Action selection using cooperative occupancy grid map for humanoid soccer robot

In the highly dynamic field of humanoid soccer robot competitions, successful ball possessing and scoring more goals are of the main challenges. But these concerns are not completely resolved due to the existing limitations of the robot such as limited processing power, limited Field of View (FOV), noisy environment, and etc In this paper, we propose a Co-operative Occupancy Grid Map in which teammates send the information of existing obstacles in their FOV to each other so that a vast area of the environment including the obstacles, will be shared and modeled. Then we propose an Action Selection method which uses this co-operative model for managing the ball movement, kicking, passing and dribbling. The results of the proposed approach show a dramatic increase in ball possession and rate of successful kicks.