Generation of roles in reinforcement learning considering redistribution of reward between agents | IEEE Conference Publication | IEEE Xplore