper, is that
it can apply a constant gripping force on cylindrical assembly parts with various
surface material (Table 1) and can perform positioning and twisting with relatively
high speed. Therefore, it can guarantee a trade-off between the effectiveness of
parallel manipulator and dexterity of human-sized anthropomorphic robotic hands.
It makes the robot applicable for more advanced assembly tasks while maintaining
the reliability with industrial manufacturing requirements.
This paper is organized as follows. In Section 2, after a brief introduction of the
6 Fei Chen, et al.
Fig. 2. A category of assembly parts that are generously cylinder shaped. A brief introduction to
the common feature extraction of these objects is presented in Table 1.
Table 1. Assembly parts category
Category External Appearance Surface Material Feature

a Straight Plastic Rigid
b Irregular Rubber Deformable
c Straight Plastic Rigid
d Bending Rubber Deformable
e Straight Metal Rigid
design requirements and assumptions, mechanical design of TP gripper is discussed.

In Section 3 and Section 4, the kinematics and dynamics of this gripper is studied
respectively. In Section 5, both simulation and experimental study are carried out to
evaluate the performance of the gripper. Conclusions and future works are presented
in Section 6.
2. Mechanical Design
2.1. Problems, Requirements and Assumptions
Small production has to keep step with the growing social demanding for High-Mix,
Low-Volume manufacturing, it has to adopt various robotic manipulators with dif-
ferent end-effectors to carry out different assembly tasks. A typical small production
shown in Fig. ?? can be divided into three parts: 1) Feeding system. It is usually
constructed by conveyors. 2) Robotic manipulators. The key point for this part is
to design various functional end-effectors, or grippers. 3) The assembly zone. The
basic assembly action is: picking, fixing, placing, mating, and so on.
The basic assumption in this research is that the general shape of the assembly
parts is cylindrical, and with different surface materials. Usually when this kind of
assembly parts lie on the conveyor, there could be several possible postures (Fig.
2-(e)). However, due to the design of the assembly system (Fig. 1), only limited
number of postures are fitting for correctly assembly. If the posture is already fit
enough, what leaves for the gripper is just to grasp it and perform the predesigned
assembly action. If the posture is not fit, there remains some possible solutions: 1)
Fig. 3. A typical architecture of TP gripper. (a) Concept design of TP gripper. (b) Twisting
action inspired by human. (c) Fingertip design for TP gripper. (d) Dexterity of TP gripper. (e)
Specification of TP gripper.
Table 2. Function of Different Parts Within TP Gripper Shown in Fig. 3
Motor Function
DC Motor 1 Control the stable contact on assembly parts
between two inner grippers
DC Motor 2 Control the distance between two outer fingers
Linear DC Motor Control the twisting movement in vertical direction
using the inner fingers
change the posture of the assembly parts before they are grasped by the manipula-
tor, or 2) adjust the posture of the assembly parts after they are grasped. There are
a lot of related researches about how to estimate the postures of the assembly parts
related with the first solution by using some specific sensor systems and estimation
algorithms, and then robotic manipulator grasp them from appropriate direction.
8 Fei Chen, et al.
There are also special machines designed to support adjusting the posture of the
assembly parts with external force, like using vibration. Due to some technical d-
ifficulties, seldom of these researches have ever considered adjusting the posture of
the gripper online using the second solution. The second solution is very adaptive
and reliable, and it can cope with more situations than the first solution.
Inspired by human being’s hands (Fig. 3-(b)), it is better the traditional parallel
jaw like gripper which can do the twist action by sliding one fingertip on the surface
of another finger. A twisting action refers to the action grasp of an object in a
human hand, which adjust its posture by twisting the fingers. It is easier to do so if a
complete human-sized anthropomorphic robotic hand is built. It is difficult to apply
this technology on hand shape like robotic hands, but it is relatively easier to make
a jaw shape like robotic gripper to do so but only if it is equipped with additional
movement freedom. Therefore, It is important to design a robotic gripper not only
reserve the simplicity of jaw like gripper but also keep some feature like dexterous
manipulation. One key issue in designing such a gripper is how to guarantee a
constant and reliable twisting movement by the two fingers without dropping the
gripped assembly parts.
2.2. TP Gripper Mechanical Design

