Reward-Reinforced Generative Adversarial Networks for Multi-Agent Systems