You are on page 1of 16

Robotica

http://journals.cambridge.org/ROB
Additional services for Robotica:
Email alerts: Click here
Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here

Design and simulation of an underactuated nger mechanism for LARM


Hand
M. Ceccarelli and M. Zottola
Robotica / FirstView Article / November 2015, pp 1 - 15
DOI: 10.1017/S0263574715000648, Published online: 21 July 2015

Link to this article: http://journals.cambridge.org/abstract_S0263574715000648


How to cite this article:
M. Ceccarelli and M. Zottola Design and simulation of an underactuated nger mechanism for LARM Hand. Robotica,
Available on CJO 2015 doi:10.1017/S0263574715000648
Request Permissions : Click here

Downloaded from http://journals.cambridge.org/ROB, IP address: 134.58.253.30 on 06 Nov 2015

Robotica: page 1 of 15. Cambridge University Press 2015


doi:10.1017/S0263574715000648

Design and simulation of an underactuated finger


mechanism for LARM Hand
M. Ceccarelli and M. Zottola
LARM: Laboratory of Robotics and Mechatronics, University of Cassino and South Latium,
Cassino, Italy
(Accepted June 26, 2015)

SUMMARY
An underactuated mechanism is presented as a new finger design for improving grasp adaptability
of LARM Hand fingers. Underactuation is discussed as a feasible solution through several
design structures for finger adaptability to shape and size of objects to be grasped. The proposed
underactuated solution for a new LARM finger is characterized through simulation results in
ADAMS environment for operation feasibility and performance.
KEYWORDS: mechanism design, finger design, simulation, underactuated mechanisms.

1. Introduction
Grasping of objects with fingered dexterous robotic hands is a fundamental operation with still great
interest in industrial contexts and service robot applications, as indicated for example in refs. [14].
Since a recent past, in order to develop anthropomorphic finger mechanisms researchers have used
two different approaches, namely complex mechanisms in order to perform manipulation tasks with
high dexterity, or mechanisms with a reduced number of degrees of freedoms (DOFs) and actuators
with limited performance but a fairly simplified device operation.1,3
A design of an anthropomorphic finger must fulfill basic requirements for human-like motion
and grasp, compact size actuation, lightweight, efficiency, position, and force control, by respecting
functional requirements of manipulative dexterity, grasp robustness, and human operability as reported
in refs. [5,6]. Reducing the number of actuators but keeping hand capability for adapting its
configuration to grasped objects addresses problems of underactuation. In this way, it is possible
to simplify also the control without losing much of the versatile grasp capability. Examples
of underactuated robotic hands are reported in refs. [711], even with metamorphic mechanism
designs,12,13 and related general design considerations are outlined in refs. [7,14].
By using underactuated mechanisms it is possible to achieve an adaptive grasp that mimics the
human grasping action. Indeed, the human hand can also be considered as underactuated, since the
distal phalanges of the fingers are not independently controllable. In order to achieve an underactuated
grasp it is possible to consider two kinds of structure, namely using the flexibility of links or designing
underactuated mechanisms as pointed out in ref. [14]. Both mechanism structures will give a certain
adaptability of mechanism configuration to the shape of grasped objects. Flexible links can be
obtained by means of suitable mechanical design for flexible behavior of link elements. Practical
design solutions can be advised in building flexible links or joints through suitable materials and/or
light design for elastic response as outlined in ref. [15]. The deformation of compliant elements is
actuated by an input actuator, and the joints and/or links will be stressed by reaction forces. It should
be noted that in general this kind of solution gives a very limited extra motion due to constraints
regarding with strength and fatigue limits of materials in the links and joints, while considerable
actuation energy is needed. An underactuated finger with compliant joints for a prosthetic hand is
described in ref. [8].

