Dynamic Conditions of Reflection-Type Tactile Sensor.
ABSTRACT Considering tactile sensors there are two ways to acquire object information. These are spatial sensing with two dimensional devices and fast sensing with simple devices. Because the reflection-type tactile sensor uses a reflection image, both methods can be employed. Though there exist some dynamic characteristics in these two ways.
In this study, we first validate the hysteresis of the reflection-type tactile sensor. The results show the sensor can evaluate displacement less than 2 mm. Then we propose a novel interface called "fibratus tactile sensor." Secondly we construct a fast sensing device using reflection image and a combination of light emitting diodes (LEDs) and photodiodes (PDs), and validate the sensor's reactivity. It can distinguish 300 ms interval between two signals. Moreover the correlation between the standard deviations of the acquired outputs from the sensor and the centerline average roughness is 0.90.
Conference Paper: Sensing method of total-internal-reflection-based tactile sensor.IEEE World Haptics Conference, WHC 2011, 21-24 June 2011, Istanbul, Turkey; 01/2011
- Robot and Human Interactive Communication, 2009. RO-MAN 2009. The 18th IEEE International Symposium on; 11/2009
Dynamic Conditions of
Reflection-Type Tactile Sensor
Satoshi Saga1, Satoshi Tadokoro1, and Susumu Tachi2
6-6-01, Aoba, Aramaki-Aza, Aoba-ku Sendai, Japan
2University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan
Abstract. Considering tactile sensors there are two ways to acquire ob-
ject information. These are spatial sensing with two dimensional devices
and fast sensing with simple devices. Because the reflection-type tactile
sensor uses a reflection image, both methods can be employed. Though
there exist some dynamic characteristics in these two ways.
In this study, we first validate the hysteresis of the reflection-type
tactile sensor. The results show the sensor can evaluate displacement
less than 2 mm. Then we propose a novel interface called “fibratus tac-
tile sensor.” Secondly we construct a fast sensing device using reflection
image and a combination of light emitting diodes (LEDs) and photodi-
odes (PDs), and validate the sensor’s reactivity. It can distinguish 300
ms interval between two signals. Moreover the correlation between the
standard deviations of the acquired outputs from the sensor and the cen-
terline average roughness is 0.90.
Keywords: Force and Tactile Sensing, Visual Tracking, Virtual Reality
Considering the measuring method of tactile sensor, there are two ways to ac-
quire the object information. These are spatial sensing method and fast sensing
Conventional tactile sensors acquire the two-dimensional tactile information
by fast surface scanning due to the presence of a simple component [4,5,6,7].
These sensors use piezoelectric devices in order to acquire the surface information
from the vibration, which is caused by the roughness of the measured object.
Therefore, a sensor cannot prevent deterioration due to the stress of repeated
measurements; further, the process of amplification of the electric signal makes
M. Ferre (Ed.): EuroHaptics 2008, LNCS 5024, pp. 464–473, 2008.
c ? Springer-Verlag Berlin Heidelberg 2008
Dynamic Conditions of Reflection-Type Tactile Sensor465
Fig.1. Section of a reflection-type sensor
Other tactile sensors use image information as a typical method of scanless,
two dimensional sensing. These sensors use elastic bodies in order to utilize
the deformation information, therefore hysteresis problems occur occasionally.
In particular the measurement of small forces aggravates this problem because
the sensors use an elastic body with a small Young’s modulus. There exists no
perfect elastic body; therefore the viscosity draws hysteresis in these bodies with
small Young’s modulus.
The reflection-type tactile sensor proposed in previous study also uses two-
dimensional visual information (figure 1). This sensor amplifies the deformation
of the sensor surface by utilizing an optical lever, measures the distortion of the
reflection image and reconstructs the shape of the sensor surface. This amplifi-
cation method realizes the reconstruction of small deformations, although, the
hysteresis and repeatability may cause large errors.
On the other hand the sensor can utilize another method for sensing because
of its simplicity. The method is a high-speed and simple measurement method
by using a combination of light-emitting diodes (LEDs) and photodiodes (PDs).
