Conference PaperPDF Available

Body of a high-speed anthropomorphic table-tennis robot with a linkage mechanism

Authors:
  • OMRON SINIC X Corporation

Figures

Content may be subject to copyright.
Body of a high-speed anthropomorphic table-tennis robot
with a linkage mechanism
Kazutoshi Tanakaa
aGraduate School of Information Science and Technology, Mechano-informatics,
The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku 113-0033, Japan
tanaka@isi.imi.i.u-tokyo.ac.jp
1 Introduction
Robots help us understand and utilize mechanisms of
human adaptive behavior. This behavior includes not only
movements, such as the manipulation of objects and loco-
motion, but also interactions with other humans, such as in
table tennis. A robot can be a valuable tool for studying
high-speed human-robot interactions and human-human in-
teractions, such as human movement/shot prediction in table
tennis, which has an anthropomorphic body shape, moves
fast, and hits a ball anywhere on the table-tennis court.
Many researchers have developed robots that can play
table tennis [1, 2]. These robots can hit a ball anywhere on
the court, but they only have an arm, making it difficult for a
human to predict their future movements in the same way as
in a human-human rally. Xiong et al. presented impedance
control for a humanoid robot playing table tennis [3]. This
robot had an anthropomorphic body but did not step to hit a
ball, and could hit a ball only in a limited area.
For these reasons, Developing a table-tennis robot is
aimed, which has an anthropomorphic body shape so that
its future movement can be predicted by a human player,
moves fast, and can hit a ball anywhere on the table-tennis
court. The main contribution of this paper is to present the
concept and a part of a prototype of such a robot’s design.
2 Design of the table-tennis robot
Our table-tennis robot has an anthropomorphic appear-
ance (the top right of Figure 1) so that its motion can be pre-
dicted by a human player, such as the robots in [3]. These
robots had several joints and actuators to execute various
striking motions and could hit a ball anywhere on the court.
The robot’s weight increases and its speed decreases as the
number of actuators increases. Therefore, one actuator in
our robot is used to drive three joints in its arm to decrease
the number of actuators. Consequently, the robot can exe-
cute only two striking motions so that it moves its position
along sliders as in [1] to hit a ball anywhere on the court. It
was assumed to be crucial for a human player to predict the
other player’s motion from the motion of the arm and trunk,
so that the robot has an arm and trunk. This design approach
can be used for other robots, which move fast and interact
Hand/Racket
Forearm
Upperarm Chest
Hip
Base Y
X
Z
la
lb
l0
l2
θ2
θ1
φ1
φ2
Figure 1: Joint composition of the robot (top left). Appear-
ance of the upper body of the robot (top right). Linkage
mechanism of the arm of the robot (bottom).
with humans, and attaching a head and eyes to the robot en-
ables the player to predict its motion using other ques.
The robot has six links and eight joints (the top left of
Figure 1). The six links are the base, hip, chest, upper arm,
forearm, and hand holding a table-tennis racket. The eight
joints are the Xand Yjoints of the sliders and Zjoints of
the base, hip, chest, shoulder, elbow, and wrist. The Xand
Yjoints move the base on the sliders in front of the table to
change the location of the robot. The Zjoint on the base
changes the height of the hip. The hip joint rotates the chest
horizontally when the robot swings its arm. A linkage mech-
anism coordinates the shoulder, elbow, and wrist. Electric
motors actuate the Xand Yjoints to control the position of
the robot accurately, while pneumatic cylinders move the Z
joint, chest, and joints in the arm for fast striking motions.
The height and weight of the body are 0.6 m and 7.7
kg, respectively. An external compressor (453×682×875
(WxDxH), 118 kg, SLP-221EBD; ANEST IWATA Corp.)
sends compressed air to the robot to actuate its cylinders
(CM2B20-100Z, SMC). Twelve two-port valves (EXA-C6-
Figure 2: Snapshot of the robot swinging: driving motion (top) and pushing motion (bottom).
02CB, CKD Corporation) close and open air flow to control
the air pressure in the cylinders. A computer (BeagleBone
Black, Beagleboard.org) sends commands to these valves.
The robot’s striking motions are pushing motion and
driving motion. The robot swings its racket by extending
the arm downward in the pushing motion and by bending its
arm upward in the driving motion. The design parameters of
the linkage mechanism (the bottom of Figure 1) was deter-
mined such that the shoulder, elbow, and wrist of the robot
move from one end of their range of motion to another in
these motions. The variable φ2in Figure 1 was calculated,
minimizing fas
f= (lb(θ1[0],θ2[0]) lb(θ1[1],θ2[1]))2,(1)
where lb=qd2
x+d2
y,dx=lacosθ1+l2cos(θ1+θ2+
φ2)l1cos(θ1+φ1),dy=lasinθ1+l2sin(θ1+θ2+φ2)
l1sin(θ1+φ1).The variable lbwas was calculated using
these equations and φ2. The arm was designed such that the
shoulder, elbow, and wrist coordinate and moved through
the ranges of π/4–3π/4 rad, 0–π/2 rad, and π/4– pi/4
rad, respectively. The variables lband φ2were thus cal-
culated for the upper-arm settings of la=300[mm],l1=
50[mm],l2=50[mm],φ1[0] = π/4[rad],θ1[0] = π/4[rad],
θ1[1] = 3π/4[rad],θ2[0] = 0[rad], and θ2[1] = π/2[rad]and
forearm settings of la=300[mm],l2=50[mm],θ2[0] =
π/2[rad], and θ2[0] = π/2[rad], respectively. The values
of l1,φ1[0],θ1[0], and θ1[2]were used in the calculation of
the forearm and values of l2,φ2,θ2[0], and θ2[1]in the cal-
culation of the upper arm.
3 Swing experiments
The motions of the robot were measured in experiments
to verify the design. The robot swung its arm in two different
patterns, namely, driving and pushing patterns, in the exper-
iments. The commands for this swinging were determined
through trial and error. The pressure of supplied air was set
at 0.7 MPa. The robot’s motions were measured using eight
motion-capture cameras (Prime13W; NaturalPoint, Inc.). A
marker was attached on the racket to measure the position of