* Corresponding author. E-mail: ceccarelli@unicas.it

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

The underactuated mechanisms are more practical solutions and indeed they are widely used in
finger structures for several applications, as pointed out in refs. [4, 15, 16]. A mechanism is defined as
underactuated when its number a of actuators is smaller than the number n of DOF of the mechanism.
Thus, the degree of underactuation of the mechanism is defined as (na). In order to obtain a statically
determined system, (na) elastic elements are used together with mechanical limits, which allow
given ranges of motion and prevent branching and other undesirable effects.14
It is possible to identify two types of underactuated finger mechanisms, depending on whatever
a tendon or link transmission is used. Tendon systems are generally adopted to minimize the
transmission dimensions but they are limited in giving small grasping forces. Examples of tendon
finger mechanisms are reported in refs. [17, 18]. A finger mechanism with tendons is presented also
in ref. [19], while a pulley-cable solution is described in ref. [20]. For applications in which large
grasping forces are required, the underactuated linkage mechanisms are usually preferred, like for
example in ref. [7] where a 1-DOF mechanism with suitable four-bar linkages and flexible elements
is used to move all phalanxes of fingers; in ref. [9] where an underactuated linkage is presented as
combined with force control; and in ref. [21], where a five-fingered underactuated hand is proposed
as prosthetic hand.
Since the end of 1990s at LARM in Cassino design and research activities have been carried out in
order to design a low-cost easy-operation robotic hand with anthropomorphic fingers. The developed
solution is denominated LARM Hand and its design evolution is outlined in ref. [22]. At LARM a
research line is devoted to design fingers with 1-DOF underactuated linkage mechanisms that can
adapt their configurations to the shape of grasped objects. Examples of underactuated solutions for
finger mechanisms developed at LARM are reported in refs. [10, 11].
An underactuated mechanism design can be based on passive joints that are used to kinematically
constrain the motion transmission through linear and/or torsional spring elements within the
mechanism structure. One of the most complex issues for developing an underactuated mechanism is
the design of a suitable driving mechanism that can be embedded in the finger body and will remain
within the finger body also during its movement as stressed in ref. [10]. Usually the underactuated
mechanisms for anthropomorphic fingers have a large size and it is difficult to keep them within the
finger body.
Several design solutions have been proposed in the literature for multi-fingered robotic grippers
and hands with attempts to have high flexibility multi-purpose operation. Significant examples can
be mentioned as Stanford/JPL hand, DLRs Hand II, Manus Colobi, Utha/MIT Hand, Cifu Hand, and
Barrett Hand. Available multi-fingered robotic hand prototypes are still not able to fully reproduce
the highly flexible multi-purpose operation of the human hand. Most of the available prototypes
have a high number of DOFs, a complex control, and a high cost. On the other side, hands with
underactuated finger mechanisms have the capability to adapt their shape to grasped objects although
a finger is controlled by a reduced number of actuators and it can be built with low-cost and easyoperation features. Currently, few underactuated fingers have been reported in the literature. An
underactuated finger mechanism is developed in TBM hand (Toronto/Bloorview MarcMillan Hand)
hand, as reported in ref. [23]. The thumb of TH-1 (TsingHua) hand of Tsinghua University with a
prismatic spring-driven joint is described in ref. [9].
In this paper, a design of a new underactuated finger mechanism is proposed for robotic hands
with focus on requirements for 1-DOF actuation, anthropomorphic grasp, and mechanism compact
size as a specific solution for an improvement of the LARM hand. The LARM Hand and its finger
mechanisms are described in ref. [22], and problems for adaptable grasp have been discussed in ref.
[14]. A new underactuated kinematical solution is proposed by using an analysis of the LARM Hand
finger structure and underactuated finger mechanisms that were previously developed at LARM. A
mechanical design has been elaborated with numerical simulations in ADAMS environment and it is
presented to characterize the functionality of the new prototype.

2. LARM Hand
The last version of LARM Hand is reported in Fig. 1(a), with three 1-DOF fingers, a palm and a
standard flange for connection with robots.22 The size of this prototype is 1.2 times larger than an
average human hand. The phalanx bodies are sized with height of 2 cm, length of 6 cm, 3.7 cm, and
3.5 cm for phalanx 1, 2, and 3, respectively.10,22 The actuation system consists of three DC motors

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Table I. The structural parameters of the finger mechanism of the LARM Hand in Fig. 1.
[mm]
l21

l22

l3

l51

l52

l6

l8

24.1

3.9

28.5

19.9

25

6.9

83.5

51

129

l1
8.8

[deg]

Fig. 1. The LARM Hand IV: (a) the prototype built on 2007; (b) a scheme of finger mechanism.

