SMIX(λ)

PyTorch implementation of "SMIX(λ): Enhancing Centralized Value Functions for Cooperative Multi-Agent Reinforcement Learning". If you use our code please cite the paper.

Installation

SMIX(λ) is tested on SMAC environment. The installation procedure is the same as pymarl.

Run SMIX(λ)

Run SMIX(λ) on 3s5z map:

python3 src/main.py --config=smix --env-config=sc2 with env_args.map_name=3s5z

Results

Replays of SMIX(λ)

Videos of SMIX(λ)'s performance on 3s5z scenario are available respectively at:

• https://youtu.be/mY13RDVy-qc
• https://youtu.be/2Eut_y-9neQ.

Citing SMIX(λ)

In BibTeX format:

@inproceedings{wen20smix,
  title={SMIX($\lambda$): Enhancing Centralized Value Functions for Cooperative Multi-Agent Reinforcement Learning},
  author={Wen, Chao and Yao, Xinghu and Wang, Yuhui and Tan, Xiaoyang},
  booktitle={Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence},
  year={2020},
}