Whole-body tactile sensing through a force sensor using soft materials in contact areas
© Tsuji et al.; licensee Springer. 2014
Received: 14 January 2014
Accepted: 4 July 2014
Published: 16 September 2014
A force/torque sensor is a useful tool for detecting an external force acting on a robot. Techniques to detect a contact position from a single force/torque sensor have also been developed, but these have used rigid materials in the contact areas. In terms of safety, the material should have shock-absorbing characteristics. Hence, this paper investigates the use of a urethane sponge in the contact areas and evaluates the performance of contact point calculation. First, the relationship between the external force and the displacement of the urethane sponge is measured and a model of the deformation is discussed. Second, a compensation method for soft material deformation is proposed. Finally, the performance of the whole-body tactile sensing system is verified through several experimental results.
KeywordsForce sensing Tactile sensing Haptics
Recent years have seen the use of robots in an increasing range of areas. However, the characteristic features of robots, namely, high-speed motion and high-load operation, are hardly utilized in daily life. It is expected that our lives will be enriched by the spread of human support robots that possess these characteristics. But still numerous issues remain with regard to popularizing human support robots. One of these is the necessity of ensuring safety, as these robots will work in environments where they will come into contact with people. Vacuum cleaning robots, one of the few winning examples of commercialized domestic robots, exert only small forces to avoid potential injury to humans. Hence, it is not imperative for this type of robot to be provided with a force-sensing mechanism.
On the other hand, human-support robots – that can exert large forces should have a force-sensing mechanism in view of safety. Since any part of the robot can come into contact with a human or object in the environment, whole-body tactile sensation technology is needed. Robots intended for human support are required to have a “sense of touch”. The term “sense of touch” used here indicates the capacity to detect the point of application and direction of an external force. Moreover, as contact may occur at any points on the robot, it is desirable that the robot have a sense of touch over its entire body. Whole-body tactile sensing technology is thus essential for the popularization of human support robots.
Several studies have proposed whole-body tactile sensing methods for robots. These methods consider the following tactile sensing data: (a) contact location and (b) magnitude and direction of contact force. The most common whole-body tactile sensing method employs a tactile sensor array as a “skin” . Additionally, conformable and scalable tactile sensors , and flexible and stretchable tactile sensors  have been developed. A skin can be easily customized to a robot by covering non-flat surfaces. Although whole-body tactile sensing can be realized, doing so would warrant the use of a large number of devices on the surface.
One approach to the issue is the use of a tactile camera system. “Gelforce”  is composed of a transparent elastic body, two layers of blue and red markers, and a CCD camera. Force vectors are calculated based on the captured marker movement. The use of a camera reduces the number of sensor devices required, while the system requires space for projection.
Some studies have shown that a robot’s sensing region can be expanded by equipping it with a force sensor. Salisbury proposed a tactile sensing method that identifies the contact point and the force vector on an insensitive end-effector using a six-axis force sensor . Bicchi proposed the concept of intrinsic contact sensing that involves the identification of the contact point and force vector . A few researchers have focused on the use of six-axis force sensors for realizing whole-body haptics. Iwata and Sugano developed an interface for human symbiotic robots. They realized whole-body tactile sensation by molding the end-effector of the force sensor into a shell shape with touch sensors . The authors have proposed “haptic armor,” a tactile sensation mechanism based on a force sensor having a shell-shaped end-effector without any devices . The mechanism has the advantage of cost and wiring reduction, while allowing for six-axis resultant force detection. It is possible to apply the technology for a haptic interface  or for personal authentication .
