The Intelligent Table Tennis Serve Robot

Robotics research is gathering momentum and robots gradually steps from sci-fi movies into people’s daily life. Table tennis robots as one part of robot Family is pushing forward to develop this family. In accordance with the characteristics of table tennis, it is necessary to solve the problems as follows:Enable table tennis robots to have reliable tracing and path prediction capability for a quick moving ping-pong ball, to hit the ball with the appropriate time、location、gesture and speed.At present, robots based on human experiences has been able to play table tennis with people, but those robots have not solidified this knowledge and lack the ability to learn.How to enable the table tennis robots to acquire the ability to learn hitting the ball has turned to be a popular research topic in table tennis robot.In this paper, the writer constructed a simply table tennis robot.In the ball service training progress, a learning mechanism was introduced, so as to form a motion mode to control the hitting of the ball, to serve a ball going along a predicted line and reaching the target point. The progress in details is stated below:1Construct the locomotor system in the table tennis robot. Make mechanical arms with two degree of freedom.Choose a device to shoot balls. Produce a software control system concerning the mechanical arms, in every degree of freedom, achieving specific acceleration、uniform velocity、deceleration and maximum speed control.The locomotor system is qualified to serve balls different with various control combinations.2Construct the vision system for the table tennis robot. Select two industrial cameras to serve as the binocular system, which are calibrated to exactly locate the special coordinates for the ball in angles of view, record its track in the air and extract the features so as to establish a virtual space and reappear the ball flying in the air three-dimensionally. 3Establish the learning system based on artificial neural system, and the RBF neural system is of better generalization capability. The input data for this neural system is the ball track features during flying provided by the vision system while the output datas is the control approach of the corresponding locomotor system.Experiments show that the table tennis ball service robot is capable of serving a ball going along a set curve and reaching the target point. This research provides a new direction for table tennis robotics research, and move forward in personifying robots. At the same time, it has a profound practical value which can be directly applied in table tennis interaction training, and similarly in other ball games, as tennis and baseball, etc.