with a reduction gear train on each axis. These DC motors are attached to the base frame. DC motors
are controlled independently in close loop by means of a PLC. The control architecture is able to
operate the 3 DC motors and to manage at least 3 force sensors (one for each finger) with easyoperation user-friendly features. Piezoresistive force sensors are installed on the fingers of the LARM
Hand to monitor the grasping force that is exerted with about 15 N. Piezoresistive effect consists
in a resistance variation of a suitable material when a contact force is applied on it. Therefore, it is
necessary to use a proper conditioning board with the aim to obtain a tension output from a resistance
variation. The LARM Hand IV can be easily equipped with more or less than three fingers just by
adding or removing one independent module that is made of a finger with its own DC motor.
By analyzing several videos of the human cylindrical grasping of a given object it has been observed
that there are almost constant ratios between the motions of phalanxes during the approaching motion
to an object. Therefore, this motion can be reproduced with a suitable 1-DOF finger mechanism.
However, it is necessary to design a driving mechanism with suitable link lengths and transmission
ratios between phalanxes to produce a human-like cylindrical grasping. The driving mechanism has
been designed to remain completely embedded in the finger body during the whole movement of
the finger,6,24 in order to achieve an anthropomorphic shape of the finger, Fig. 1(a). The abovementioned design considerations have been taken into account for designing a 1-DOF human-like
finger mechanism for LARM Hand, whose scheme is shown in Fig. 1(b).6,24 Each finger is composed
of two cross four-bar linkage mechanisms. The phalanx 1 is the input bar of the first four-bar linkage
mechanism. The phalanx 1 is also the base frame of the second four-bar linkage mechanism. The
phalanx 2 is the input bar of the second four-bar linkage mechanism and it is also the coupler of
the first four-bar linkage mechanism. Then, the phalanx 3 is the coupler of the second four-linkage
mechanism. The LARM Hand design is characterized by link sizes that are reported in Table I. But
this kinematic solution is limited in adaptability to object shapes. In order to improve the capability
of grasping objects with different size and shape, solutions with underactuated mechanism have to
be considered.23
In order to characterize the LARM Hand behavior and to evaluate its grasping forces, experimental
tests were carried out at LARM, as reported in ref. [25], by grasping objects with different shape and
material. Experimental results have been compared with human grasping forces that are obtained in
experimental tests with human hand as described in ref. [24].
The paper describes solution with an underactuated linkage mechanism that preserves the main
feature of the original design of LARM finger. The feasibility of the proposed underactuated design is
characterized by simulations results that will be also used in future work for a prototype construction
and its testing.

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Fig. 2. Examples of final grasp configurations with original LARM finger: (a) all phalanxes in contact with the
object; (b) only phalanxes 1 and 3 in contact with the object.

Fig. 3. Examples of final grasp configurations with finger driving mechanisms: (a) articulated mechanism; (b)
underactuated mechanism.

3. Problems for Adaptable Grasps


The articulated finger mechanism in LARM Hand IV is composed of rigid links so that the phalanxes
have predetermined configurations during finger motion. In this way, the finger mechanism has no
possibility to adapt its configuration to the shape of grasped objects, with the exception of those
sizes and shapes that were considered during the design process for a predetermined finger grasping
configuration. However, once the finger mechanism is properly sized as function of the object size
and shape, it can conveniently wrap a prescribed family of objects by achieving grasp contacts at
each phalanx, like in the example in Fig. 2(a). Objects that are different from those considered during
the design process could be grasped with no all phalanxes in contact like in the example of Fig. 2(b),
where only phalanxes 1 and 3 are in contact.
In order to improve the capability of grasping objects with different size and shape, a solution
with underactuated mechanism can be used with conditions and peculiarities that are discussed in
previous sections. By using underactuation in the driving mechanisms phalanxes can adapt their
configurations to object shape according to kinematic constrains. Figure 3 shows an example of how
an incomplete grasp configuration of an articulated mechanism in Fig. 3(a) can be adjusted to a
wrapping configuration in Fig. 3(b) when the finger is driven by an underactuated mechanism.
The area that a finger can sweep with its phalanxes can be identified as the reachable space.
An articulated mechanism has less grasp configurations with all phalanxes in contact than an
underactuated mechanism, because a finger with underactuated DOFs can grasp any object that
is within its reachable space, independently of its shape and size, as pointed out in ref. [16].
Self-adaptive fingers can envelope objects to be grasped by adapting mechanism configuration to
their shape with only one actuator and without complex control strategies. While a finger closes against
an object, the configuration of the finger is determined by external constraints that are associated with
the object. When the object is fully grasped, the force that is generated by the actuator is distributed
among all phalanges in contact with it, as pointed out in ref. [16, 26]. As an example of underactuated
grasping a closing sequence of an underactuated 2-DOFs finger is shown in Fig. 4.16 The finger is
actuated through the smaller link, as shown by the arrow in the figure. Since there are 2-DOFs and
1 actuator, 1 elastic element must be used. Thus, a spring is used between two links to maintain the
finger fully extended. In Fig. 4(a), the finger is in its initial configuration and no external forces are

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Fig. 4. An example of a closing sequence of an underactuated 2-DOFs finger.

