Iterative learning system to intercept a ball for humanoid soccer player

Abstract

Soccer for humanoid robots has been a field of study for a long time, and the majority of the teams that compete in a tournament only focus until now in reaching the ball and drive it to score. That is the reason why we think that a more collaborative work would be a real improvement towards accomplishing the RoboCup 2050 ultimate goal of a fully autonomous humanoid team defeating the winning team of the FIFA World Cup Championship of the same year. In this paper, we propose a training system for humanoid-type soccer robot, that will learn to precisely intercept of a ball when is kicked by one robot of the same team. Vision system for ball detection is used as input to predict trajectory of the ball. A knowledge based learning algorithm enables the player to get higher chance to intercept the ball. We confirmed that the proposed approach can be a part of intelligent robot in the field of humanoid soccer.