Scholarly article on topic 'A Basic Paper Handling Task Experiment Using Tri-axial Tactile Data'

A Basic Paper Handling Task Experiment Using Tri-axial Tactile Data Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Computer Science
OECD Field of science
Keywords
{"Tactile sensor" / "Three-axial data" / "Paper handling" / "Force control" / "Multi-articulated finger"}

Abstract of research paper on Economics and business, author of scientific article — Kenji Sugiman, Masahiro Ohka, Mohammad Azzeim bin Mat Jusoh

Abstract In advanced missions, handling thin and soft membranes is difficult for robots. This task is composed of several sub-tasks, including turning and removing a sheet from a pile of papers on a table, folding paper, and sticking two pieces of paper together. In this study, a hand robot turns and removes a sheet from a pile of papers, and ensures that only one sheet is grasped. To perform this task, two robotic fingers compress the piled papers in the normal direction of the table with a specific force as they close. We therefore incorporated position-based force control into our system; for the force controller, we adopted stiffness control that calculates position modification proportionally to the difference between the target force and an external force measured by the tactile sensor. With this controller, the robot was able to remove a sheet from a pile of papers and ensure that only one sheet was removed. Using this discrimination, the maximum-minimum shearing forces are better classified than with normal force distribution deviation.

Academic research paper on topic "A Basic Paper Handling Task Experiment Using Tri-axial Tactile Data"

(8)

CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedia Computer Science 105 (2017) 270 - 275

2016 IEEE International Symposium on Robotics and Intelligent Sensors, IRIS 2016, 17-20 December 2016,

Tokyo, Japan

A Basic Paper Handling Task Experiment Using Tri-axial Tactile Data

Kenji Sugimana, Masahiro Ohkaa*, Mohammad Azzeim bin Mat Jusohb

aGraduate School of Information Science, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, 464-8601, JAPAN bFakulti Kejuruteraan Mekanikal (FKM), UiTMShah Alam, Selangor, MALAYSIA

Abstract

In advanced missions, handling thin and soft membranes is difficult for robots. This task is composed of several sub-tasks, including turning and removing a sheet from a pile of papers on a table, folding paper, and sticking two pieces of paper together. In this study, a hand robot turns and removes a sheet from a pile of papers, and ensures that only one sheet is grasped. To perform this task, two robotic fingers compress the piled papers in the normal direction of the table with a specific force as they close. We therefore incorporated position-based force control into our system; for the force controller, we adopted stiffness control that calculates position modification proportionally to the difference between the target force and an external force measured by the tactile sensor. With this controller, the robot was able to remove a sheet from a pile of papers and ensure that only one sheet was removed. Using this discrimination, the maximum-minimum shearing forces are better classified than with normal force distribution deviation.

© 2017 The Authors. Publishedby ElsevierB.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of organizing committee of the 2016 IEEE International Symposium on Robotics and Intelligent Sensors(IRIS 2016).

Keywords: Tactile sensor, Three-axial data, Paper handling, Force control, Multi-articulated finger

1. Introduction

The application of advanced robots to not only the industrial field, but also nonindustrial fields such as in-house, medicine, nursing care, and disaster sites,1,2is inevitable. We expect that the use of robots will alleviate issues arising from the decreasing working population caused by the declining birth rate and aging population, and protect humans from danger. To perform such tasks, robots need to process surrounding information through visual and tactile sensors.

Turning flexible and thin objects such as paper is a well-known challenge because it is difficult for robots. Several studies have been conducted on this issue3-5. One solution is the use of a special machine for turning paper4,5. However, using a specific machine to perform each specific task requires too many machines in one work place. In our daily life, we prefer versatile machines such as humanoid robots. Therefore, we aim to create an articulated-fingered hand that handles papers based on three-axis tactile information.

We developed tactile sensors capable of measuring three-axis force distribution to precisely recognize contact situations6 and created a hand robot equipped with tactile sensors to perform the abovementioned advanced tasks7. Recently, this robot was used for human-robot communication via tactile and visual sensations8.

* Corresponding author. Tel.: +81-52-789-4861; fax: +81-52-789-4800. E-mail address: ohka@is.nagoya-u.ac.jp

1877-0509 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of organizing committee of the 2016 IEEE International Symposium on Robotics and Intelligent Sensors(IRIS 2016). doi:10.1016/j.procs.2017.01.221

In this study, we show that this robot can be used for handling paper. This task is composed of several sub-tasks, such as turning and removing a sheet from a pile of papers, folding paper, and sticking two pieces of paper together. Turning and removing a sheet is the first step and is the most difficult task of the sub-tasks. In our preceding study9, we performed a sheet counting task in order to derive classifiers. However, in that investigation, although force control was essential technology for paper turning and counting, we were unsuccessful with the force control.

