Cooperative Deep $Q$-learning Framework for Environments Providing Image Feedback