chaobiubiu / DCC

Official code for the paper "Coordinating Multi-Agent Reinforcement Learning via Dual Collaborative Constraints".

Coordinating Multi-Agent Reinforcement Learning via Dual Collaborative Constraints

Code for the paper "Coordinating Multi-Agent Reinforcement Learning via Dual Collaborative Constraints" submitted to Neural Networks.

This repository develops DCC algorithm on both Multi-agent Particle Environments and StarCraft Multi-Agent Challenge benchmarks, and compares it with several baselines including LDSA, MAVEN, ROMA, QMIX, HSD and RODE_nr.

Requirements

To install requirements:

pip install -r requirements.txt

Training

To train the approach in the paper, run this command:

python main.py

You can select the training task between MPE and SMAC by setting --env-config='sc2' or 'mpe'. Also you can select the training algorithm by setting --config='dcc'

Hyper-parameters

To modify the hyper-parameters of algorithms and environments, refer to:

src/config/algs/dcc.yaml
src/config/default.yaml

src/config/envs/mpe.yaml
src/config/envs/sc2.yaml

Note

This repository is developed based on PyMARL. And we have cited the SMAC paper in our work.