Multi Agent Constrained Attention Actor Critic

Code for our paper Attention Actor-Critic algorithm for Multi-Agent Constrained Co-operative Reinforcement Learning.

Base code is adopted from Shariq Iqbal's PyTorch implementation of the paper Actor-Attention-Critic for Multi-Agent Reinforcement Learning.

Requirements

Python 3.7.1

PyTorch, version: 1.1.0

OpenAI Baselines

OpenAI Gym, version: 0.9.4

Tensorboard, version: 2.1.0

Tensorboard-PyTorch, version: 1.9

Training

To train on default hyperparameters, run the following commands:

For Constrained Cooperative Navigation: python main.py

For Constrained Cooperative Treasure Collection: python main_treasure_collection.py