Paddle Juggling by Robot Manipulator with Visual Servo

Juggling is a typical example to represent dexterous tasks of humans. There are many kinds of juggling, e.g., the ball juggling and club juggling in Fig. 1. Other kinds are the toss juggling, bouncing juggling, wall juggling, wall juggling, box juggling, devil sticks, contact juggling and so on. These days, readers can watch movies of these juggling in the website youtube and must be surprised at and admire many amazing dexterous juggling. The paddle juggling is not major juggling and may be most easy juggling in all kinds of juggling. The paddle juggling means that one hits a ping-pong ball iteratively in the inverse direction of the gravity by a racket as shown in Fig. 2. Most people can easily hit a ball a few times as the left figure since it is easy to hit the ball iteratively. However, most people finally miss out on the juggling as the right figure since it is difficult to control the hitting position. Therefore, realizing the paddle juggling by a robot manipulator is challenging and can lead to solution of human dexterity. As mentioned previously, the paddle juggling of a ball by a robot manipulator is composed of three parts: the first part is to iterate hitting the ball, the second part is to regulate the incident angle of the ball to the racket, and the third part is to regulate the hitting position and the height of the hit ball. M. Buehler et. al. (Buehler, Koditschek, and Kindlmann 1994) proposed the mirror algorithms for the paddle juggling of one or two balls by a robot having one degree of freedom in two dimensional space, where the robot motion was symmetry of the ball motion with respect to a horizontal plane. This method achieved hitting the ball iteratively. R. Mori et. al. (R. Mori and Miyazaki 2005) proposed a method for the paddle juggling of a ball in three dimensional space by a racket attached to a mobile robot, where the trajectory of the mobile robot was determined based on the elevation angle of the ball. This method achieved hitting the ball iteratively and regulating the incident angle. These algorithms are simple and effective for hitting the ball iteratively. However, their method does not control the hitting position and the height of the ball. On the other hand, S. Schaal et. al. (Schaal and Atkeson 1993) proposed a open loop algorithm for onedimensional paddle juggling of one ball. S Majima et. al. (Majima and Chou 2005) proposed a method for the paddle juggling of one ball with the receding horizon control base on the impulse information of the hit ball. Their methods achieved regulating the height of the ball. However, the control problem is only considered in one dimensional space. This study propose a method to achieve paddle juggling of one ball by a racket attached to a robot