In this paper, we design a robotic gripper for positioning and twisting to meet some
industrial assembly requirements. The architecture of this TP gripper is shown in
Fig. 3. A typical TP gripper (just half shown in Fig. 3-(a)) is driven by four motors
(two DC motors and two linear DC motors). The functions of these motors are
shown in Table 2. Compare with the conventional gripper, this 4-DOF TP gripper
can not only do open and close for positioning, but also do twisting for effective
assemble cylindrical parts (Fig. 3-(d))).
A TP gripper also includes inner fingers and outer fingers (Fig. 3-(c))). With
the help of DC motor 2, this TP gripper is suitable for almost every diameter of the
cylindrical shaped electronic assembly parts by adjusting the distance between its
outer fingers. The inner fingers, driven by the linear motors, can slide on the track
inside the outer fingers. DC motor 1 can adjust the contact point upon the object
and guarantee a constant gripping force applied on the assembly parts. It provides
sufficient friction force with which the cylinder assembly parts can be twisted by
two inner gripper fingers. Both of the linear motors drive the inner finger movement
with independent velocity. In this case, the rotation angle can be precisely controlled
by both inner and outer fingers.
2.3. Twisting by TP Gripper

It is assumed that the robot can always grasp the object at or close to the centre of
gravity, so we do not have to consider the misalignment of the object along y-axis. A
typical grasp with in-hand transition process is divided into four phase (Fig. 6-(b)):
(1) approaching, (2) grasping, (3) picking up, and (4) twisting with an angle theta
c2
c1
B2 d1
γ1 B1
D1≡C2
C1
b2 γ2
b1
y a1
A2 α2
A1
α1
O1 x
a2 O2
Fig. 4. Kinematic schema of a 4-DOF TP gripper. The first four bar link (pink solid line) deals with
vertical movement of the point D1 . The rotation of the fingertip (point E) is given by the second
four-bar-link (green solid line). The center of the motor can be considered coincident. The arrows
(semi-transparent) represent the vectors for position analysis. In letter in caps are the joints, in
bold the vectors and in Greek the angles.
or repositioning with a desired distance. During the second phase (b − 2), how in
deep (d) the gripper can grasp the object will determine how the gripper will do
the twisting. If d is very large, most of the twisting case will take place upon the
inner fingers. If d is very small, which means the TP gripper only grasps the object
close to the end of the fingertip (b − 3). In this case, if a large rotational angle θ is
required, the TP gripper has to twist the objects upon the outer finger (b − 4).
3. Kinematics Analysis
In order to control the position of the manipulated object, the kinematics of the
finger should be solved. As shown in Fig. 4, two four bar linkages are used for the
actuation of the finger. To get rid of the knowledge of passive angles, two distance
equations can be used:
T
(B1 − A1 ) (B1 − A1 ) = b21
T (1)
(B2 − A2 ) (B2 − A2 ) = b22
where A1 , B1 , A2 and B2 can be expressed in terms of only actuated angles and
outputs of the mechanism:
[ ] [ ]
O1,x + a1 cos α1 C1,x − c1 cos γ1
A1 = B1 =
O + a1 sin α1 C − c1 sin γ1
[ 1,y ] [ 1,y ] (2)
O2,x + a2 cos α2 C2,x − c2 cos γ1
A2 = B2 =
O2,y + a2 sin α2 C2,y − c2 sin γ2
10 Fei Chen, et al.
It should be noted that points C2 and D1 coincide, so that:
[ ]
C1,x + d1 cos γ1
C 2 ≡ D1 = (3)
C1,y + d1 sin γ1
A Weierstrass substitution is used to put coordinates (3) in polynomial form:
Θ1 = tan α1 /2, Θ2 = tan α2 /2, Γ1 = tan γ1 /2, Γ2 = tan γ2 /2, to obtain:
   