Fig. 5. A typical sequence of human-like grasping.

present. The finger mechanism behaves as a four-bar linkage. In Fig. 4(b), the first phalanx reaches
contact with the object. In Fig. 4(c), the second phalanx moves with respect to the first one and the
finger closes against the object since the first phalanx is constrained by the object. During this phase,
the actuator produces the force that is required to stress the spring. Finally, in Fig. 4(d) both phalanxes
are in contact with the object and the finger has completed the shape adaptation. The actuator force is
distributed among the two phalanxes in contact with the object. It should be noted that the sequence
occurs with a continuous motion of the actuator as related to continuous deformation of the spring as
pointed out in ref. [16].
Thus, underactuation is useful in anthropomorphic fingers because with underactuated mechanisms
it is possible to achieve an adaptive grasp that mimics human grasping with a proper sequence like
that the one shown in Fig. 5. Phalanxes 1, 2, and 3 will touch the object, one after the other, while the
underactuated mechanism will adjust its configuration to the object size and shape in 4 phases:
(1)
(2)
(3)
(4)

The finger moves freely.


Phalanx 1 reaches contact with the object and phalanxes 2 and 3 move freely.
Phalanx 2 reaches contact with the object and phalanx 3 continues to move towards the object.
Phalanx 3 reaches contact with the object.

The finger reaches a final grasp configuration with a full wrap of the object with all the phalanxes
in contact. Due to different shape or position of objects, sometimes a distal phalanx will be in contact
with the object before the proximal phalanx. Thus, the phase (b) or (c) in Fig. 5 can differ in the
grasping sequence, but the final full grasping configuration will still be achieved with a full wrap.
The only difference is that the final configuration can have less contact points, as commented in
ref. [10]. However, it is possible to reach singular positions, where the grasp is not stable and the
finger could lose contact with object, as explained in ref. [26].
In order to obtain an underactuated finger for LARM Hand, it is possible to slightly modify the
original linkage mechanism shown in Fig. 1(b). The goal is to preserve the main features of LARM
finger design with a 1-DOF mechanism structure, anthropomorphic grasping behavior, and a size that
permits a mechanical design within the finger body during all grasping phases.

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Fig. 6. Solutions for underactuated finger mechanisms,14 : (a) a general kinematic scheme; (b) practical solutions
for underactuation by using linear and torsional springs (l.s. and t.s.).

Fig. 7. A new solution for LARM finger mechanisms with its design parameters.

Figure 6 shows feasible solutions with a conceptual general scheme as proposed in ref. [14] in
which combinations of flexible elements and springs in the links can be considered for design of
underactuated finger mechanisms.

4. A New Underactuated Finger Mechanism


In order to design a new 1-DOF underactuated finger for LARM Hand, new mechanisms have
been experienced by using considerations that are summarized in Fig. 6. In order to obtain a new
underactuated finger mechanism for LARM Hand a mechanism search has been worked out to identify
several possible solutions.25 A feasible solution for a new mechanism has been selected as shown in
Fig. 7 with linear springs that make underactuated the rotation joints in C and G. The mechanism is
composed by 8 links, 9 revolute joints, and 2 springs. Phalanxes are links 2, 6, and 8. This mechanism
has a fairly limited manufacturing complexity because of the reduced number of bodies in the linkage
structure and it has an efficient operation since it ensures proper distribution of contact forces among
the finger phalanxes.
Because of underactuation this mechanism is able to grasp objects with different shape remaining
within the finger body during its movement. The closure sequence of the finger motion while grasping
a cylinder is reported in Fig. 8 as a characteristic operation example. Linear springs are properly
modeled in ADAMS environment for the underactuated rotation joints of the scheme in Fig. 7. The
characteristic operation of the finger mechanism can be described with typical situations for specific
contacts through suitable virtual equivalent mechanisms.10 Equivalent mechanisms can be identified
as in Fig. 9 for the cases with no phalanxes in contact, only phalanx 1 in contact with an object, and
phalanxes 1 and 2 in contact. Referring to the first situation, a phalanx is free when there is no contact
force and torque acts to move it. Generally a phalanx is free before it will touch an object. In this
case, the two links that are connected by a spring can be considered as a single virtual link. Here,
links 3 and 4 can be considered as one virtual link 9 as shown by dashed line segment BD in Fig. 9(a).
Link 6 and 7 can be considered as virtual link 10 with segment FI. Therefore, the proposed finger

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Fig. 8. A closure sequence of cylinder grasping by the new finger mechanism in Fig. 7 as simulated in ADAMS
environment.