In this paper, we first describe the hysteresis and repeatability, and propose a
new tactile interface; we then consider the possibility of employing a high-speed
measurement method by utilizing LEDs and PDs. The method of reconstructing
the shape from the reflection image is explained in , therefore the details are
omitted in this paper. Moreover a method of reconstructing the force distribution
from the shape distribution remains to be deteremined.
2Reconsideration of Tactile Sensor
The human receptors of cutaneous sensation can detect pressure, temperature,
pain, etc. Mechanoreceptors acquire touch, pressure and deformation informa-
tion, while thermal receptors and nociceptors acquire temperature information
and pain information, respectively. Comparing with these human receptors con-
ventional tactile sensors treat mainly touch, pressure and deformation informa-
tion which mechanoreceptors acquire.
466S. Saga, S. Tadokoro, and S. Tachi
Considering these information, the receptors measure distribution of planar
deformation or distribution of planar strain information. If the sensors measure
a similar situation, the measurement value by the sensors should be a constant.
In other words, the sensor that determines the state should have low hysteresis
and high repeatability. In our previous paper  we have already indicated that
the proposed sensor can resolve problems pertaining to wiring, resolution and
processing time. However, the hysteresis and repeatability problems should be
Additionally the human receptors can detect high frequency vibration (up to
approximately 300 Hz). Artificial sensors should also aim at these high reactivi-
ties. Our sensor can utilize another sensing method because of the simplicity. The
method is high-speed and simple measurement one by employing a combination
of LEDs and PDs. Therefore we validate how the reactivitiy of the sensor.
3 Construction and Validation of Sensing under Dynamic
3.1 Hysteresis of the Sensor
The reflection-type sensor comprises a reflective surface and an imaging device.
Our proposed tactile sensor, “RefShape,” is designed as shown in figure 2. In this
study to simplify the contact state with an object, we use the sensor as shown
in figure 3 whose contact plane with an object can be one point.
The sensor is made of addition-polymerization-type silicone rubber (KE109A,
B and KE1052A, B). The Young’s modulus of the sensor surface and the interior
of the sensor are approximately 1.6 MPa and 0.08 MPa. The thickness of the
Harder layer is 0.5 mm. The length of the salience is 50 mm, and the salience is
implated into the surface with a depth of 10 mm. This construction can confine
the deformation of the sensor surface to a small region . As a single contact
point, we use a piano wire with a diameter of 1 mm.
Fig.2. Tactile sensor “RefShape”
1 6MPa 0 5mm
Fig.3. Sensor with fibratus salience
Dynamic Conditions of Reflection-Type Tactile Sensor467
Fig.4. Experimental setup for measuring
Fig.5. Before deformation
Fig.6. After deformation
The procedure of the experiment is as follows. We attach a force sensor to
the fibratus salience of the reflection-type sensor, and slide the force sensor ap-
proximately 16 mm horizontally, and incline the salience. In each experiment we
maintain the sliding speed to be 1.6 mm/s, 4.8 mm/s, 8.0 mm/s, 11.2 mm/s,
and 16.0 mm/s. The sliding movement is reciprocation. We place the force sen-
sor to be unstick from the salience when the movement comes to one peak of
reciprocation (figure 4).
In a recursive movement (figure 5 and 6) we record the force, the position dis-
placement of the force sensor, and the position displacement of the characteristic
point acquired by the camera.
In order to validate the hysteresis and repeatability of the reflection-type sen-
sor in repetitive measurements, we first plot the correlation between the value of
the force sensor and the displacement of the characteristic point on the reflection
image; subsequently, we re-plot the correlation between the given displacement
of the force sensor position and the displacement of the characteristic point on
the reflection image (figures 7, and 8).
-0.016-0.011 -0.006 -0.001
Shift of characteristic point [pixel]
Measured force [N]
Fig.7. Correlation between measured force and measured shift of charasteristic point
468S. Saga, S. Tadokoro, and S. Tachi
Shift of characteristic point [pixel]
Movement of force sensor [mm]
Fig.8. Correlation between movement of the force sensor and measured shift of char-
Figure 7 shows that there is a strong correlation between the value of the force
sensor and the displacement of the characteristic point on the reflection image.
