Research On The Control System And Hit Strategy Of Ping Pong Robot

Ping pong robot, which covers a series of key robotics technology, such as intelligent control, fast vision tracking, human computer interaction and machine learning, is a typical intelligent and real time system. Due to its high scientific value, it's very effective to research the ping pong robot to improve the development of robotics technology.In this paper, as one part of the"high performance humanoid robot"project which is supported by National High-tech R&D Program (863 Program), we develop a ping pong robot control system and study the control strategies to rally with people.Firstly, to meet the requirements of ping pong robot, a control system, with embedded processor and FPGA as central processor cell, which possesses some characteristic of small size, high integration, good real-timing and high speed of communication, is designed in this paper. Based on the platform, we implant a real-time operating system called VxWorks and Construct high-speed communication network based on Socket and LVDS. As a result, the necessary hardware and software environment of ping pong robot is built.The following is the kinematics of 6-DOF robot and the motor control algorithms. The closed analytical solution of inverse kinematics of 6-DOF manipulator is solved by D-H method. We propose the method of solving the Jacobian matrix which combining the advantages of definition and vector product method and solve the velocity kinematics, and then we study the quintic polynomial trajectory planning and the PD control method with the state observer to realize point to point position control of the motor.In the rest of the paper, we focus on the hit strategy of ping pong robot. We use system identification method to predict the trajectory of ping pong based on the data of visual system, and analyze the collision between the ball and table. And then, we design a protocol based on STNP to synchronize the clocks of the visual PC and embedded processor with millisecond precision. A physical model, which solves the speed and attitude of the pad according to the speed and the predicted position of the coming ball, is established to active stroke the ball. Finally, we calculate the three elements of hitting ball, that is the time, position, attitude and speed of the pad, and propose the task plan to achieve a safe and accurate shot.At last, some experiments are designed to verify the key part of this paper. We test the performance of the control system, conduct a joint position control experiment, and analyze error of ping pong trajectory prediction. Finally, we verify the ability of the robot to hit the ball, test the success rate of single shot and make it rally with human for several round. All of the experiments prove that the algorithm proposed in this paper is correct.