1−Θ2 1−Γ2
O1,x + a1 1+Θ21 C1,x + c1 1+Γ21
A1 =  2Θ1
1  B1 = 
2Γ1
1
O1,y + a1 1+Θ 2 C1,y + c1 1+Γ 2
 1  1  (4)
1−Θ2 1−Γ2 1−Γ2
O2,x + a2 1+Θ22 C1,x + d1 1+Γ21 − c2 1+Γ22
A2 =  2Θ2
2  B2 = 
2Γ1
1
2Γ2
2
O2,y + a2 1+Θ 2 C1,y + d1 1+Γ 2 − c2 1+Γ2
2 1 2
After substitution of (4), factorization and elimination of non-vanishing factors,

the system of equations (1) can be rewritten in the form:
∑
Π1 : fi,j Θi1 Γj1 , Π1 = 0
i, j = 0..2
i+j ≤3
∑ (5)
Π2 : gi,j,k Θi2 Γj1 Γk2 , Π2 = 0
i, j, k = 0..2
i+j+k ≤5
where the coefficients fi,j and gi,j,k only depend on the geometrical parameters of
the mechanism.
The kinematics of the two four bar linkages can be solved independently due
to the fact that the first of equations (5) is devoid of the actuated angle α2 . Then,
substitution of the vanishing set of Π1 into Π2 allows finding a solution also for
kinematics of the second linkage. It should be noted that polynomials Π1 and Π2
can be used for solving both the direct and the inverse kinematics problems of the
mechanism: the former aims at finding the values of Γ1 and Γ2 once the actuation
angles are given, the latter at determination of Θ1 and Θ2 for a given configuration
of the finger.
For the direct kinematics, the equality Π1 = 0 provides two solutions for Γ1
fully depending on mechanism geometric parameters and which are not explicitly
shown here for space reasons. Analogously, Π2 = 0 provides two solutions for Γ2
depending from both geometry of the mechanism and from Γ1 : substituting one at
a time such values in the solutions of Π2 = 0, up to four solutions are found for
the kinematics of the device, corresponding to four different assembly modes. The
above explanation can be easily presented in Fig. 6.
A similar path can be followed for the solution of the inverse kinematics problem
that provides, for a given configuration of the finger (i.e. a given set of Γ1 and Γ2 ),
four different combinations of the actuators (i.e. four sets of Θ1 and Θ2 ). Even in
this case, the explicit formulas of the four solutions are not given for the sake of
conciseness.
joints space regions with γ2>0
-5
γ2 [deg]
-10
-15
γ2>0
180
γ2>0
-20 90
90
0 0 ]
g
[de
α2 [de -90 -180
-90 α 1
g]
-270 -180
Fig. 5. Analytical finger configurations γ2 for every set of actuation angles α1 and α2 . The joints
space regions providing a positive value for γ2 are indicated in blue. It shows four main regions
for α1 and α2 combination to reach a specific γ2 .
4. Twisting Dynamics
The knowledge of the gripper kinematics allows performing a study on the actua-
tion forces needed for realization of particular motions. Such study, aimed at the
the optimization of the whole gripper structure, also realizes the first step in the
direction of the realization of a force control algorithm of the device. The knowledge
of the force exerted during grasping operations plays a role of paramount impor-
tance for different reasons. Firstly, such information is necessary for minimization
of actuation burden and energy consumption. As a side effect, the knowledge of the
minimal grasping wrench can be exploited for an efficient sizing of actuators, with
advantages in terms of whole device mass and compactness. At last, minimizing
the interaction force among the gripper fingers and the payload is helpful for the
manipulation of delicate objects whose surfaces may be damaged by a repetitive or
an excessively firm contact.
In the following, an analysis of the twisting dynamics is performed with par-
ticular attention to the action/reaction forces among the gripper fingers and the
payload. For simplicity, it is assumed that the cross section of the gripped object is
a perfect circle with a radius of R. So during the twisting phase, the configuration
12 Fei Chen, et al.
vl y
(b-1) (b-2)
x
Tl
Nl
Nr
Tr θ
γ2
(b-3) (b-4)
(a) vr (b)
Fig. 6. Dynamic of twisting movement using TP gripper with 4-DOF configuration. (a) Dynamic
analysis of twisting phase. (b) Four phases of a typical grasping manipulation, including approach-
ing, grasping, picking up, and twisting with an angle θ.
of the finger influences the direction of the tangential force which is responsible of
the rotation of the cylinder (see Fig. 6-(a)).
For an effective twist operation, the cylinder is supposed not to slip so that
the fingers constraint the cylinder with two holonomic constraints of pure rotation.
According to Fig. 6-(a), the following Newton-Euler formulation can be used to
describe the dynamics of the phenomenon for an object of mass m and a moment
of inertia Iz :
Iz θ̈ = (Tl + Tr ) R
mẍ = mg − Nr sin(γ2 ) + Tl − Tr cos(γ2 ) (6)
0 = Nl − Nr cos(γ2 ) + Tr sin(γ2 )
where ẍ and θ̈ respectively represent the acceleration of the center of mass and
the angular acceleration of the grasped object. As mentioned, the two contact forces
Tl and Tr must guarantee a non-slipping contact, thus:
{
|Tl | ≤ µ|Nl |
(7)
|Tr | ≤ µ|Nr |
where µ is the coefficient of static friction. With the aim of finding the condition
of minimum actuation effort needed for the grasping operation, the system should
be analysed in a condition of incipient slipping, so when one of the two tangent
forces Tl and Tr reaches the respective limit value. This consideration yields to the
four slipping conditions:
{ {
Tl = µNl Tl = −µNl
1: 2:
|Tr | ≤ µ|Nr | |Tr | ≤ µ|Nr |
{ { (8)
|Tl | ≤ µ|Nl | |Tl | ≤ µ|Nl |
3: 4:
Tr = µNr Tr = −µNr
The four systems expressed in (8) must be put together the dynamics equations
(6) one at a time, yielding to many distinct solutions of the unknown Tl , Tr , Nl and
Nr . Thus the limit force that the finger must exert on the cylinder in the four cases
is:
2 ∓1)(g−ẍ)+Iz ω̇(µ sin γ2 ±cos γ2 )

