Reinforcement learning (RL) has been significantly successful in playing Atari games with human performance. Our project aims to implement and analyze Deep Q-networks (DQN) in an Atari environment, evaluating their effectiveness and potential improvements.