JP2008036383A - Arm type robot for playing match of table tennis - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a table tennis game machine which is an arm type similar to a man and can play a match using a head. <P>SOLUTION: A robot with a plurality of visual sensors and a plurality of mechanical arms is proposed. Low torque smooth control of an arm is obtained by early estimation of ball trajectory and a ball position and dynamic construction of a continuous curve trajectory from an initial position to a mirror position and a hitting position. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a robot having a plurality of visual sensors and a plurality of machine arms, and an intelligent table tennis game apparatus.

  Table tennis is one of the popular sports. The game is played by players on both sides of the table tennis table serving ping-pong balls, hitting them back, and rallying them. However, sometimes a player may want to play table tennis without an opponent. To that end, a number of table tennis serving devices, robots or game machines have been developed. These devices enable the player to hit the robot back by serving the ball to the player. However, many of the known devices are non-interactive mechanisms that do not react to a human player's ball game, such as catching a ball returned by the player and re-supplying the ball to the robot server (Patent Document 1). .

  The speed of the table tennis ball is fast, and the measurement and prediction of the trajectory requires real time. Although it has already been researched and developed, it generally requires expensive equipment (Patent Document 2). Since the time allowed for performing the hitting operation is short, it is difficult to realize an arm type robot (Patent Documents 3, 4, and 5).

  In recent years, research on robots capable of interactive battle with humans has been reported (Non-Patent Document 1). This representative configuration realizes an operation of hitting table tennis with four degrees of freedom by combining an X, Y linear motor actuator, a racket pitch, and a yaw angle. The position of the flying table tennis is judged by two cameras, and the result is collated with the learned database. By controlling the four degrees of freedom of the racket based on experience data, it is possible to hit the ball back to a certain position.

  However, such an apparatus has a strong image of an industrial Cartesian coordinate robot, and is not interesting and practical.

In addition, such a system generally uses a high-performance actuator, a dedicated hardware image processing apparatus, and the like, and there is a problem that a system as a game machine becomes expensive.
Japanese Patent Application No. 4-355276 Japanese Patent Application 2000-234552 Japanese Patent Application No. 11-116565 Japanese Patent Application 2001-290008 JP-A-06-052316 Journal of the Robotics Society of Japan, Vol. 21, no. 1, pp. 81-86 (2003) “Computer Vision”, Satoshi Sato, Corona, 1999

  The present invention has been made in view of the above-described conventional problems, and is to provide a highly entertaining product. Aiming to be an intelligent table tennis game device that is arm-type and far from humans, away from X and Y linear motor actuators. In addition, taking full advantage of the parallel computing capabilities of personal computers (PCs), which have been remarkably developed in recent years, data from a plurality of image sensors are processed by general-purpose PCs, and a low-priced practical game system is targeted.

  The present invention is characterized by an intelligent robot having a plurality of visual sensors and a plurality of mechanical arms.

  The arm is fixed to the body and has a scalar structure, and a racket can be attached to the tip of the arm. The arm is composed of at least 4 degrees of freedom in accordance with 2 degrees of freedom of the shoulders and elbows and 2 degrees of freedom of the right and left and up and down rotation of the racket. The center of the racket can be freely moved in a parallel plane separated by a certain distance A on the table tennis table surface. The table tennis ball collides with the table tennis table and passes through this surface, passes again, or reaches a distance range shorter than the racket size. At this time, the presence position of the table tennis is defined as “hitting position”.

  Since a plurality of visual sensors are calibrated in advance, the internal parameters and external parameters of the sensors are recorded (Non-Patent Document 2). The three-dimensional coordinates of the table tennis can be obtained from the two-dimensional information of each sensor. The position and speed of the ball can be calculated from the three-dimensional coordinate difference and the time difference of two consecutive frames. By calculating the aerodynamics, the trajectory and hit position of the ball can be predicted. By measuring immediately after the opponent (human player) hits the ball, the robot arm movement time is increased, and the load on the actuator due to sudden acceleration or the like is reduced. The trajectory and hitting position predicted from the information of at least two frames are not highly accurate. However, this method is effective because the initial motion of the hitting arm has low accuracy requirements.

  The robot can quickly determine the arm that is most likely to reach the striking position among the plurality of arms.

  Multiple frames of data are obtained from the visual sensor. Similarly, a new ball trajectory and hitting position can be predicted by aerodynamic calculations. As a result of continuous averaging, the accuracy of the hitting position and the like can be improved with time (the following mathematical formula).

  The robot arm finally reaches a highly accurate impact position using the above data.

  Further, when returning the ball to the opponent, the trajectory can be predicted by the same method, so that the ball player's hitting ability adjustment function can be realized according to the opponent's level by setting the drop position and speed in advance.

