Detc2004 57126
Detc2004 57126
net/publication/37537780
CITATIONS READS
38 1,381
2 authors, including:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Richard Michael Crowder on 26 August 2014.
DETC2004-57126
ABSTRACT INTRODUCTION
Research into robotic grasping and manipulation has led to Efforts are being made world wide to remove human
the development of a large number of tendon based end operators from hazardous environments. Such environments
effectors. Many are, however, developed as a research tool, include, space exploration, underwater mining, and operations
which are limited in application to the laboratory environment. in nuclear and chemical industries. The objective is to reduce or
The main reason being that the designs requiring a large remove the risk from the operators. In order to achieve this one
number of actuators to be controlled. Due to the space and has to rely on robotic or autonomous systems that have robust
and flexible mechanical structure to adapt to the environment’s
safety requirements, very few have been developed and
needs. However, the current level of technology does not
commissioned for industrial applications. This paper presents provide off-the-shelf solutions for every possible situation
design of a rigid link finger operated by a minimum number of because such environments are unknown and there is a degree
actuators, which may be suitable for a number of adaptive end of uncertainty involved. The existing systems can meet the
effectors. The adaptive nature built into the end effector (due to needs that are well defined and known in advance, but if the
limited number of actuators) presents considerable problems in operations are to be performed autonomously in such
grasping and control. The paper discusses the issues associated environments the unstructured and non-deterministic nature of
with such designs. The research can be applicable to any the environment precludes all this. This requires special
adaptive end effectors that are controlled by limited number of purpose system to be developed, which may put stringent
actuators and evaluates their suitability in industrial specification on the material and components of the robotic
environments. systems.
The robotic system required to operate in such
Keywords: End Effectors, Finger Design, Robotic Hand, environment should be versatile enough to handle unknown
Adaptive Finger, Robotic grasping objects and situations i.e. the gripper of the robotic system
should be able to grasp objects of different shapes and mass
more like the human hand than a conventional robotic end
effector. The design considered in this paper has evolved from
an industrial need for a tele-operated system to be used in
Tip
The finger mechanism can be considered to have two degrees
of freedom:
• Bending, where displacement of joint B bends all the three
Link 1
0.18
Fig. 4. The differential finger drive mechanism (0.006, 0.03)
0.16
As shown in the figure brakes 1 and 2 control the two lead 0.14
screws, while brake 3 controls the finger orientation relative to (0.03, 0.03)
the end effector frame. Thus the mechanism can provide three
fingertip Y, meters
0.12
FINGER CAPABILITY
CONTROL ISSUES
A prototype finger mechanism based on the differential
Even though grasping is not an exact science, some sort of
gearbox has been built to evaluate the performance of the
pre-planning is required to achieve successful grasping. To
system [13]. This is the mechanism used for controlling the
grasp an object, the end effector must be positioned and
of the fingers and not sufficiently restrained, the effect of the 0.045
first contact might simply be to move the object. There is also a
0.04
possibility that fingers can touch the object at points that are not
exactly lying on the central axis of the end effector as shown in 0.025
Fig. 6. This may result in formation of an unsymmetrical grasp
0.02
with respect to the points of contact on the grasped object. For
the unsymmetrical grasp the centriod of the triangular grasp 0.015
plane does not lie on the central axis of the end effector which 0.01
means the grasp is unbalanced with respect to finger position 0.005 bend component (d2)
(see the figure). Further due to the complex kinematic chain of
0
the finger, it has been shown that the direction of the fingertip- 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Y-location of the fingertip at a fixed X=0.07 m, in meters
normals are not co-planar at the plane of contact [17], this will
result in force and moment unbalance [18]. The fingers need to
Fig. 7. Leadscrews displacement for vertical motion
move in horizontal and vertical directions co-operatively to
minimize this unbalance and in the proposed end effector this The finger motion with respect to the base joint A can be
has to be achieved by a single motor per finger. Thus it is seen in a simple kinematic chain representation in Fig. 8.
important to see how the leadscrews displacement changes Throughout the vertical motion it has been found that the
when the finger has to track a precise path. leadscrews displacements change in such a way that there is
one and only one vertical fingertip position available for a set
finger2 of d1, d2 displacements. However, the leadscrews displacement
finger2 is seen to be taking place in opposite direction (Fig. 7). This
finger1 finger1 means that for controlling the motion by a single motor using
the differential unit, the motor has to change the direction of
rotation for each individual leadscrew in succession. Due to the
backlash in the system linear motion cannot be achieved and
finger3 the accurate position control in vertical direction is difficult.
finger3
Symmetrical Unsymmetrical
0.14
Fig. 6. Position unbalance in grasp formation
0.12
0.025
curl component (d1) implementation.
• Equilibrium of the grasped object has been simulated for tip
0.02 grasp. In the case of wrap around grasp, the object is over-
constrained by several contacts, thus only the contact forces
0.015
need to be controlled to avoid crushing.
bend component (d2)
0.01
The steps involved in grasping by the three fingered end
effector starts with fingers allowed to move in the enclosing
direction. Each finger motion is impeded as soon as it records a
0.005
-0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 contact event with the object. The fingers under this condition
X-location of the fingertip at a fixed Y=0.14 m, in meters
apply a minimum force to the object. At this situation the lift is
attempted and the control is passed to the fuzzy logic which
Fig. 9. Leadscrews displacement for horizontal motion generates appropriate motor torque for holding the object [20].
