An Algorithm to Train Person to Person Defense for RoboCupSoccer

Abstract

RoboCup is an international scientific initiative to advance the state-of-the-art in intelligent robots. There are 2D and 3D simulation leagues in RoboCupSoccer focus on artificial intelligence and team strategy in RoboCup. In 2D league, many people have proposed many models about artificial intelligence of eleven agents of their team. But no one has considered about person to person defense in real soccer game. Then we construct an algorithm to train person to person defense using Q-learning.