#This repo is made for a project in Deep Learning and Multimedia Data Analysis in MCS Artificial Inteligence of Aistotle University of Thessaloniki.

DRL Doom Project

This project implements Deep Reinforcement Learning (DRL) techniques to train agents to play Doom using the ViZDoom environment. The repository includes training scripts, utility functions, and custom gym environments for experimentation.

Project Structure

• DRL_Doom/train/: Contains training scripts and configurations.
• utils/: Includes utility functions such as DoomGym.py for custom gym environments.
• DRL_Doom/github/ViZDoom/: Contains the ViZDoom environment files.
  • scenarios/: Includes predefined Doom scenarios (basic.cfg, defend_the_center.cfg, deadly_corridor_s1.cfg) for training and testing.

Setup Instructions

1. Clone the Repository:

   git clone <repository-url>
   cd DRL_Doom

2. Install Dependencies: Ensure you have Python installed. Then, install the required packages:

   pip install -r requirements.txt

3. Set Up ViZDoom:

   • Download and install the ViZDoom engine from ViZDoom GitHub.
   • Ensure the scenarios/ folder contains the necessary .cfg files for the Doom scenarios.

Usage

In the testing baseline.ipynb you can see how the agent and env is initialized and how to run the project.

Custom Gym Environments

The DoomGym.py file provides custom OpenAI Gym environments for different Doom scenarios:

• VizDoomGym: For the basic.cfg scenario.
• VizDoomGym_DC: For the defend_the_center.cfg scenario.
• VizDoomGym_DeadlyCorridor: For the deadly_corridor_s1.cfg scenario.

Example usage:

import gym
from utils.DoomGym import VizDoomGym

env = VizDoomGym(render=True, config="github/ViZDoom/scenarios/basic.cfg")
state = env.reset()
# Interact with the environment
action = env.action_space.sample()
state, reward, terminated, truncated, info = env.step(action)

Reward Shaping

The VizDoomGym_DeadlyCorridor environment includes reward shaping based on:

• Movement reward.
• Damage taken and inflicted.
• Ammo usage.

This helps the agent learn more complex behaviors in challenging scenarios.

Visualization

To visualize the logs, run the following on the log file.

tensorboard --logdir=.

License

This project is licensed under the MIT License. See the LICENSE file for details.