Once the fuzzy control has ensured that the grasped object does
It is important to note here that the unbalance minimization not slip, the equilibrium of the grasped object is tested by
in position can mostly be achieved by horizontal movement of conducting the position, force and moment unbalance test. If
the finger, and fortunately this does not require motion reversal the grasp is unbalanced with respect to any of these parameters,
sequence as it was in the previous case. Thus this is an the algorithm attempts to modify the unbalance as far as
interesting feature of the finger, which makes it more suitable possible within the allowed constraints. To ensure successful
for grasping application than manipulation. grasping it is necessary to ascertain that each finger touches the
object so that sensors can provide the measure of contact force.
GRASPING ISSUES The sensing and control part of the end effector have been
One of the most basic and necessary qualities of a grasp is presented in our earlier work [19, 20], here only mechanical
that the grasped object is in equilibrium. A grasp is said to be in aspect of the end effector grasp is considered for unbalance
equilibrium if the components of forces and moments acting on minimization.
the grasped object all add up to zero. In case of the object From the preceding discussions it is clear that the structure
grasped by multifingered end effectors, the equilibrium of grasp of the finger does not allow the control of the finger motion in a
requires that no resultant force acts on the object and the object truly three dimensional space. Thus force and moment balance
is fully restrained. If the resultant force is not zero, there will be can be achieved by changing the fingertip forces and contact
unbalance in force and moment, which will tend to move and positions over the grasped object. Note that the finger rotation
rotate the object. can only be used to change the grasping configurations (from
For the present end effector, unbalances in force and concentric to two fingers and opposing thumb grasp), this
moment may arise even at the equal motor torque at each finger motion cannot be used for changing the finger position over the
due to the unsymmetrical location of fingertips from the central grasped object.
axis of the end effector. This is due to the fact that the structure In order to simulate object grasping, the model allows initial
of the finger offers different contact forces at different locations position unbalance to be introduced into the system and the
for the same motor torque [17]. Also the direction of fingertip- algorithm based on least square method attempts to minimize
normals are different at different fingertip locations, which the distance between grasp centroid and the central axis of the
again results in unbalance in force and moment. Thus in order end effector. This in actual end effector will be achieved by co-
to achieve equilibrium of the grasped object, these unbalance ordinated movement of the three fingers, mostly in horizontal
need to be minimized without slipping the object and satisfying direction, so that object remains in grasp. Fig. 10 shows
the related constraints. position unbalance minimization for the case when finger 1 is
In order to minimize the system unbalance (position, force displaced by 30% inward, finger 2 by 25% outward and finger
and moment), a grasping model of the end effector has been 3 by 15% inward with respect to the corresponding symmetrical
developed which is based on the following assumptions: grasp. As seen in the figure the algorithm is able to minimize
the unbalance to the extent of 80%.
• Direction of the fingertip force is same as the normal vector
at the contact point, which can be defined from the bend of
each finger.
• Fingertips make full contact with the object during grasping
and the force at the fingertip can be provided by the
attached sensors.
• Frictional forces arising at the fingertips are not considered
for the derivation of the equilibrium equations since the
9
8 13 and 14 where unbalance in force and moment components
7 can be seen to be closing to zero at the end of the process. The
6
fingertip forces considered in the simulation are 3N, 2N and 1N
5
4 for fingers 1, 2 and 3 respectively with the position unbalance
3 as shown in Fig. 10.
2
1
0
number of i te rati ons
1 6 11 16 21 26 31 36 41 46 51
1 16 31 46 61 76 91 106
number of i terati ons
0.8
0.4
force components, N
0
Fx
-0.8 Fz
see from Fig. 11 that the torque required by the motor to drive -1.6
each fingers is also reduced. The leadscrews displacement
during this minimization is shown in Fig. 12. Non-intersecting
and identical slopes of the displacement lines suggest that for Fig. 13. Minimization of the force components
the position unbalance minimization both the leadscrews of the
finger moves in the same direction, thus direction reversal is
not required and can balance position as well as fingertip force
in one minimization process (Fig. 11). The number of iterations number of iterations
on the axis represents the incremental rate at which the 1 16 31 46 61 76 91 106
simulation is being carried out. 0.004
moment components, Nm
0.002
60
Mx
58
0 My
motor torque, mNm
56
fing1 Mz
54
fing2 -0.002
52
fing3
50 -0.004
48
46
1 6 11 16 21 26 31 36 41 46 51
0.025 drive unit within the end effector’s body with the use of solid
0.02 mechanical linkage and the transmission through the toothed
fing1_d1 belts, thus offering a strong and reliable system for use in
0.015
fing1_d2 industrial environments where safety is an important
0.01 consideration.
0.005 The control of the finger is easier with one motor and three
0
brakes as compared to individually actuated finger joints, since
1 6 11 16 21 26 31 36 41 46 51 only one motor needs to be controlled together with brakes. The
number of i te rati ons fingers can be driven in adaptive as well as precise control
mode which can be rotated about its own axis allowing it to
form either concentric or two fingers and opposing thumb
grasps. Thus the finger design offers a practical solution to the
Fig. 12. Leadscrews displacement during minimization specific tasks of grasping objects of specified shape and size
securely within the structure of the end effector. However, the