Table tennis robot is a hand-eye coordination system. It percepts the flying states of the table tennis ball through the visual system, predicts future flying trajectory using the visual measurement data and artificial intelligence algorithms, then designs the optimal trajectory for the robot, finally controls the racket hitting the ball at the right time, position with the best speed and attitude through the mechanical system. This process involves many research problem in the filed of visual perception and intelligent control. Based on the background of table tennis robot, accurate visual perception and intelligent control of high speed moving object are explored and some specific problems such as the rotation estimation, modeling and flying trajectory prediction are studied in this paper. The main contributions of this paper are as follows: Firstly, a simple pose measurement method is designed for rocket to estimate rotation. With two auxiliary lines drawn on the racket pad, the intersections of the auxiliary lines and the pad edge are selected as feature points. A high efficient image-processing algorithm and a feature points extracting method are proposed according to the characteristics of the pad and the auxiliary lines. The racket pose in the camera frame is estimated via PnP positioning approach based on the intrinsic parameters of the camera, the image coordinates of the feature points and their Cartesian coordinates in the racket frame. Secondly, another rotaion estimation method is proposed based on spinning pattern classification of table tennis ball’s flying trajectory. On the basis of force analysis, this paper discusses how the Magnus force influences the flying trajectory under different spinning patterns, and then two fuzzy neural network classifiers are designed to estimate the spinning patterns. Thirdly, the rebound model between spinning table tennis ball and table/racket is given. Based on analyzing the stress, impulse, and impulse moment of the rebound process, a concept of “critical friction angle” is introduced and a condition is put forward to distinguish the type of the friction effect, and then a physical rebound model is gotten. In addition, a linear rebound model is proposed through the learning algorithm and multiple linear regression. Fourthly, the flight model and the rebound model are designed for the spinning table tennis and a fuzzy controller is designed for online rectification of the...
修改评论