The Research On The Trajectory Prdiction And Classification Of The Table Tennis Trajectory

Currently, there are two puzzles of Ping-Pong robot system:It’s hard for robot to response in a short time, when table tennis movement in high speed condition; Ping-Pong robot cannot judge the type of opponents hit the ball (e.g. Backspin, Topspin, Regular), can’t judge the rotation and the direction of rotation, which will cause the robot to return the ball in single strategy, poor adaptability. To solve this two problems, this paper launches the research work on the trajectory of the Ping-Pong on Ping-Pong robot system, mainly including two parts:the prediction of trajectory and classification of rotation ping-pong trajectory.Prediction of the trajectory is the base of research in Ping-Pong robot. This paper presents an approach to predict the trajectory of ping-pong ball based on the Extreme Learning Machines(ELM) algorithm. Firstly, we get the real experiment data between human and machine. Then we extract the postion and speed information of10consecutive frames as the characteristic, which is the input of ELM algorithm. Then we preprocess the input trajectory data and creat a new ELM network for prediction, whose output is the postion information of after20frames. Through several experiments we found deficiency of original ELM algorithm, so we improve ELM algorithm.The improved ELM learns a lot of historical data off line and saves weight matrix of the hidden layer to save the trained ELM model.At last, we verify the feasibility and applicability of ELM in prediction of trajectory by experiments.Predict of the trajectory by Error Back Propagation(BP) neural network. Select the same data and method with ELM algorithm and creat BP prediction network, whose output is also postion information of after20frames. The neural network learn a lot of historical data offline, and the best can be saved to predict the new trajectory, which proves the feasibility of BP neural network. Finally, we compare BP neural and improved ELM algorithm, verifying the superiority of BP in accuracy and superiority of ELM in time consuming.In summary, we consider the ELM algorithm is more suitable for the Ping-Pong robot in predicting of trajectory.When using ELM to classify the ro trajectory. Similarly select the characteristic, position, speed of every frame of rotation ping-pong trajectory as the characteristic. Experiments verify the feasibility and applicability of ELM in classification of rotation ping-pong trajectory. Finally, we compare BP neural and improved ELM algorithm. Experiments show that compared with BP. the accuracy of FLM is relatively a little low. but the consuming time is much less.