chenlheng / mapr2

Multi-Agent Probabilistic Recursive Reasoning (MAPR2)

Multi-Agent Probabilistic Recursive Reasoning is a multi-agent reinforcement learning framework. The algorithms are based on the paper PROBABILISTIC RECURSIVE REASONING FOR MULTI-AGENT REINFORCEMENT LEARNING in ICLR 2019.

The learning path of PR2-AC in differential game:

Local Installation

To get the environment installed correctly, you will first need to clone rllab, and have its path added to your PYTHONPATH environment variable.

1. Clone rllrb

cd <installation_path_of_your_choice>
git clone https://github.com/rll/rllab.git
cd rllab
git checkout b3a28992eca103cab3cb58363dd7a4bb07f250a0
sudo pip3 install -e .

2. Intsall other dependencies

sudo pip3 install joblib,path.py,gtimer,theano,keras,tensorflow,gym, tensorflow_probability

3. Intsall maci

cd maci
sudo pip3 install -e .

Implemented Algorithms

• PR2-AC/Q
• DDPG
• DDPG with Opponent Modelling
• DDPG with Symplectic Gradient Adjustment Optimization
• MADDPG
• MASQL
• IGA
• WoLF-IGA/PHC
• LOLA-Exact