Whole-body tactile sensation, and thus, improved human safety, can be realized using a tactile sensor array, a camera, and a force sensor. However, in the event of a collision, a robot can damage a human even if the robot is equipped with a whole-body tactile sensing system. There are a few solutions to this problem that involve the use of image sensors for avoiding collisions , but these methods are not always effective at blind angles. As previously mentioned, the end-effector of a whole-body tactile sensing system is composed of a soft material, except for the force sensor. Force sensing methods have used end-effectors made of acrylic, polyvinyl chloride, and aluminum, all of which are highly rigid materials. It is unknown whether a system with a soft material can provide good performance. Therefore, in this paper, we propose a whole-body tactile sensing system using a force sensor with soft material in contact areas. Since an error produced by a deformation of the material is the main issue, this paper presents a method of compensating for soft material deformation. In our previous publications, we have confirmed the advantage of the method through experimental trials ,. This paper consolidate the theory and verify the method through the experimental results.
Whole-body haptics with force sensor
where p denotes the number of surfaces. f k is a function that calculates the distance from the surface the input point. If the end-effector is composed of a highly rigid material such as acrylic, polyvinyl chloride, or aluminum, its shape would remain unchanged upon the application of an external force. Thus, the shape equation becomes (4), which is the same as the equation that does not include external force. Under the assumption that a force is applied from outside the end-effector, contact point position P e can be estimated by the simultaneous solution of (3) and (4).
However, it is difficult to calculate an accurate contact point for an end-effector made of a soft material because the end-effector is deformed by the applied external force. In the next subsection, we discuss compensation for soft material deformation.
Compensation for soft material deformation
Here, d n denotes the component of d, which is vertical to the surface.
Here, F n denotes the normal-direction force on the soft material surface, and K and b denote linear approximation parameters, which are provided by a preliminary test recording the external force acting on and the displacement characteristics of the soft material.
Here, F r and d r are external force and displacement vectors in contact point coordinates, respectively. R k is a rotational matrix that transforms absolute coordinates to relative coordinates based on the k th surface. R k is derived from the kinematic information of each surface. Since F n is the normal direction component of the force,. By substituting the equation to (8), displacement of the soft material d n is estimated.
Equation (14) shows that P e can be calculated from the force information if the shape of the end-effector and contact parameters K and b are known.
Structure of haptic armor with soft material
Figure 3 shows a structural diagram of a haptic armor composed of soft material. An end-effector made of soft material is supported by an acrylic frame. Figure 4 shows diagrams of mechanisms for fixing soft material on a force sensor. As proposed in , consider that the fixation method used is a force sensor covered with soft material, as shown in the right example in Figure 4. This fixation method has two disadvantages:
For an external force applied close to the edge of the end-effector, the correct contact point is not determined because the external force and displacement characteristics are different.
Therefore, this study employs the mechanism of the left example in Figure 4. Under the assumption that the thickness of the soft material is even, accurate compensation is available. It is quite common that some areas have different thickness owing to irregularity of the shape. The performance of the error compensation will be degraded in such area, while the accuracy is still better than the result without any compensation.
Soft material characteristics
Linear approximation based on the data measured when increasing the external force from 2.0 to 20 N,
Linear approximation based on the data measured when increasing the external force from 2.0 N to 20.0 N and decreasing the external force from 20.0 to 2.0 N, and
Linear approximation based on the data measured when decreasing the external force from 20.0 to 2.0 N.
Linear approximation parameters
Linear approximation 1
Linear approximation 2
Linear approximation 3
This paper introduces a displacement estimation method based on linear approximation, while it is also possible to have more accurate estimation with other precise models. However, this study estimates the displacement by a linear function for the simplicity. Simplicity is an overriding matter because this method is useless with time consuming calibration. Additionally, estimation with a nonlinear model does not improve the performance so much because the hysteresis effect, which is difficult to consider in the estimation, is larger than nonlinearity effect.
Results and discussion
Experimental system parameters
Distance from center of Surface 1 edge
Distance from center of Surface 1 edge
d z 1
Distance from center of Surface 2–5 edge
d z 2
Distance from center of Surface 2–5 edge
Thickness of urethane sponge
Here, the third term in (16) shows the soft material displacement estimated from the linear approximation. In the case where the end-effector is composed of a highly rigid material, parameters K and b are zero because such materials hardly deform under the application of an external force. Therefore, when K and b equal zero, (16) is the equation that corresponds to (4). We conducted four experiments for validating the proposed method. First, an experiment to evaluate the aging degradation is shown. Second, an experiment to compare the deformation with different thickness of the urethane sponge is shown. Third, an experiment was conducted for evaluating the difference between the linear approximation and the variation of characteristics across the end-effector. In the last place, an experiment was conducted for validating whole-body tactile sensing capability with soft material.
