In this project, you will generate PDDL2.1 Minecraft maps. The output includes two versions of the same task - basic (item counts model) and advanced (all blocks model) and you can read more about them in this link.
pip install all the requirements for this project:
python -m pip install -r requirements.txt
- The original domain and problem are located in the directory:
planning/
- To generate new maps (to dataset directory) just run:
python constructor.py
required options:
-h, --help show this help message and exit
--map_size int Size of the map (must be >= 6)
--num_maps int Number of maps to generate
--task {0,1} Task type: 0 for WoodenSword, 1 for PogoStick
You can change the starting item range in the config.json file
If you find our work interesting or the repo useful, please consider citing this paper:
@misc{benyamin2025integratingreinforcementlearningaction,
title={Integrating Reinforcement Learning, Action Model Learning, and Numeric Planning for Tackling Complex Tasks},
author={Yarin Benyamin and Argaman Mordoch and Shahaf S. Shperberg and Roni Stern},
year={2025},
eprint={2502.13006},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2502.13006},
}
@inproceedings{benyamin2023model,
title={Model Learning to Solve Minecraft Tasks},
author={Benyamin, Yarin and Mordoch, Argaman and Shperberg, Shahaf S and Stern, Roni},
booktitle={PRL Workshop in ICAPS},
year={2023}
}