The Trajectory Prediction And Analysis Of Rotary Table Tennis

Currently, the identification problem of the rotary table tennis has not been solved well. This paper carried out some relevant researches on the flight trajectory of rotary table tennis, including the trajectory prediction of the rotary table tennis and the analysis of the rotary table tennis.This paper studies the trajectory prediction problems of table tennis of ping pong robot system, and propose two new trajectory prediction methods:the trajectory prediction method of the rotary table tennis based EKF and the trajectory prediction method of the rotary table tennis based UKF,both methods can estimate the ball’s motion state in the state of the angular velocity is unknown. Further, the EKF prediction method to predict first derived in the motion model of the rotary table tennis ccordance with the force of circumstances, and then construct the equations of process and equations of observation by the model, at last predict the flight trajectory of the rotary table tennis. For UKF prediction method, first of all do the U transform of sampling points, and then through the iterative of prediction process and update process, finally get the predicted trajectory of the rotary table tennis.Respectively, experiment with the rotation data made by pitching machine and the rotation data made by human, the Matlab experimental results show that the prediction accuracy of EKF and UKF can both meet the ping pong robot system needs, but the latter cost a shorter time, and analyzes the pros and cons of them.In addition, according to the predicted flight trajectory, we designed a pattern recognition classifier of trajectory based BP neural network. First, preprocessing the input trajectory data, and then create a new BP network for pattern recognition, the input is the three-dimensional speeds of the table tennis on five designated section of the flight trajectory, the output is the corresponding rotation type, and the neural network learn a lot of historical data offline, and finally the trained network model can be used for the table tennis trajectory pattern recognition.