Evaluation of aging degradation
One important issue of using soft material for force sensation is performance degradation owing to aging. Sugiura et al. have realized force sensing using a photo reflector in soft material and have shown the linearity between material density and voltage response of a photo reflector . Although this result may be one of the results to support small degradation, the evaluation does not include the effect of the degradation by the variation of material density. Hence, this study shows the effect of aging through the experimental result of an old urethane sponge. The sponge was used in the contact area of the robot for 6 months. In spite of the 6 month use, the sponge was in the state without any visible plastic deformation.
Evaluation with different thickness of soft material
Evaluation of compensation
Whole-body tactile sensing capability with soft material
An experiment was conducted for determining the point of action of an external force on five end-effector surfaces. As shown in Figure 13, the point of action was set at the center of each surface for Surfaces 1–5. Additional points of action were set along the grid at 200-mm increments, thus resulting in a total of nine points per surface. A total of 45 such points were set over the entire end-effector. An external force of 15 N was applied at these points along the z-axis with a digital push–pull gauge. Figure 14 shows three types of average errors for the 45 points. These results were obtained under the following conditions:
End-effector with soft material (urethane sponge), as shown in Figure 13, without compensation for linear approximation;
End-effector with soft material (urethane sponge), as shown in Figure 13, with compensation for linear approximation 2; and
In the case of (c), the experiment was conducted using equipment measuring 500 × 500 × 500 mm, while 45 points of action were set over the entire end-effector and along the grid at 200-mm increments, resulting in nine points per surface. As the dark bars in Figure 14 show, the average errors of these 45 points for (a), (b), and (c) were 27.2, 16.3, and 13.2 mm, respectively. The average error of (a) was larger than that of (c) because end-effector displacement was not estimated. For (b), the displacement was estimated by linear approximation. The light bars in Figure 14 show the relative errors by the ratio of absolute error to the size of the end-effector. Since (a) and (b) have 20% larger end-effector, the relative error become smaller in these two cases. The difference between (b) and (c) is quite small in relative error. Therefore, we conclude that whole-body tactile sensing capability with a soft material can be realized by using compensation based on external force and displacement characteristics. Incidentally, there were errors in all cases, i.e., (a), (b), and (c). These errors are unavoidable because commercial force sensors have a maximum of 1% nonlinearity and 1% hysteresis.
This paper proposed a whole-body tactile sensation system with soft material in the contact areas and a method of compensating for soft material deformation. The error of a calculated contact point with the soft material is similar to that with a highly rigid material, when the compensation for the deformation is introduced. Experimental results verified the validity of the proposed mechanism. Future work will focus on evaluating the compensation for the deformation with other soft materials.
This study was supported in part by a Grant-in-Aid for Scientific Research (C) (24560539).