The sensor can evaluate forces less than 0.001 N toward the end of salience by
using sub-pixel information. Because the minimum evaluation unit of the force
sensor is 0.001 N, the hysteresis and repeatability of the sensor are not very
Figure 8 shows the correlation between the given displacement of the force
sensor position and the displacement of characteristic point on the reflection
image, it can evaluate less than 2 mm displacement of the end of salience. From
each figure, the sensor can detect similar deformation pattern without the de-
pendency of these five deformation speeds. At the same time, the displacement
of the characterisitic point is 0 pixel until the displacement of the sensor is 5 mm.
At one peak of reciprocation the force sensor is away from the salience. Therefore
the displacement of characteristic point becomes 0 pixels. This shows that the
reflection-type sensor can keep its zero point under repetitive measurement.
This result shows that the reflection-type sensor has low hysteresis (less than
5%) and similar deformation patterns for the five deformation speeds. By uti-
lizing this low hysteresis and high repeatability we have proposed a new type
sensor named “fibratus tactile sensor” [2,3]. This sensor uses not only the optical
lever based on the reflection image, but also a “physical lever,” By using these
two levers, this sensor can record the small displacement caused by the small
forces. The physical lever can be used on any object whose endpoint is stick
shaped and which has a hardness greater than that of the silicone rubber.
At the demonstration in Emerging Technology SIGGRAPH2007 , we dis-
play the sensor using feathers as the fibratus salience (figure 9). During the five
days of demonstration, more than 4,000 people have witnessed the sensitivity
and stability of the sensor. The sensitivity realize the sensor to detect the sense
of gentle touch. Because the gentle touch is intuitive operation, some people
have commented “The sensor can be used as a new interface for children, elderly
people, and disabled persons.” For example, by stroking fibers gently, one can
create dynamic flows with one’s fingers, and intuitional senses may be used as
Dynamic Conditions of Reflection-Type Tactile Sensor 469
Fig.9. Fibratus tactile sensor
inputs for computers. If softer fibers are utilized, this sensor is capable of sensing
even the blowing of wind.
3.2Reactivity of the Sensor
The reflection-type sensor comprise a reflective surface and an imaging device.
Our proposed tactile sensor, “RefShape”, is designed as shown in figure 2, and a
simplified version of the sensor is shown in figure 3. In these sensors we use a web
camera as an imaging device. However, we can combine the LEDs and PDs as
a simple component instead of camera and emply faster sensing method. In this
case, we consider the possibility of utilizing a high-speed measurement method
by LEDs and PDs (figure 10). In this study, to simplify the contact state with
an object, we use the sensor as designed in figure 11 whose contact plane with
an object can be one point. The material is the same silicone rubber that was
used for the sensor in the previous camera system. As a single contactpoint, we
use a piano wire with a diameter of 1 mm.
In order to acquire impulse response of the sensor we flick the end of the
salience. The maximum amplitude of the salience is 35 mm. Figure 12 shows the
result and approximate curve; it shows that the sensor can distinguish a 300 ms
interval between two signals under the impulse condition. Moreover if we assume
the vibration as damped vibration, we get the approximate curve (eq. 1):
y = 1.5sin(2π
From this curve the half life of the amplitude is calculated as 0.069 s. Thus, a
shorter time is required for convergence under a smaller displacement.
Next, we measure the output of the sensor by considering some objects and by
using the rotating machine (figure 14). In this study we use 9 objects (figure 13);
acryl plates(flat, smoke, fog, diamond), a cork plate, cotton waste, a rubber net,
470S. Saga, S. Tadokoro, and S. Tachi
Fig.10. High speed sensing system
Fig.11. High speed sensing system with a
-0.100.1 0.20.3 0.4
Output voltage [V]
Amp. Max 35 mm
y = 1:5sin(2ù0:05
Fig.12. Impulse response
Fig.13. (From left) Acryl(flat), Cork, Cotton
waste, Rubber net, Filter sponge, Acryls (Fog,
Fig.14. Sequential scan
and a plastic filter. Each centerline average roughness is displayed in table 1. The
roughness is measured using a laser ranging sensor (Keyence LB-02).
