Dynamic modeling based on fuzzy Neural Network for a billiard robot

Abstract

Considering that the motion law of billiards is complicated, a dynamic model based on fuzzy neural network is proposed to predict the final position of cue ball after stroking and collision. The collision coordinate system is established to descript the cue ball location after colliding with the target ball. Based on the relationship between the system's input and output, the Monte-Carlo method is adopted to record the large amounts of data collected by the billiard robot program. Back propagation Neural Network (BPNN) method is used to train the data to establish a fuzzy dynamic model. In the verification test, the billiard robot is able to correctly predict the final position of cue ball after colliding. The statistic result shows that a lower value of the input geometric features is more easily for the robot to learn, which is tallied with the behavior of human beings to play.