Nr,1..2 = ∓ mR(µ sin γR(sin γ2 ∓µ( cos γ2 +1))
mR(g−ẍ)+Iz ω̇
(9)
Nr,3..4 = ± R(sin γ2 ±µ( cos γ2 +1))
In order to determine the minimum grasping force necessary to efficiently operate

the manipulation task, the four solutions of the problem must be compared and the
most severe condition is to be taken into consideration:
Nr,min = max{Nr,1 , Nr,2 , Nr,3 , Nr,4 } (10)
As visible in equation (8), several parameters are involved in the minimum

required grasping force: the physical properties of the payload (mass, inertia and
geometry), the configuration of the finger (represented by the angle γ2 ) and the law
of motion which determines the linear and the angular accelerations to whom it is
subject. Moreover, such parameters are not independent by each other due to the
kinematics of the manipulation which is by nature a closed kinematics task.
Due to that, it is not easy to figure out the influence of each term without
considering a particular application. In order to improve the understanding of
the phenomenon, several simulations have been performed with different velocity
and acceleration.[ A cylindrical
] object is considered, with mass m = 25 [g], inertia
Iz = 2.8×10−6 kg m2 and the coefficient of the static friction µ = 0.5. The manip-
ulation consists of twisting the cylinder of 360o with a given profile of time-angle.
The simulation has been repeated for several combinations of finger inclination,
cylinder radius and execution time. This last parameter in particular influences the
acceleration of the object and therefore it also affects the minimum required grasp-
ing force. Results are shown in Fig. 7. As visible, the dimension of the payload has
marginal contribution if the angular acceleration is low. On the other hand, the
configuration of the finger influences the grasping force and it allows to decrease
the contact force among the finger and the payload.
14 Fei Chen, et al.
Minimum grasping force required - Nr [N]
0.24 0.25 0.26 0.27