- Mukai T, Hirano S, Nakashima H, Sakaida Y, Guo S: Realization and safety measures of patient transfer by nursing-care assistant robot riba with tactile sensors. J Robot Mechatron 2011,23(3):360–369.Google Scholar
- Ueki S, Kawasaki H, Ito S, Nishimoto Y, Abe M, Aoki T, Ishigure Y, Ojika T, Mouri T: Development of a hand-assist robot with multi-degrees-of-freedom for rehabilitation therapy. IEEE ASME Trans Mechatron 2012,17(1):136–146. 10.1109/TMECH.2010.2090353View ArticleGoogle Scholar
- Guidali M, Schlink P, Duschau-Wiche A, Riener R (2011) Online learning and adaptation of patient support during adl training In: Proc. IEEE Int. Conf. on Rehabilitation Robotics.View ArticleGoogle Scholar
- Tsuji T, Yamada M, Kaneko Y: A robot measuring upper limb range of motion for rehabilitation database. J Robot Mechatron 2013,25(3):515–520.Google Scholar
- Schmitz A, Maiolino P, Maggiali M, Natale L, Cannata G, Metta G: Methods and technologies for the implementation of large-scale robot tactile sensors. IEEE Trans Rob 2011,27(3):389–400. 10.1109/TRO.2011.2132930View ArticleGoogle Scholar
- Ohmura Y, Nagakubo A, Kuniyoshi Y (2006) Conformable and scalable tactile sensor skin for curved surface. In: Proc. IEEE Int. Conf. on Robotics and Automation, 1348–1353.Google Scholar
- Tada Y, Inoue M, Kawasaki T, Kawahito Y, Ishiguro H, Suganuma K: A principle and characteristics of a flexible and stretchable tactile sensor based on static electricity phenomenon. J Rob Soc Jpn 2008,26(2):210–216. 10.7210/jrsj.26.210View ArticleGoogle Scholar
- Sato K, Kamiyama K, Kawakami N, Tachi S: Finger-shaped gelforce: sensor for measuring surface traction fields for robotic hand. IEEE Trans Haptics 2010,3(1):37–47. 10.1109/TOH.2009.47View ArticleGoogle Scholar
- Salisbury J (1984) Interpretation of contact geometries from force measurements. In: Proc. IEEE Int.Conf. on Robotics and Automation, 240–247.Google Scholar
- Bicchi A (1990) Intrinsic contact sensing for soft fingers. In: Proc. IEEE Int. Conf. on Robotics and Automation, 968–973.Google Scholar
- Iwata H, Sugano S (2002) Whole-body covering tactile interface for human robot coordination. In: Proc. IEEE Int. Conf. on Robotics and Automation, 3818–3824.Google Scholar
- Tsuji T, Yasuyoshi K, Abe S: Whole-body force sensation by force sensor with shell-shaped end-effector. IEEE Trans Ind Electron 2009,56(5):1375–1382. 10.1109/TIE.2009.2014748View ArticleGoogle Scholar
- Tsuji T, Ito T (2009) Command recognition by haptic interface on human support robot. In: Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, 3178–3183.Google Scholar
- Arakawa J, Tsuji T, Tan D (2010) Personal identification method for robot with whole-body sensing mechanism. In: Proc. IEEE Int. Conf. on Industrial Electronics, Control and Instrumentation, 1590–1595.Google Scholar
- Cervera E, Gracia-Aracil N, Martinez E, Nomdedeu L, Pobil A (2006) Safety for a robot arm moving amidst humans by using panoramic vision. In: Proc. IEEE Int. Conf. on Robotics and Automation, 1348–1353.Google Scholar
- Kurita N, Hasunuma H, Sakaino S, Tsuji T (2013) Whole-body tactile sensing by a force sensor using soft material at contact areas. In: Proc. JSME Robotics and Mechatronics Conference. 2A1-A09. (in Japanese).Google Scholar
- Kurita N, Hasunuma H, Sakaino S, Tsuji T (2013) Simplified whole-body tactile sensing system using soft material at contact areas. In: Proc. IEEE Int. Conf. on Industrial Electronics, Control and Instrumentation, 4264–4269.Google Scholar
- Lankarani HM, Nikravesh PE: Continuous contact force models for impact analysis in multibody systems. Nonlinear Dynam 1994,5(2):193–207.Google Scholar
- Hayashi M, Sagisaka T, Ishizaka Y, Yoshikai T, Inaba M (2007) Development of functional whole-body flesh distributed three-axis force sensors to enable close interaction by humanoids. In: Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, 3610–3615.Google Scholar
- Sugiura Y, Kakehi G, Withana A, Lee C, Sakamoto D, Sugimono M, Igarashi T (2011) Detecting shape deformation of soft objects using directional photoreflectivity measurement. In: Proc. ACM Symp. User Interface Software and Technology, 509–516.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.