Fig. 9. Equivalent mechanisms for the finger mechanism in Fig. 7: (a) for free phalanxes condition; (b) for
phalanx 1 in stop condition.

mechanism can be considered as the equivalent mechanism in Fig. 9(a), which recalls the original
four-bar linkage in LARM Hand in Fig. 1(b).
In the second situation, phalanx 1 is stopped while phalanxes 2 and 3 are free. In this case, link
2 and joints E and F are fixed and they act as a virtual frame as shown in Fig. 9(b). Spring 1 will
start to be deformed because of motor push. But spring 2 will be not activated because phalanxes 2
and 3 are still free to move. Links 6 and 7 can still be considered as one single virtual link 10 with
segment FI. Thus, the finger mechanism can be considered as the equivalent mechanism in Fig. 9(b).
When phalanx 2 is stopped because in contact with object, link 6 and joints E, F, and G are fixed,
and phalanxes 1 and 2 act as a virtual frame. Thus, also spring 2 will start to be deformed to move
phalanx 3 towards a contact against the object.
A kinematic characterization of the new finger mechanism has been worked out through simulations
in ADAMS environment with the aim to test the operation feasibility and to characterize the grasping
performance of the proposed underactuated solution. Considering the mechanical design and finger

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand


Table II. Main parameters referring to kinematic solution in Fig. 7.
[mm]
l1
5.4

[deg]

l21

l22

l3

l4

l51

l52

l6

l7

l8

50.8

17

51.3

5.4

46.5

34.3

12.3

14.5

83.5

87.7

128.6

Table III. Friction values and spring parameters for ADAMS model in Fig. 11.
Friction

Static
Dynamic

Springs

Joints

Object-phalanx

0.1
0.01

2
1.1

Stiffness [N/mm]
Damping [N s/mm]

0.2
1

Fig. 10. Finger structure with phalanxes design dimensions and embedded mechanism.

dimensions in Fig. 10, main parameters have been determined for the kinematic scheme in Fig. 7 as
reported in Table II.
By using the model of Fig. 10 closure-equations of links can be expressed through polygons
BCDEA and GFEHI as
l1 ei1 + l21 ei2 = l3 ei3 + l4 ei(3 ++m1 ) + l51 ei5 .
l52 e

i(5 +5 )

+ l8 e

i8

= l22 e

i(2 +2 )

+ l6 e

i6

+ l7 e

i(6 +m2 )

(1)
.

(2)

They can be used to obtain input-output motion relationships for motion control of the finger
mechanism.
The new underactuated finger mechanism has been tested by using ADAMS View dynamic
simulation software in simulating grasps with different objects with the aim of characterizing the
new design also in term of exerted actions. By using values in Table II a CAD model has been built
in Solidworks as shown in Fig. 11. In ADAMS model friction values and characteristics have been
defined as reported in Table III.
The input link has been actuated by using the solution of Eqs. (1) and (2) with an angular velocity
given by
= Kv sin(time),

(3)

where time is the clock variable and Kv is the velocity factor that can be specifically adjusted
for each grasped object for a proper smooth motion. This input motion has been implemented in
numerical simulations with ADAMS model and results are reported in the next section to validate
and characterize the proposed new finger design.
After the functional characterization of the new finger design through numerical simulations, a
first prototype has been built as shown in Fig. 12. The prototype is actuated by a crank, and it has
been useful to validate the mechanism behavior through first experimental tests of operation practical
feasibility. Fig. 12 reports a closing sequence that is the same of the simulated one in Fig. 8.

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

Table IV. Material properties for the grasped objects in the simulated tests in Fig. 8.

Aluminum
Rubber

Young modulus [N/mm2 ]

Density [kg/mm3 ]

Poisson ratio

7.1 104
2.8 103

2.7 106
1.7 106

0.33
0.4

Fig. 11. An ADAMS model for the new finger design.

Fig. 12. An experimental test with a built prototype performing the closure sequence in Fig. 8.

5. Numerical Performance Evaluation


A numerical evaluation has been worked out by a dynamic simulation in ADAMS environment
by running the designed underactuated mechanism with design features and anthropomorphic
closing motion as previously presented. Referring to Fig. 13, simulations are computed to
determine contact force values Fc0 , Fc1 , Fc2 , Fc3 , actuation torque at joint B, joints reactions
Ra , Rb , Rc , Rd , Re , Rf , Rg , Rh , Ri , phalanxes contact point accelerations af1 , af2 , af3 , and springs
deformations from initial lengths 1 and 2 . The objects that have been considered for
simulations are made by aluminum (rigid objects) and rubber (deformable objects), with mechanical
properties that are reported in Table IV.
Five simulations in ADAMS environment have been worked out in grasping objects of different
shape as shown in Table V with closure sequence of 1 s duration. All the simulated cases have shown

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

10

Design and simulation of an underactuated finger mechanism for LARM Hand


Table V. Characteristics of the objects used in numerical ADAMS simulations.

Fig. 13. A finger scheme with variables for numerical simulations in ADAMS environment.