ω ≈ 0.5 [rad/s2] ω ≈ 1 [rad/s2]
1.0 1.0
0.5 0.5
0.0 0.0
γ2 [deg]
γ2 [deg]
-0.5 -0.5
-1.0 -1.0
-1.5 -1.5
-2 -2
-2.5 -2.5
7.5 10 12.5 15 7.5 10 12.5 15
R [mm] R [mm]
ω ≈ 5 [rad/s2] ω ≈ 30 [rad/s2]
1.0 1.0
0.5 0.5
0.0 0.0
γ2 [deg]
γ2 [deg]
-0.5 -0.5
-1.0 -1.0
-1.5 -1.5
-2 -2
-2.5 -2.5
7.5 10 12.5 15 7.5 10 12.5 15
R [mm] R [mm]
Fig. 7. Effect of angular acceleration (ω̇), geometry of manipulated object (radius R) and finger
configuration (angle γ2 ) on the minimum grasping force N2 required for efficient manipulation.
5. Simulation and Experiments

5.1. Dynamics twisting simulation
Fig. 8 shows the simulation results for successful twisting. Due to the structure
constraints, the force applied on the object increases when the object moves upstairs.
b-2
b-3
b-1
(b-1) (b-2) (b-3)
(a) (b)
Fig. 8. Gripping force of successful twisting within different phases by simulation.
b-3
a-3 c-3
b-2
a-2 a-4 c-2
b-4 c-4
a-1 b-1 c-1
(a-1) (a-2) (a-3) (a-4) (b-1) (b-2) (b-3) (b-4) (c-1) (c-2) (c-3) (c-4)
(a) (b) (c)
Fig. 9. Gripping force of successful twisting within different phases for different objects.
The position of the object is calculated based on the kinematics.
5.2. Twisting and repositioning experiments

To validate the simulation results, experiments are carried out to exam the gripping
force changing for successful twisting. It shows that the simulation results are well
matched.
The generic application on different objects are carried out in Fig. 9. The twisting
on inner fingers and outer fingers as described in Section III-C is also studied. In this
experiment, in order to prove the basic scheme of the twisting action, we did not
exam the twisting process under minimum gripping force as described in equation
(8), but just apply higher gripping force to prevent sliding. The measured gripping
force matches the one calculated in the simulation. In addition, one may notice that
there is an alignment displacement on the two fingers after twisting, which may
16 Fei Chen, et al.
Fig. 10. A test of TP Gripper with three phases including (a) grasping, (b) picking up, and (c)
twisting.
prevent the gripper from moving down towards the assembly area. The solutions
are mainly two, 1) drop the assembly parts to the assembly slots when the gripper
is getting close enough, or 2) adjust the posture of the robot gripper, and make one
part of the assembly part touch the slot. The assembly part will slide into the slot
due to its compliance structure when it is released.
5.3. High speed assembly

We also test the gripper performance within high acceleration and high speed move-
ment by attaching it to an industrial robotic manipulator. Fig. 10 shows the basic
three phases within a complete twisting movement. According to the experimental
results (Table 3), it costs less than 0.6s to accomplish twisting by rotating within
180 degrees. The results show that this TP gripper can accomplish the twisting
during high-speed and high-acceleration assembly.
6. Conclusion
In this paper, a novel jaw like gripper for precise twisting and positioning is designed,
analyzed, and demonstrated. It is significant to introduce this TP gripper for flexible
assembly to small sized manufacturing. This mechanical design can guarantee both
simplicity and dexterity, the effectiveness of which is proved by the experimental
data.
Future work will focus on in deep analysis of some specific assembly case and
improvement of the mechanical design. At the same time, control algorithm for
adaptive and reliable twisting is also should be designed.
7. Acknowledgement
This research has received funding from the European Union’s Seventh Framework
Programme FP7/2007-2013 under grant agreement number 285189 - AUTORE-
CON.
Table 3. Time cost for different stages within twisting (Fig. 10)
Rotational angle θ (a)-(b) (b)-(c) Total time cost Rolling distance

