A topological approach to present input data to perceptron neural network for passing in middle size robot soccer

Abstract

One of the problems in middle size robot soccer, which also can be applied to scanner robots routing in unpredictable environments, is leading passing among robots that comprises choosing the best team-mate to receive the ball without any need to explicit communication among robots. In this paper we have developed an algorithm based on Perceptron neural network for this problem which determines the best passing angle based on the topological data of the play field (i.e. position of robots). With a modification to Perceptron structure and proper data presentation approach a considerable improvement in solution performance has been achieved.