DQN-BattleCityNes

Multi-agent Reinforcement Learning on simulated environment of BattleCity, an NES game. The game environment is generated using the OpenAI Gym Retro Game Integration Tool. This tool generates a state file containing the two player start state, a JSON data file describing the game variables like enemies left, lives left, player 1 score, player 2 score. Deep Q-Networks are used to perform the RL task to teach the two agents to play the game successfully and a playback of the agents is generated as a video.