The Research On The Trajectory Prediction Of The High Speed Moving Object

The study on ping-pong robot has a history of over twenty years. At the same time, the trajectory prediction algorithm is investigated, which is an important branch of ping-pong robot. A complete flying trajectory can be divided into following four stages:ping-pong ball trajectory tracking, ping pong ball trajectory prediction before rebound, rebound on the table and ping-pong ball trajectory prediction after rebound.3D position of the ball can be detected in the trajectory tracking stage. In the research of the trajectory prediction before rebound, the coordinates of the ball can be predicted according to the sampling points which are detected in the tracking. The direction and the velocity can be changed after rebound on the table. The method of prediction after rebound is adopted by the same method of the prediction before rebound, only the change of the initial velocity.The trajectory prediction of Ping-pong has been a research focus in the ping-pong robot project. In this thesis, the emphasis is giving an introduction of the two trajectory prediction methods:the trajectory prediction of ping-pong based on BP neural network and the trajectory prediction based on kinematic model. When predicting the trajectory based on BP neural network, predictive model is set up. Then by the application of the ping-pong trajectories, the nodes and parameters in BP neural network can be learned off-line. Finally, apply the input-output model to predict the trajectory. In the trajectory prediction based on kinematic model, first of all, RANSAC algorithm is presented to remove the noise points of the sampling points. Secondly, the analytic solution of the flying model is deduced. And by the study of the rebound mechanism, the rebound model can be proposed.According to the experimental data, the prediction accuracy of the two approaches is high, and the time consumed by the approach itself satisfied the requirement of the real time performance, and their advantages and disadvantages are analyzed. A3D simulation vision system is designed to analyze the trajectory prediction.