Dynamic Conditions of Reflection-Type Tactile Sensor471
Fig.15. Sequential response
Table 1. Centerline average roughnesses and standard deviations (Stdev) of the ma-
Acryl (flat) Cork Cotton wastel Rubber net Plastic filter
Acryl (smoke) Acryl (fog) Acryl (diamond)
Centerline average roughness
STDEV of PD response [V]
Fig.16. Correlation between centerline average roughness and standard deviation
472S. Saga, S. Tadokoro, and S. Tachi
From a comparison between the centerline averageroughness and the standard
deviations of acquired output (figure 15), the correlation between two parameters
is 0.90 (In table 1, and figure 16, the roughness of flat acryl is assumed to
As applications, by scanning the sensor it can be used an identifing tool of
textures of real environment. Additionally by engaging the previos method of
using camera, the spatial and temporal scanning ranges can be extended si-
multaneously. With these results the sensor can be used as an intuitive sensor
for interactive teddy bears, novel intuitive interfaces for computers, simple but
sensitive sensors for humanoid robots.
4Results and Discussion
This paper describes the difference between spatial sensing with two dimensional
devices and fast sensing with simple devices for the tactile sensor, and considers
the characteristics of each sensor. Furthermore we validate the dynamic charac-
teristics and problems of the proposed reflection-type tactile sensor under each
Firstly we mention the importance of hysteresis and repeatability and validate
the hysteresis of the reflection-type tactile sensor. Figure 7 shows that the sensor
can evaluate forces less than 0.001 N toward the end of the salience by using sub-
pixel information. Figure 8 shows that it can evaluate displacement less than 2
mm toward the end of salience. This result shows that the reflection-type sensor
has low hysteresis (less than 5%) in 35 mm movement condition and similar
deformation pattern for the five deformation speeds.
Next we show that the sensor can distinguish a 300 ms interval between two
signals under the impulse condition. From the approximation curve of the exper-
iment result it is observed that a shorter time is required for convergence under a
smaller displacement. Moreover the correlation between the standard deviations
of the acquired outputs and the centerline average roughnesses is 0.90 (In table
1 and figure 16, the roughness of flat acryl is assumed to be 0).
Based on these results the reflection-type tactile sensor can be used as a
high-speed tactile sensor, too. The importance of this sensor is the reflection
characteristic. In this study we validate the reactivity of one combination of sili-
cone rubbers, though, we can choose some other transparent material in case of
mismatch of the sensing range toward measured object. Moreover, in this study
we use a camera or the combination of LEDs and PDs individually, we can use
a combination of a camera, LEDs, and PDs as sensing elements simultaneously.
This sensing method enables the sensor to be precise and high reactivity mea-
threshold and determine a method for reconstructing the force distribution from
information on the displacement of the shape.
Dynamic Conditions of Reflection-Type Tactile Sensor 473
1. Saga, S., Kajimoto, H., Tachi, S.: High-resolution tactile sensor using the deforma-
tion of a reflection image. Sensor Review 27, 35–42 (2007)
2. Saga, S., Kuroki, S., Kawakami, N., Tachi, S.: Fibratus tactile sensor using reflection
on an optical lever. In: Conference Abstracts and Applications of SIGGRAPH 2007,
San Diego, America (2007)
3. Saga, S., Kuroki, S., Tachi, S.: Fibratus tactile sensor using reflection image -the
requirements of fibratus tactile sensor-. In: Proceedings of 16th International Sym-
posium on Measurement and Control in Robotics, Warsaw, Poland (2007)
4. Shimojo, M., Ishikawa, M.: An active touch sensing method using a spatial filtering
tactile sensor. The transactions of the Institute of Electronics, Information and
Communication Engineers J74-C-II(5), 309–316 (1991)
5. Tada, Y., Hosoda, K., Asada, M.: Learn to grasp utilizing anthropomorphic fin-
gertips together with a vision sensor. In: Proceedings of IEEE/RSJ International
Conference on Intelligent Robots and Systems 2005 (2005)
6. Tanaka, M., Chonan, S., Jiang, Z.-W.: Measurement and valuation of touch sensa-
tion (tactile perception of forefinger compared with ae sensor readings. Journal of
the Japan Society of Applied Electromagnetics and Mechanics 5(3), 28–34 (1997)
7. Zhang, Y., Mukaibo, Y., Maeno, T.: A multi-purpose tactile sensor inspired by
human finger for texture and tissue stiffness detection. In: Proceedings of 2006 IEEE
International Conference on Robotics and Biomimetics, pp. 159–164 (2006)