  As a known method for orbit control of a robot arm, there is a “mirror law”. When hitting a ball, the ball and the racket move in a facing direction so that a mirror exists at the hitting position. That is, the racket (arm) can be controlled using the ball and the hitting position. Details of the known method are omitted here (Non-Patent Document 1).

  However, in the case of an arm type robot, there is an initial position of the arm. It is difficult to move to the mirror position by the opponent's ball trajectory. The present invention realizes smooth control of the arm by constructing a continuous curved track from the initial position to the mirror position and the striking position (FIG. 1). The present invention names this the “smooth prism law”.

  The smooth prism law uses the “mirror law” to calculate the curve length λ and calculates the image (mirror) position of the ball. Rather than directly using this position for racket control, a continuous function that maps to λ is constructed.

As shown in FIG. 1, the racket starts toward the mirror position where the ball is located, and continuously changes both the position and speed to the mirror position and the striking position via a spline curve or a smooth function such as a circle (continuous to the second derivative). . The tangent of the curve near the hitting position coincides with the ball trajectory direction. The racket speed in the vicinity of the hitting position follows the “Mirror Law”. With the variable λ as a parameter, the equation for the curved trajectory can be expressed as:

  The robot arm can be controlled by the “smooth prism law” using the racket 2D coordinates of the above formula.

  The robot's line of sight is linked to track table tennis to improve entertainment. The table tennis trajectory data is used for arm control and also for driving the eyeball. The calculation method is as follows. At a certain time T, the world coordinates of the table tennis are (X, Y, Z). When converted to a coordinate system with the center of the eyeball as the origin, (X ′, Y ′, Z ′) is obtained. Since the center position of the eyeball is fixed and known, this conversion is only possible. Next, it is converted into a spherical coordinate system with the center of the eyeball as the origin, and becomes (ρ ′, θ ′, φ ′). The eyeball rotates (or moves) within the movable range of the two-dimensional space according to the above (θ ′, φ ′). Since the center positions of the plurality of eyeballs are slightly different, there is a slight difference between the amount of rotation and the direction.

  Such an eyeball is similar in appearance to the eyes of a human or animal, but is not a device that senses an optical signal. In order to ensure the calculation accuracy of the flight trajectory of the table tennis, a plurality of visual (image) sensors may be installed outside the robot body at a distance left and right.

  The robot of the present invention is characterized by a humanoid shape having a plurality of arms and a plurality of visual sensors. Complex ball hitting tasks are realized by early prediction of ball trajectory, proper use of arms, and continuous low torque trajectory generation. According to the present invention, a low-priced practical table tennis battle robot can be configured.

  FIG. 1 is a diagram showing a configuration of a humanoid robot table tennis game apparatus having a trunk (movable on a rail), two arms, and both eyes according to the first embodiment of the present invention when viewed from above and from the side. .

In FIG. 2, (8) is a table tennis table, and there are a human player (9) and a robot player (1) at both ends. The robot has two arms (5) and (6), and the body can be moved on the rail (2). A plurality of visual sensors (3) and (4) can be attached to the robot body. In this example, in order to ensure the calculation accuracy of the flight trajectory of the sphere, it is installed outside the fuselage at a distance left and right.
The plurality of visual sensors (3) and (4) can be attached to the robot body. In this example, in order to ensure the calculation accuracy of the flight trajectory of the sphere, it is installed outside the fuselage at a distance left and right.

Arm hitting trajectory-smooth prism law. It is explanatory drawing which showed the system structural example figure of this invention. Control block diagram.

Explanation of symbols

1 Robot Body 2 Moving Rail 3 Left Vision Sensor 4 Right Vision Sensor 5 Robot Left Arm 6 Robot Right Arm 7 Table Tennis in Flight 8 Table Tennis Table 9 Human Player

Claims (4)

  A robot having a plurality of visual sensors and a plurality of mechanical arms charges an intelligent table tennis game device as a player.   2. The system according to claim 1, wherein the position and trajectory of the table tennis is predicted by aerodynamics from the image data of at least two frames immediately after the human player launches by a plurality of visual sensors, and the early determination of the arm to be used thereby, the arm trajectory Control method. Also, a method of improving the accuracy of table tennis position and trajectory prediction with time from the weighted average value of multiple frame data.   Starting at the mirror position where the ball of the arm (racquet) is located, a continuous smooth trajectory up to the mirror position and hitting position is dynamically generated using data from the method described in claim 2. A motion trajectory control method (“smooth prism law”) that continuously changes the position and speed from the arm standby position to the ball hitting position by continuously mapping the mirror position of the ball to the above curve.   A robot with multiple eyes. A method that uses coordinate transformation to link the robot player's eyes to table tennis.

Images