Fig. 14. Computed simulation of the finger closing sequence with the object for the case 3 of Table II: (a) initial
configuration; (b) 0.01 s; (c) 0.25 s; (d) 0.59 s; (e) 0.69 s; (f) 0.78 s.

the driving underactuated mechanism moving smoothly while remaining within the finger body. An
example is shown in Fig. 14 with a sequence of configurations that are evaluated during the ADAMS
simulation of the finger motion.
The computed results from the simulation are summarized in the plots of Figs. 1520 and Tables
VIVIII. In particular, the case 3 of Table V is considered in details and Fig. 14 reports the grasp
closing sequence during which all the phalanxes are in contact with the object surface at the end of
the grasping sequence.
Figure 15 shows the computed results for the contact forces at palm and phalanx points in which it
is possible to identify values corresponding configurations (b), (d), (f) in Fig. 14, when phalanxes 1,

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

11

Table VI. Values of joint reaction forces at static equilibrium for the cases of study in Table V.

Phalanx 1
Phalanx 2
Phalanx 3
Palm

Object 1 [N]

Object 2 [N]

Object 3 [N]

Object 4 [N]

Object 5 [N]

59.23
6.12
10.21
41.54

9.83
8.24
4.95
14.62

68.54
5.01
15.98
32.69

11.12
1.72
4.37
6.23

10.45
8.51
4.64
14.52

Table VII. Maximum torque as computed during simulations for grasping tests of Table V.

max [Nmm]

Object 1

Object 2

Object 3

Object 4

Object 5

5,612.58

4,706.26

7,218.59

6,750.86

10,256.46

Table VIII. Computed maximum spring deformations as function of the objects in the grasping tests of Table V.

1max [mm]
2max [mm]

Object 1

Object 2

Object 3

Object 4

Object 5

1.74
4.45

1.87
3.07

2.59
5.71

2.36
5.92

1.69
4.53

Fig. 15. Simulation results for the grasping test in Fig. 14 in terms of contact forces on the phalanxes and palm.

2, and 3 reach contact, respectively. At (b) configuration forces are indicated in the zoomed view with
values of about 35 N on phalanx 1 and about 11 N on palm. At (d) configuration the contact forces
on palm and phalanx 2 reach about 72 N. After 0.95 s from the beginning of the simulated closure
the final grasp configuration is reached and stationary forces values are computed as shown in Fig. 15
with maximum value of about 70 N for the phalanx 1. It is to note that the contact forces on phalanxes
and palm show a sudden increase at the moment of each contact and successively they decrease as
related to the smooth adjustment of the finger configuration through the action of the underactuation
of the springs. This behavior can be observed also in the results of the other characteristics in
Figs. 16 and 18.
In Fig. 16, the computed input torque is reported with a maximum value of 7.22 N m. The torque
evolves similarly to the contact forces since they are related to each other as per force transmission
capability of the finger mechanism. The force redistribution makes the torque decreasing up to the
next jump that is required by the action on the spring because of an additional contact of phalanxes.
In Fig. 17, the fingertip accelerations are reported while grasping the object and maximum values
are computed with peaks that correspond to the contact of phalanxes with the object. These values of
accelerations of phalanx contact points are computed of about 9.0 m/s2 for contacts at the phalanx 1
and 2 and about 50.00 m/s2 at the phalanx 3 contact. The peak values are experienced in all phalanxes,
but only the phalanx 3, which is the extremity of the finger, is the most sensitive with large values
of acceleration up to the maximum of about 50.00 m/s2 . Actually the peaks indicate impacts of a

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

12

Design and simulation of an underactuated finger mechanism for LARM Hand

Fig. 16. Simulation results for the grasping test in Fig. 14 in terms of input torque.

Fig. 17. Simulation results for the grasping test in Fig. 14 in terms of phalanxes accelerations.

Fig. 18. Simulation results for the grasping test in Fig. 14 in terms of joints reaction forces.

phalanx with the object at the first contact instant and null or very small values are computed in the
remaining time of the finger motion as an indication of the smooth movement of the finger towards
the final grasp configuration.
The computed results for reaction forces are reported for each joint in Fig. 18 as characteristics
that are helpful to understand and appreciate the stress of the finger structure during the motion and
grasping action. At the static equilibrium the maximum reaction force is computed of about 540 N
at E joint. It is remarkable how the reaction forces at the joints show a similar evolution with values
near to each other as an indication of the distribution and share of the finger stress during the motion

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

13

Fig. 19. Simulation results for the grasping test in Fig. 14 in terms of springs deformations from initial length.

