Multifaceted Experience-Driven PCG via Reinforcement Learning (MFEDRL)

This is the code for the paper "The Fun Facets of Mario: Multifaceted Experience-Driven PCG via Reinforcement Learning" accepted by the 13th Workshop on Procedural Content Generation.

Please use this bibtex if you use this repository in your work:

@inproceedings{wang2022mfedrl,
  title={The Fun Facets of Mario: Multifaceted Experience-Driven PCG via Reinforcement Learning},
  author={Wang, Ziqi and Liu, Jialin and Yannakakis, Georgios N},
  booktitle = {13th Workshop on Procedural Content Generation at the 2022 International Conference on the Foundations of Digital Games},
  year={2022},
  pages={Accepted},
  organization={ACM}
}

Environments that have been tested

• Python 3.9.6
• JPype 1.3.0
• pygame 2.0.1
• dtw 1.4.0
• scipy 1.7.2
• torch 1.9.0+cu111
• numpy 1.20.3
• gym 0.21.0

How to use

Training GAN generator:

Run command line instruction:

At the root path of this project> python train.py generator

You can check the running arguments (to specify algorithm parameters) by:

At the root path of this project> python train.py generator --help

Training CNet:

Run command line instruction:

At the root path of this project> python train.py cnet

You can check the running arguments (to specify algorithm parameters) by:

At the root path of this project> python train.py cnet --help

Training CNet:

Run command line instruction:

At the root path of this project> python train.py designer

You can check the running arguments (to specify algorithm parameters) by:

At the root path of this project> python train.py designer --help

Play a level:

You may modify the path of level file in line 322 of the smb.py, and then run smb.py to play any level stored in a Mario-AI-Framework-supported text file.

• Line 322 of smb.py

    lvl = MarioLevel.from_file(Path of your level file)

The path can be either related (to project root) or absolute, and the file type won't be checked.