Multi agent reinforcement learning: PyTorch implementations of several algorithms for Multi Agent domains

• MADDPG extends the DDPG algorithm to multi-agent settings by using centralized learning and decentralized execution. Each agent has a centralized critic that evaluates all agents' actions, while having a decentralized actor for decision-making.
• Paper: Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments

Demo	Results

git clone [email protected]:mobinajamali/MARL.git

pip install -r requirements.txt

Shout out to Phil Tabor for his brilliant courses!