The purpose of this study is to enable the robot to turn and remove a sheet from papers piled on a table, and ensure that only one sheet is grasped. Since these procedures require feedback of three-axis tactile data to achieve force control, we improved the design of the tactile sensor for high accuracy. Furthermore, the robotic fingers should compress the piled papers in the normal direction of the table with a specific force, while they move to the center of the papers along the longitudinal axis of the papers with a constant velocity. To perform this, we incorporated position-based force control into our system for position-control motor driver usage.

2. Dual Hand-arm Robot Equipped with Optical Three-axis Tactile Sensors

Figure 1 shows the dual hand-arm robot used for this experiment. It is equipped with optical three-axis tactile sensors. Since the robot and the tactile sensors have been previously explained in our earlier papers6'7, we explain briefly herein. Although the robot has two eyes, as shown in Fig. 1, these are used for another experiment8.

Each arm has five degrees of freedom (DOF) and each hand has two articulated fingers. Three of the four fingers have three motorized joints and one has one motorized joint. Since the joints installed on the hand base are used for the wrist joint, each arm's DOF is six.

Fig. 1. Dual hand-arm robot equipped with three-axis tactile sensors. Fig. 2 Optical three-axis tactile sensor.

The three fingers are equipped with the optical three-axis tactile sensor shown in Fig. 2, which has 41 sensing elements capable of sensing three-axis force. Since the tactile sensor is dynamically moved to perform page turning, we redesigned the tactile sensor to replace the optical fiber with LEDs because the optical fiber inhibited free movement9. Here, one fingertip on the right hand finger has a dummy tactile sensor.

3. Position-based Force Control

Since Coulomb friction depends on normal force, force control is essential for page turning. On the other hand, this robot's motor controller is built as position control. We therefore adopt position-based force control10 as the controller, as shown in Fig. 3. This system obtains both target position Pp_ref and target force Fref as inputs. For the force controller, the target position for satisfying target force PF_ref is calculated from target force Fref and measures the external force from tactile sensor Fext. The sum of target position Pp ref and the target position calculated by force controller Pp_ref is applied to inverse kinematics to obtain joint angle )nf (i: joint number). In each joint motor controller, which is shown by a dashed-line square, reference torque r^y^ is output from reference joint angle yef and the measured angle by encoder 9(fyur.

In the force controller, we adopt stiffness control as shown in the following equation:

PF _ ref = C [p.ext ~ Fref ) , (1)

where C is compliance decided through a series of experiments.

Theoretically, this system can maintain target force Fref, but it cannot preciously keep target position Pp_ref because target position PP_ref is modified by PF_ref, as shown in Fig. 3. Since the degree of this modification is mainly determined by the deformation of the sensor element, it is negligibly small.

Fig. 3 Block diagram of position-based force control.

To determine the optimum compliance C, we performed constant compression force tests with different Cs (C = 0.3, 0.5, 0.8, 1.0, 1.2, 1.7, and 2.0 mm/N), as shown in Fig. 4. In these tests, the robot uses two fingers to directly push papers on a table, as shown in Fig. 5, and maintains a compression force of 1.0 N on each finger. In Fig. 4, three Cs cases (C = 0.3, 0.8, and 2.0 mm/N) are exemplified; these force variations are observed on one finger of the hand. As shown in Fig. 4, if C = 2.0 mm/N, although force is converged at 1 N, variation in force is bumpy at the early stage. If C = 0.3 mm/N, it slowly reaches a stable state after 11 seconds. Therefore, we decided that C = 0.8 mm/N is the optimum value.

0 10000 20000

Time [msec]

Fig. 4 Tuning of compliance C via step response tests.

Fig. 5 Procedure for turning a sheet.

Fig. 6 Procedure for taking a sheet.

4. Turning and Taking Paper

In order to turn a sheet, the piled sheets are compressed by two fingers, as shown in Fig. 5, to maintain a constant compressive force. Then, each finger moves toward the center of the paper in 0.5 mm/sec and under constant compressive force. The constant compressive force was decided by a series of experiments, from which we adopted Fref = 0.1, 0.4, 0.7,1.0,1.3, and 1.6 N as the target forces. Figure 5 illustrates a successful case. As shown in Fig. 5, only the top sheet is bent into an arch. If Fref = 0.1 N, the robot took only one, but several times was unable to take any (the percentage of successful cases was 40 %). However, if Fref > 0.7 N, the robot sometimes took more than one sheet. We found that the optimum Fref was 0.4 N; the robot took only one sheet from the piled sheets all 10 times of the 10 trials at 0.4 N. We used this value in the following experiment.

In the turning sheet process, the robot took a sheet according to the following procedure, as shown in Fig. 6: The robot first bends the sheet into an arch using the left hand to get enough arch height for the finger of the right hand; the finger of the right hand is inserted into the arch; and the robot removes the sheet by closing its right hand.