Fig. 20. Computed torque for cases of study in Table V.

and action of the grasping. The largest values are computed for the joints of the driving linkage that
is the part of the finger dedicated to the force transmission from the actuator.
In Fig. 19, the deformation history is reported for the two linear springs in the finger mechanism.
Spring 1 starts to be deformed when the phalanx 1 is in contact, while spring 2 is deformed when the
phalanx 2 reaches contact. Then, both the springs reach their maximum compression when also the
phalanx 3 is in contact with the object and the final grasp configuration is achieved. While spring 1
shows a smooth deformation evolution with small values, spring 2 suffers a larger deformation even
with a sudden change that is due to the severe spring conditions of the final closing configuration of
the finger mechanism.
The values of grasping forces from simulations with different objects are listed in Table VI as a
summary of the numerical values at static equilibrium for the cases of study in Table V. The large
differences in each phalanx as function of the grasped object can be understood as a measure of
the efficiency of the underactuated mechanism in adjusting the grasp in different configurations and
objects. It is to note that the maximum values refer to the first phalanx and palm that are indeed the
part of the finger with strongest contact with a grasped object.
In Fig. 20, plots of input torque are summarized as computed during simulations of grasp closing
sequence for the five cases of study in Table V. Common behaviors can be appreciated from the
computed peaks of the contacts between phalanx and object that are related to the activation of the
underactuation springs, as previously commented. The final value of the torque varies as depending
of the size and shape of the grasped object influencing the final finger configuration. Underactuation
is experienced in all the cases with a torque evolution whose typical time history is shown in more
details in Fig. 16. Only for the case with the parallelepiped object the two characteristic peaks of the
spring activation are quite similar in values and evolution. Table VII summarizes maximum torque
values that are computed during simulations and a maximum value is determined as 10.26 N mm.

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

14

Design and simulation of an underactuated finger mechanism for LARM Hand


Table IX. Structural parameters for mechanisms in Fig. 1(b) and in Fig. 7.
[mm]

LARM Hand
New finger

[deg]

l1

l21

l22

l3

l4

l51

l52

l6

l7

l8

8.8
5.4

24.1
50.8

3.9
6

28.5
17

51.3

6
5.4

19.9
46.5

25
34.3

12.3

6.9
14.5

83.5
83.5

51
87.7

129
128.6

The reported simulation results can be useful also to choose suitable commercial springs. In fact,
the data in Table VIII with maximum deformations of springs for the five objects give the maximum
value of 5.92 mm for spring 2. It is to note that the deformation range of both the springs is limited
within a reasonable range that shows a practical feasibility of the spring underactuation.
As a summary of the design characterization, Figs. 1(b) and 7 can be used to compare the structure
of the original finger mechanism in the LARM Hand with the solution of the new underactuated
finger mechanism. Table IX lists structural parameters characterizing the two mechanisms with the
aim of facilitating a comparison in terms of dimensional sizes of the LARM finger design. It is to note
that l3 and l6 of the original finger in Fig. 1(b) correspond to the virtual links BD and FI, respectively,
as outlined in Fig. 9. The overall size of the original finger is preserved in the new underactuated
solution. The underactuated solution of Fig. 7 can be considered as properly sized by looking at
the link ratios that permit to appreciate the similarities in the kinematic designs that are listed in
Table IX. In addition, the reported simulations show that the characteristic operation of LARM
fingers is preserved also in the new design with an anthropomorphic closing to grasping configuration
and mechanical efficiency due to suitable contact forces in all the phalanxes.

6. Conclusions
An underactuated finger mechanism is presented with a mechanical design that is an evolution
of the linkage design in the original LARM finger. The kinematic design is characterized by the
underactuation that is obtained by linear springs as passive actuators between two connected links.
The new design is characterized to be underactuated with the capability to have finger configuration
adaptable to object shape and size while preserving the characteristic functioning and size of the
original design. The underactuated mechanism has been simulated in ADAMS environment to check
its mobility within the finger body and to operate grasping with efficient performance. The smooth
human-like operation has been tested through numerical simulations in ADAMS environment whose
results show satisfactorily capabilities in a characterization of a feasible implementation as shown
also in first tests of a prototype.

