Playing MsPacman with Reinforcement Learning
INF581 project, done in collaboration with 3 classmates.
This project is done in the purpose of implementing and applying a Reinforcement learning algorithm, to play the arcade game MsPacman. Particularly, we will be developing the Q-learning model, in order to reach a policy that maximizes the expected value of the earned points. Our agent will learn to eat pellets and avoid ghosts, by making an optimal search based on the current state in the RAM of the game.
A report is available in the files.