the racket. The motion-capture system measured racket po-
sitions at 120 frames per second. The velocity of the racket
Drive
Push
t [s]
v [m/s]
Drive
Push
Push
Push
Drive
Drive
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
2
4
6
8
Figure 3: Speeds of swinging motions of the robot.
was calculated from the measured positions.
Snapshots of the two swinging motions executed by the
robot (Figure 2) indicates that the robot could execute these
different motions. Figure 3 shows the speed of the racket.
The motions had a maximum racket speed of 8.5 m/s.
4 Conclusion
In this paper, the body of a high-speed anthropomor-
phic table-tennis robot with a linkage mechanism was pre-
sented. Experimental results indicate that the robot executed
two different swings with a maximum racket velocity of 8.5
m/s. This speed was half the humans’ speed of 19.4 m/s [4].
Therefore, the structural parts are going to be replaced with
lighter ones to increase the speed. The whole system of the
robot around the body presented here is being developed,
and whether humans are able to predict the next move of the
robot in table is going to be investigated.
References
[1] Michiya Matsushima, Takaaki Hashimoto, Masahiro Takeuchi, and
Fumio Miyazaki. A learning approach to robotic table tennis. IEEE Trans-
actions on robotics, 21(4):767–771, 2005.
[2] Katharina M¨
ulling, Jens Kober, Oliver Kroemer, and Jan Peters.
Learning to select and generalize striking movements in robot table tennis.
The International Journal of Robotics Research, 32(3):263–279, 2013.
[3] Rong Xiong, Yichao Sun, Qiuguo Zhu, Jun Wu, and Jian Chu.
Impedance control and its effects on a humanoid robot playing table tennis.
International Journal of Advanced Robotic Systems, 9(5):178, 2012.
[4] Yoichi Iino and Takeji Kojima. Kinematics of table tennis topspin
forehands: effects of performance level and ball spin. Journal of Sports
Sciences, 27(12):1311–1321, 2009.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This paper proposes an impedance control scheme used on humanoid robots for stability maintenance when the robot is expected to carry out fast manipulatory tasks. We take table tennis playing as an example to study this issue. The fast acceleration required by table tennis rallying will result in an unknown large reaction force on the robot, causing the body to swing back and forth in an oscillating motion and the foot to lose complete contact with the ground. To improve the stability during fast manipulation and in order to resist disturbances due to the reaction force, we introduce impedance control to absorb the impact and decrease the amplitude of body swinging. The system's adjusting time is also reduced and the oscillations are eliminated according to the experimental results, which show the effectiveness of our scheme.
Article
Full-text available
We propose a method of controlling a table tennis robot so as to return the incoming ball to a desired point on the table with specified flight duration. The proposed method consists of the following three input–output maps implemented by means of locally weighted regression: 1) a map for predicting the impact time of the ball hit by the paddle and the ball position and velocity at that moment according to input vectors describing the state of the incoming ball; 2) a map representing a change in ball velocities before and after impact; and 3) a map giving the relation between the ball velocity just after impact and the landing point and time of the returned ball. We also propose a feed-forward control scheme based on iterative learning control to accurately achieve the stroke movement of the paddle as determined by using these maps. Experimental results including rallies with a human opponent are also reported to demonstrate the effectiveness of our approach.
Article
Learning new motor tasks from physical interactions is an important goal for both robotics and machine learning. However, when moving beyond basic skills, most monolithic machine learning approaches fail to scale. For more complex skills, methods that are tailored for the domain of skill learning are needed. In this paper, we take the task of learning table tennis as an example and present a new framework that allows a robot to learn cooperative table tennis from physical interaction with a human. The robot first learns a set of elementary table tennis hitting movements from a human table tennis teacher by kinesthetic teach-in, which is compiled into a set of motor primitives represented by dynamical systems. The robot subsequently generalizes these movements to a wider range of situations using our mixture of motor primitives approach. The resulting policy enables the robot to select appropriate motor primitives as well as to generalize between them. Finally, the robot plays with a human table tennis partner and learns online to improve its behavior. We show that the resulting setup is capable of playing table tennis using an anthropomorphic robot arm.
Article
The purpose of this study was to investigate whether performance level and ball spin affect arm and racket kinematics of the table tennis topspin forehand. Nine advanced and eight intermediate male table tennis players hit topspin forehands against light and heavy backspins. Five high-speed video cameras were used to record their strokes at 200 fps. Contributions of joint rotations to the racket speed, the racket kinematics at ball impact, the time required for racket acceleration and the maximum slope of the racket speed-time curve (s(max)) were determined. The advanced players showed a significantly larger contribution of lower trunk axial rotation to the racket speed at impact and a significantly larger value of s(max), and tended to require a less time for racket acceleration than the intermediate players. The racket speed at impact was not significantly different between the two player groups. The players adjusted the racket face angle rather than the inclination of the racket path at impact to the different ball spins. The results suggest that the ability to accelerate the racket in less time in the topspin forehand against backspin balls may be an important factor that affects the performance level.