MADDPG

Reproduce MADDPG with PARL

Based on PARL, the MADDPG algorithm of deep reinforcement learning has been reproduced.

Paper: MADDPG in Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments

Multi-agent particle environment introduction

A simple multi-agent particle world based on gym. Please see here to install and know more about the environment.

Benchmark result

Mean episode reward in training process (totally 25000 episodes).

Experiments result

simple	simple_adversary	simple_push	simple_crypto
simple_speaker_listener	simple_spread	simple_tag	simple_world_comm

How to use

Dependencies:

• python3.7+
• paddlepaddle>=2.0.0
• parl>=2.1.1
• PettingZoo==1.17.0
• gym==0.23.1

Start Training:

# To train an agent for simple_speaker_listener scenario
python train.py

# To train for other scenario, model is automatically saved every 1000 episodes
python train.py --env [ENV_NAME]

# To show animation effects after training
python train.py --env [ENV_NAME] --show --restore

# To train and evaluate scenarios with continuous action spaces
python train.py --env [ENV_NAME] --continuous_actions
python train.py --env [ENV_NAME] --continuous_actions --show --restore