Professional Documents
Culture Documents
Ubh 3 A Biologically Inspired Robotic Ha
Ubh 3 A Biologically Inspired Robotic Ha
Figure 5: One-dof hinge (a) and two-dof hinge (b) Figure 6: Setup for finger structure characterization
based on compliant mechanism. and position sensors calibration.
desired deformations of hinges, friction inside the ten- the curvature of the hinge (because of reflections), and
dons’ sheaths, etc.) and enhance the capability of the the sensor output is “proportional” to the bending an-
anthropomorphic hand. A set of position sensors lo- gle. The signal coming from the sensor is reported in
cated into the joints must be add to the systems, even Fig. 8, where it is compared with the one produced by
if, in principle, the motors’ sensors could be enough to the Hall-effect sensor. The characteristic of the sensor
perform some kind of control (e.g. impedance control) is again strongly non-linear, but in this case the max-
and some operations (grasping). Moreover, to perform imal sensitivity is achieved for small angle values (the
fine manipulation tasks, which need to exactly know light comes directly from the emitter to the receiver),
the magnitude of the exchanged forces and the exact where most part of the manipulation operations hap-
position of contact points, force and tactile sensors pens. By properly adjusting some parameters of the
must be inserted. sensor (this activity is currently ongoing), i.e. by given
an initial bending which prevent a direct view of the
4.1.1 Position sensors two elements of the sensor or by changing the distance
between them, it is possible to change such a char-
In order to estimate the relative positions between the acteristic and enhance the sensitivity for bigger angle
links of the finger, a measure based only on tendon values. It is worth to notice that the physical principle
lengths is not sufficient. On the other hand, the pe- adopted makes this sensor insensitive to external ele-
culiar structure of the joints (without a fixed rotation ment (the soft-cover can easily shield the sensor from
center) makes it difficult to find a suitable position the light of the environment).
sensor. In fact, the standard robotic technologies (e.g. If an high resolution, in the overall range of variation
potentiometers) require well-defined paths. In order to of the joint angle, is required, the two last sensors can
obtain a sensor, reliable (without sliding parts) and be used together: as shown in Fig. 8 their outputs are
enough repeatable, different physical principia have perfectly complementary and if properly fused they
been considered. lead to a high precision estimation in every configura-
A simple solution, reported in Fig. 7.(a), is based on tions.
the Hall effect: a linear Hall-effect sensor is located In the 2 d.o.f. joint shown in Fig. 5.(b), since the elas-
in one side of the joint and a magnet is faced in the tic hinge can move in the 3D space and not only in a
other side. While the joint is closing the distance be- plane, the sensing principles above mentioned can not
tween these elements decreases and accordingly the be exploited, and at least two information are neces-
magnitude of the magnetic field, detected by the sen- sary in order to estimate the relative position between
sor, increases. The response of the sensor, shown in the two links. In this case, the obvious solution consist
Fig. 8, is strongly non-linear, and, in particular, its of the measure of the tendon displacements. Such a
sensitivity is quite reduced when the joint is “opened” measure has to be performed as close as possible to the
(0o ). Moreover, in this configuration the effect of ex- joint, therefore a miniaturized position sensors (which
ternal magnetic or ferromagnetic elements can be very must be located into the finger base) has been devised,
relevant. In order to achieve the advantage of the hall- see Fig. 9. It is composed by a magnet (or a number
effect sensor (non-contact sensing) without the same of magnets, according to the desired resolution) fixed
drawbacks (in particular, the dependance of the out- to the tendon. When it moves the magnetic field, de-
put on an external magnetic field), an optical sensor tected by a Hall-effect sensor, changes and a position
has been developed. This device, shown in Fig. 7.(a), measure is therefore achieved.
exploits one of the elastic sheaths of the finger: a pho- A common drawback of the sensing technologies
todiode, emitting red light, is located in one end of the above introduced is the difficult procedure of calibra-
sheath while the other one hosts a photo-transistor,
which detects the amount of light received. When the
1
joint closes, the intensity of the light is modulated by light based sensor
hall−effect based sensor
0.9
0.8
0.7
Light emitter
sensor output (V/Vmax)
0.6
Elastic hinge
0.5
Light receiver
0.4
Elastic hinge
Magnet 0.3
Hall-effect
sensor 0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
(a) (b) angle (deg.)
Magnetic field
(a)
Tendon
(a) (b)
Magnetic field
Figure 9: Working principle of the tendon position location of the contact points. To this purpose, tac-
sensors based on a hall-effect sensor: one magnet (a) tile array sensors are profitably used [11].
and multiple magnets(b). To make the sensory equipment consistent with the
overall project (inspired by criteria of simplicity and
tion. In order to simplify this process, the setup of reliability) we have designed a sensor able to detect
Fig. 6, is used. It allows to compare the joint angle the force provide by the tendons to the joints.
values, estimated by means of the images coming from Fig. 10.(a) shows the load cell, composed by a
the video camera, with the sensors’ output. deformable structure monitored by classical strain
Alternatively, in order to keep the finger structure gauges. It has been integrated into the connection be-
(and hence the overall end-effector) as simple as pos- tween the tendon and the actuated phalange (see the
sible, a control based on visual servoing (and on the cad model in Fig. 10.(b)), allowing a direct measure
measures of tendons’ elongation given by sensors on of the forces exchanged with the environment, with-
the motors) can be implemented [10] (this possibility out the error due to the friction between the sheaths
is currently evaluated, by means of the same setup and the tendons. The output of the load cell is very
used to make the calibration). linear and does not dependent on the joint angle, as
reported in Fig. 11
Furthermore, the adoption of a soft layer of visco-
elastic gel is particularly suitable for an integration
4.1.2 Force sensors of distributed tactile sensing capabilities. Some ex-
The actuation chain based on tendons produces non- perimental activities have been performed in order to
negligible levels of static and viscous friction. In or- test the interactions between the gel and a tactile sen-
der to perform fine manipulation tasks and apply very sor (the use of a commercial single element pressure
small forces on the environment, force and tactile sen- sensors has been investigated [12]). The presence of
sors are, therefore, necessary. Such sensors can be
used for two different purposes, which require specific
characteristics: Load cell response
• force control;
6
5.5
• task planning. 5
Sensor output (V)
4.5
3.5
tion control, requires sensors with a very high band- 3
60
while more complicate sensors often adopted in dex- 40
20
20
40
terous manipulation are multi-axis force sensors and Tendon tension (N)
0 0
Joint angle (Degrees)
[12] www.tekscan.com/flexiforce/flexiforce.html.
References [13] C. Ching-Ping, B. Hannaford, “Static and dy-
[1] C.S. Lovchik and M.A. Diftler, “The Robonaut namic characteristics of McKibben pneumatic ar-
Hand: a Dexterous Robot Hand for Space”, in tificial muscles”, in Proc. IEEE Int. Conf. on
Proc. IEEE Int. Conf. on Robotics and Automa- Robotics and Automation ’94, ICRA 94, May
tion ’99, ICRA 99, Michigan, May 1999. 1994.
[2] J. Butterfass, M. Grebenstein, H. Liu, G. [14] N. Hogan, “Impedance Control: An approach
Hirzinger, “DLR-Hand II: Next Generation of to manipulation, Parts I-III”, in ASME Journal
a Dextrous Robot Hand”, in Proc. IEEE Int. of Dynamic Systems, Measurement and Control,
Conf. on Robotics and Automation ’01, ICRA 01, Vol. 107, pp. 1-24, 1985.
Seoul, Korea, May 21-26, 2001.