References
1. G. Carbone, Grasping in Robotics, (Springer, Dordrecht, 2013).
2. M. Ceccarelli, Fundamentals of Mechanics of Robotic Manipulation, (Kluwer Publ., Dordrecht, 2004).
3. B. Siciliano and O. Kathib, Springer Handbook of Robotics, Part D, Chapter 28, (Springer, Heidelberg,
2008).
4. I. Staretu, Gripping Systems, (Derc Publishing House, Tewksbury, 2011).
5. A. Bicchi, Hands for dexterous manipulation and robust grasping: A difficult road toward simplicity,
IEEE Trans. Robot. Autom. 16, 652662 (2000).
6. M. Ceccarelli, N. E. Rodriguez and G. Carbone, Optimal design of driving mechanism in a 1-DOF
anthropomorphic finger, Mech. Mach. Theory 41(8), 897911 (2005).
7. C. M. Gosselin and T. Lalibert`e, Underactuated mechanical finger with return actuation, US patent 5762390,
June (1998).
8. M. C. Carrozza, G. Cappiello, G. Stellin, F. Zaccone, F. Vecchi, S. Micera and P. Dario, A Cosmetic
Prosthetic Hand with Tendon Driven Under-Actuated Mechanism and Compliant Joints: Ongoing Research
and Preliminary Results, Proceedings of the 2005 IEEE International Conference on Robotics and
Automation (ICRA), Barcelona (Apr. 2005) pp. 26612666.
9. W. Zhang, D. Zhao, Q. Chen and D. Du, Linkage under-actuated humanoid robotic hand with control of
grasping force, Inform. Control, Autom. Robot. (CAR) 2, 417420 (2010).
10. L. Wu and M. Ceccarelli, A numerical simulation for design and operation of an underactuated finger
mechanism for LARM hand, Int. J. Mech. Based Des. Struct. Mach. 37, 86112 (2009).

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

Design and simulation of an underactuated finger mechanism for LARM Hand

15

11. S. Yao, M. Ceccarelli, G. Carbone, Q. Zhan and Z. Lu, Analysis and optimal design of an underactuated
finger mechanism for LARM hand, Frontiers Mech. Eng. 6, 332343 (2011).
12. J. S. Dai, D. L. Wang and L. Cui, Orientation and workspace analysis of the multifingered metamorphic
hand-metahand, IEEE Trans. Robot. 25(4), 942947 (2009).
13. G. Wei, J. S. Dai, S. Wang and H. Luo, Kinematic analysis and prototype of a metamorphic
anthropomorphic hand with a reconfigurable palm, Int. J. Humanoid Robot. 8(3), 459479 (2011).
14. M. Ceccarelli, C. Tavolieri and Z. Lu, Design Considerations for Underactuated Grasp with a one D.O.F.
Anthropomorphic Finger Mechanism, International Conference on Intelligent Robots and Systems IROS
2006, Beijing (Oct. 2006) pp. 16111616.
15. L. Howell, Compliant Mechanisms, (Wiley-Interscience, New York, 2001).
16. T. Lalibert`e, L. Birglen and C. Gosselin, Underactuation in robotic grasping hands, Mach. Robot. Control
4, 111 (2002).
17. J. D. Crisman, C. Kanojia and I. Zeid I, Robot arm end-effectors, US patent 5570920 (1996).
18. J. F. Mullen, Mechanical Hand, US patent 3694021, September (1972).
19. S. Bhatti, Design of a multiple finger prosthetic hand with a passive adaptive grasp system, EE 4BI6
Electrical Engineering Biomedical Capstones, Paper 28, Hamilton, Canada (2010).
20. R. Cabas, L. M. Cabas and C. Balaguer, Optimized Design of the Underactuated Robotic Hand,
Proceedings of the 2006 IEEE International Conference on Robotics and Automation (ICRA), Orlando,
Florida (May 2006) pp. 982987.
21. J. Zhao, L. Jiang, S. Shi, H. Cai, H. Liu and G. Hirzinger, A Five-Fingered Underactuated Prosthetic
Hand System, Proceedings of the 2006 IEEE International Conference on Mechatronics and Automation,
Luoyang (Jun. 2006) pp. 14531458.
22. G. Carbone and M. Ceccarelli, Design of LARM hand: Problems and solutions, J. Control Eng. Appl.
Inform. 10(2), 3946 (2008).
23. N. Dechev, W. L. Cleghorn and S. Naumann, Multiple finger, passive adaptive grasp prosthetic hand,
Mechanism Mach. Theory 36, 11571173 (2001).
24. N. E. Rodriguez, M. Ceccarelli and G. Carbone, Design and tests of a three finger hand with 1-DOF
articulated fingers, Robotica 24, 183196 (2006).
25. M. Zottola, Masters Thesis Design of an underactuated LARM finger mechanism for robotic hand,
University of Cassino and South Latium, Cassino (2013).
26. L. Birglen and C. M. Gosselin, On the Force Capability of Underactuated Fingers, Proceedings of the
2003 IEEE International Conference on Robotics and Automation (ICRA), Taipei, Vol. 1, (Sep. 2003)
pp. 11391145.

http://journals.cambridge.org

Downloaded: 06 Nov 2015

IP address: 134.58.253.30

You might also like