5. Counting the Number of Sheets

Although the robot can take one sheet from the pile of sheets under the condition Fref = 0.4 N with high probability, evaluating the number of sheets is required to confirm it. Therefore, after the procedure for removing a sheet from the pile, the robot ensures the number of sheets according to two parameters derived from our preceding paper:9

ean normal force of element #01 to #08)

K-Y —-t-r-x 100 (2)

(Normal force of element #00 j

R2 = max^Fx (t)2 + Fy (t)2 j - min^Fx (t)2 + Fy (t)2 j (3)

Equations (2) and (3) are used for different grasping modes; Equation (2) is used for the pinch mode and Equation (3) is used for the pinch-first-then-slide-fingers mode. In Eq. (2), elements #00, ..., and #08 show the number of elements and element #00 is located at the summit of the sensor. Elements #01 to #08 are aligned on the circumference at approximately a 3-mm radius around the summit. Equation (2) indicates that if the number of sheets increases, the contact forces of elements #01 to #08 increase because the bending rigidity of the sheets increases. The denominator is for normalization.

In Eq. (3), Fx(t) and Fy (t) are time functions of components of the tangential force vector generated on element #00. When we observe variation in the tangential force, it increases with an increase in finger slide and shows the maximum value; after that, it decreases rapidly to show the minimum value. If the target friction is larger, Rj becomes larger. In the case of two sheets, since slippage occurs between two sheets, R2 shows a small value.

In the preceding study, we concluded that R2 is more appropriate for checking the sheet number than Rj. Since we used contact force control, the denomination of Eq. (2) is completely controlled. In this paper, we found that Eq. (2) is almost constant even if the sheet number increases from 1 to 10. Consequently, we reconfirmed the following: Eq. (3) is more appropriate for evaluating whether one or more sheets have been grasped.

6. Conclusion

In the future society of collaborating robots and humans, robots will perform various kinds of tasks. In this study, we presented a flexible sheet handling task—a very difficult task for robots. To perform this task, force control based on tactile data is used because the friction force that occurs during sheet turning is controlled by normal force. On the other hand, since the present robot motor controller is position based, we used a position-based control system as the controller. Using this force control, the robot successfully performed sheet handling consisting of turning a sheet with the left hand, grasping it with the right hand, and evaluating whether one or more sheets were removed. Since this robot is equipped with a force controller, we re-evaluated the parameters for the sheet number counting introduced in the previous paper by controlling contact force. The difference between the maximum and minimum magnitudes of tangential force is a better parameter to evaluate how many sheets have been removed as the robot hand pinches a sheet and slides fingers along it.

Since the sheet turning and the evaluation of the number of sheets grasped were individually performed in this study, we will perform these tasks sequentially in our future work. Furthermore, we will progress this study to master biomembrane handling in the future.

References

1. Cabinet Office, Government of Japan, Comprehensive Strategy on Science, Technology and Innovation, 2013. Retrieved from

http://www8.cao.go.jp/cstp/english/doc/20130607cao_sti_strategy_provisional.pdf

2.M. Balicki, A. Uneri, I. Iordachita, J. Handa, P. Gehlbach, and R. Taylor, Micro-force Sensing in Robot Assisted Membrane Peeling for Vitreoretinal Surgery, Med Image Comput Assist Interv., 13(Pt 3), pp. 303-10, 2010.

3.K. Murakami and T. Hasegawa, Novel Fingertip Equipped with Soft Skin and Hard Nail for Dexterous Multi-fingered Robotic Manipulation, 2003 IEEE International Conference on Robotics & Automation, pp. 708-713, 2003.

4. http://www.nisic.co.jp/products/assistive/index.html

5. Y. Watanabe, M. Tamei, M. Yamada, and M. Ishikawa, Automatic Page Turner Machine for High-speed Book Digitization, IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 272-279, 2013.

6.M. Ohka, H. Kobayashi, J. Takata, and Y. Mitsuya, An Experimental Optical Three-axis Tactile Sensor Featured with Hemispherical Surface, Journal of Advanced Mechanical Design, Systems, and Manufacturing, 2(5), pp. 860-873, 2008.

7.M. Ohka, S. C. Abdullah, J. Wada, and H. B. Yussof, Two-hand-arm Manipulation Based on Tri-axial Tactile Data, Inter. Journal of Social Robotics, Vol. 4-1, pp. 97-105, 2012.

8. T. Ikai, S. Kamiya, and M. Ohka, Robot Control Using Natural Instructions Via Visual and Tactile Sensations, Journal of Computer Science, 12(5), pp. 246-254, 2016.

9.K. Sugiman, M. A. M, Jusoh, M. Ohka, H. Yussof, and S. C. Abdullah, Thin Flexible Sheet Handling Using Robotic Hand Equipped with Three-axis Tactile Sensors, Procedia Computer Science, Vol. 76, pp. 155-160, 2015.

10. The Robotics Society of Japan, Robotics Handbook (in Japanese), 2nd ed., pp. 287-295, 1990.