Predicting the weights of a weighted mean distance model using BP NN for robot simulated soccer

Abstract

To improve the effectiveness of the defense strategy of Robocup simulation robot soccer team, this paper proposes a BP neural network (BP NN) method to predict the weights of weighted mean distance model. The weighted mean distance model is first introduced into the defense strategy of one robot soccer team. A BP neural network model is then built to modify the weights of the weighted mean distance model to realize more efficient player-attaching defense, which is learned and decided according to the match situations in the court. Experiment results show that the weight predication method for the weighted mean distance model using BP neural network can effectively improve the position preciseness and reliability of the defense players in the real match and can greatly lessen the error probability in robot team's defense behaviors.