MiniStar

MiniStar is an open-source environment built on top of SMACv2, designed to support self-play.

Features

• Self-Play Environment: Built to facilitate self-play strategies for agent training.
• SMACv2 Compatibility: Extends the popular SMACv2.
• Integration with OnPolicy: Utilizes the OnPolicy project for testing and evaluation.

---

Getting Started

Follow these steps to set up and use the MiniStar environment:

1. Prerequisites

Ensure you have the following installed:

• Python 3.6 or higher
• StarCraft II (SC2) installed
• SMACv2 installed

2. Map Setup

Copy the maps provided in the sp_maps directory into your StarCraft II installation:

cp  sp_maps/* $SC2PATH/Maps/Smac_Maps/

Here, $SC2PATH refers to the root directory of your StarCraft II installation.

3. Update smac_maps.py

Use the smac_maps.py file to update the SMACv2 map list. Replace the file in your SMACv2 environment with the updated version:

cp smac_maps.py $smacv2PATH/smacv2/env/starcraft2/maps/smac_maps.py

Here, $smacv2PATH refers to the root directory of your smacv2 installation.

4. Running the Training Script

Start the training script to train a Protoss 10v10 self-play environment:

sh train_protoss_10v10_selfplay.sh

5. Testing with OnPolicy

MiniStar leverages the OnPolicy framework for testing. To test the environment using OnPolicy:

---

---

Contributing

Contributions are welcome! Feel free to submit issues or pull requests to improve MiniStar.

---

License

This project is open-source and available under the MIT License.

---

Acknowledgments

• SMACv2: The foundation for this environment.
• StarCraft II: The platform for agent training.
• OnPolicy: Used as the testing framework for this project.

For further questions or support, please open an issue or contact the maintainers.

MiniStar