Action utility prediction and role task allocation in robot soccer system

Abstract

Robot soccer game is a standard platform for Multi-agent System (MAS) research. The object of the paper is to study the cooperation mechanism of MAS in robot soccer games. In order to enhance coordination performance of the role execution subsystem, the paper studied the robot soccer decision-making system based on the dynamic role transition algorithm, and proposed a novel role task allocation mechanism based on action utility prediction. First, a new notion of influence area was presented in the paper, which was used to predict the utility value of a role action. Then, a novel method of selecting the role action based on the concepts and techniques of game theory was proposed. The results in the FIRA 2D simulation platform illuminate that the proposed method is reasonable and efficient.