[rad] [s] [s] [s] [mm]
π/2 0.12 0.22 0.34 24
π 0.13 0.40 0.53 48
References
G. Michalos, S. Makris, N. Papakostas, D. Mourtzis, and G. Chryssolouris (2010), “Auto-
motive assembly technologies review: challenges and outlook for a flexible and adaptive
approach”, CIRP Journal of Manufacturing Science and Technology, Vol. 22 No. 2, pp.
81-91.
F. Chen, K. Sekiyama, J. Huang, B. Sun, H. Sasaki, and T. Fukuda (2011), “An assembly
strategy scheduling method for human and robot coordinated cell manufacturing”,
International Journal of Intelligent Computing and Cybernetics, Vol. 4 No. 4, pp. 487-
510.
F. Chen, K. Sekiyama, P. Di, J. Huang, and T. Fukuda (2012), “i-hand: An intelligent
robotic hand for fast and accurate assembly in electronic manufacturing”, IEEE In-
ternational Conference on Robotics and Automation, Saint Paul, pp. 1976-1981.
G. Monkman, S. Hesse, and R. Steinmann (2007), “Robot Grippers”, Vch Verlagsge-
sellschaft Mbh.
A. Pettersson, T. Ohlsson, S. Davis, J. Gray, and T. Dodd (2011), “A hygienically de-
signed force gripper for flexible handling of variable and easily damaged natural food
products”, Innovative Food Science and Emerging Technologies, Vol. 12 No. 3, pp.
344-351.
M. Wagner, J. Morehouse, and S. Melkote (2009), “Prediction of part orientation error
tolerance of a robotic gripper”, Robotics and ComputerIntegrated Manufacturing, Vol.
25 No. 2, pp. 449-459.
S. Ragunathan and L. Karunamoorthy (2008), “Modular reconfigurable robotic gripper
for limp material handling in garment industries”, International Journal of Robotics
and Automation, Vol. 23 No. 6, pp. 213-219.
Y. Yokokohji, J. San Martin, and M. Fujiwara (2009), “Dynamic manipulability of multi-
fingered grasping”, IEEE Transactions on Robotics, Vol. 25 No. 4, pp. 947-954.
G. Gogu (2008), “Structural synthesis of parallel robots: Methodology. Part 1”, Springer.
I. Bonev (2001), “Delta parallel robot-the story of success”, Newsletter.
E. Brown, N. Rodenberg, J. Amend, A. Mozeika, E. Steltz, M. Zakin, H. Lipson, and H.
Jaeger (2010), “Universal robotic gripper based on the jamming of granular material”,
Proceedings of the National Academy of Sciences, Vol. 107 No. 44, pp. 18809-18814.
J. Butterfass, M. Grebenstein, H. Liu, and G. Hirzinger (2001), “Dlr-hand ii: Next gen-
eration of a dextrous robot hand”, IEEE International Conference on Robotics and
Automation, pp. 109-114.
W. Townsend (2000), “The barrett hand grasperCprogrammably flexible part handling
and assembly”, Industrial Robot: an international journal, Vol. 27 No. 3, pp. 181-188.
L. U. Odhner and A. M. Dollar (2011), “Dexterous manipulation with underactuated
elastic hands”, IEEE International Conference on Robotics and Automation, pp. 5254-
5260.
J. Huang, P. Di, T. Fukuda, and T. Matsuno (2010), “Robust model-based online fault
detection for mating process of electric connectors in robotic wiring harness assembly
18 Fei Chen, et al.
systems”, IEEE Transactions on Control Systems Technology, Vol. 18 No. 5, pp. 1207-
1215.
J. Bohg, A. Morales, T. Asfour, and D. Kragic (2013), “Data-driven grasp synthesis a
survey”, IEEE Transactions on Robotics, Vol. 30 No. 2, pp. 289-309.
M. Richard, Z. Li, and S. S. Sastry (1994), “A mathematical introduction to robotic
manipulation”, Florida: CRC Press.
D. Prattichizzo, M. Malvezzi, M. Gabiccini, and A. Bicchi (2012), “On the manipulability
ellipsoids of underactuated robotic hands with compliance”, Robotics and Autonomous
Systems, Vol. 60 No. 3, pp. 337-346.
V. Lippiello, B. Siciliano, and L. Villani (2013), “A grasping force optimization algorithm
for multiarm robots with multifingered hands”, IEEE Transaction on Robotics, Vol.
29 No. 1, pp. 55-67.
D. R. Faria, R. Martins, J. Lobo, and J. Dias (2012), “Extracting data from human
manipulation of objects towards improving autonomous robotic grasping”, Robotics
and Autonomous Systems, Vol. 60 No. 3, pp. 396-410.
P. Azad, T. Asfour, and R. Dillmann (2007), “Stereo-based 6d object localization for
grasping with humanoid robot systems”, IEEE International Conference on Intelligent
Robots and Systems, pp. 919-924.
D. Song, C. H. Ek, K. Huebner, and D. Kragic (2011), “Multivariate discretization for
bayesian network structure learning in robot grasping”, IEEE International Conference
on Robotics and Automation, pp. 1944-1950.
M. Popovic, G. Kootstra, J. A. Jorgensen, D. Kragic, and N. Kruger (2011), “Grasping un-
known objects using an early cognitive vision system for general scene understanding”,
IEEE International Conference on Intelligent Robots and Systems, pp. 987-994.
F. Chen, F. Cannella, C. Canali, T. Hauptman, G. Sofia, and D. Caldwell (2014), “In-hand
precise twisting and positioning by a novel dexterous robotic gripper for industrial
high-speed assembly”, IEEE International Conference on Robotics and Automation,
pp. 270-275.
View publication stats

