This environment is an interface with the ALE environment that simulates any ATARI game.
Related paper: Mnih, Volodymyr, et al. "Human-level control through deep reinforcement learning." Nature 518.7540 (2015): 529-533. (Hyper-parameters tuning is necessary if you want to try to replicate close performances.)