Skip to content

Latest commit

 

History

History
6 lines (6 loc) · 724 Bytes

README.md

File metadata and controls

6 lines (6 loc) · 724 Bytes

Mario-RL-Agent

The goal of this project is to create a DQN in order to play Super Mario Bros. I used the gym-super-mario-bros environment, which is built on top of the OpenAI gym library. A state of the game is represented by a list of 4 consecutive frames of 240 × 256 × 3 (height × width × 3 channels RGB). The size of the action space in this game is 256 which corresponds to 256 possible actions of the character.

An image captured from the Super Mario Bros

Screenshot (30)

Result

Screenshot (29)