AA_V6

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AA_V6

Uploaded by

Copyright:

Available Formats

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

Design of a novel dexterous robotic gripper for in-hand twisting and

Article in Assembly Automation · August 2015

Fei Chen Luca Carbonari

SEE PROFILE SEE PROFILE

Carlo Canali Mariapaola D'Imperio

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Design of a Novel Dexterous Robotic Gripper for In-Hand Twisting

Received (Day Month Year)

2 Fei Chen, et al.

ments. Manipulating several cylindrical assembly parts by robot, as an experimental

Keywords: Robotic Gripper; In-Hand Manipulation

4 Fei Chen, et al.

6 Fei Chen, et al.

Table 1. Assembly parts category

Category External Appearance Surface Material Feature

design requirements and assumptions, mechanical design of TP gripper is discussed.

Table 2. Function of Diﬀerent Parts Within TP Gripper Shown in Fig. 3

8 Fei Chen, et al.

2.2. TP Gripper Mechanical Design

2.3. Twisting by TP Gripper

10 Fei Chen, et al.

It should be noted that points C2 and D1 coincide, so that:

After substitution of (4), factorization and elimination of non-vanishing factors,

joints space regions with γ2>0

12 Fei Chen, et al.

2 ∓1)(g−ẍ)+Iz ω̇(µ sin γ2 ±cos γ2 )

In order to determine the minimum grasping force necessary to eﬃciently operate

Nr,min = max{Nr,1 , Nr,2 , Nr,3 , Nr,4 } (10)

As visible in equation (8), several parameters are involved in the minimum

14 Fei Chen, et al.

Minimum grasping force required - Nr [N]

0.24 0.25 0.26 0.27

5. Simulation and Experiments

(b-1) (b-2) (b-3)

Fig. 8. Gripping force of successful twisting within diﬀerent phases by simulation.

a-1 b-1 c-1

The position of the object is calculated based on the kinematics.

5.2. Twisting and repositioning experiments

16 Fei Chen, et al.

5.3. High speed assembly

Rotational angle θ (a)-(b) (b)-(c) Total time cost Rolling distance

18 Fei Chen, et al.

View publication stats

You might also like