Skip to content

gameaischool2021members/roguelife

Repository files navigation

Roguelife 💀

The game that both plays and develops itself

More elaborated, this project is about pitting AI agents against a procedural level generator. The level generator will adapt to the player over time. For illustration purposes, we have designed some rule-based agents who use very specific strategies, to see how the generator "responds" to that. For the full experiment, we want to run a learning agent (using a DQN), that will learn to exploit the environment, all while the environment learns to exploit the agent.

Authors

  • Jonathan Jørgensen
  • Pedro M. Fernandes
  • Even Klemsdal
  • Niels NTG Poldervaart

Architecture

Architecture

Running

Evolve Levels for the Rule Based Agents

  • Run the run_all.sh script. This will evolve a population of difficult maps for each one of the Rule Based Agents.

Train the RL Agent against the environment

  • Run the run_dqn.py file with the --train argument and the name of the training session. Example: "python3 run_dqn.py --train Q01"

Dependencies

pygame
numpy
gym
pillow
scipy
stable_baselines3

Screenshots

Video

Training Montage

Presentation Slides

Click here!

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published