Skip to content

A Deep Convolutional Q-Learning implementation that trains an AI agent to play Pac-Man using reinforcement learning techniques.

Notifications You must be signed in to change notification settings

dhritishetty/Pac-Man

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

Pac-Man: Deep Convolutional Q-Learning for Classic Arcade Play

This project demonstrates the application of Deep Convolutional Q-Learning (DQL) to train an AI agent to play the classic Pac-Man game. The goal is to develop an agent that can navigate the game environment, avoid ghosts, and eat all the pellets efficiently.

Project Overview

  • Environment: The Pac-Man environment from the Gymnasium library.
  • Algorithm: Deep Convolutional Q-Learning (DQL), an advanced reinforcement learning algorithm that combines Q-Learning with convolutional neural networks.
  • Objective: Train an agent to play Pac-Man by optimizing its actions through experience and exploration, ultimately mastering the game.

Features

  • Implementation of a Deep Convolutional Q-Network (DQN) with experience replay and target network.
  • Visualization of training progress and agent's performance.
  • Evaluation of the trained agent's ability to play Pac-Man in various scenarios.

How to Run

1. Access the Colab Notebook: Open the Pac-Man Colab Notebook to view and run the project code directly in Google Colab.

2. Run the Cells: Execute the cells in the notebook sequentially to train the agent and visualize its performance.

3. Evaluate the Agent: After training, evaluate the agent's performance by running the evaluation cells in the notebook.

Results

To see the results of this project, visit the Google Colab notebook where you can interactively view the training process and visualizations.

About

A Deep Convolutional Q-Learning implementation that trains an AI agent to play Pac-Man using reinforcement learning techniques.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published