Our project focuses on the problem of generating synthetic levels of a game such that the levels can be used to learn an optimal policy for playing the game. Given a few pre-existing game levels we want to use deep generative models (like GANs) to generate new additional game levels. We will then train an RL agent on these levels to learn a gene…