bdvllrs / marl-patrolling-agents

Project on multi agent reinforcement learning applied on patrolling agents

MARL for patrolling agents

We provide here an environment for a predator/prey game. We explore two methods: a simple DQN architecture as well as a true Multi-Agent algorithm architecture using a Policy Gradient approach: Multi-Agent Deep Deterministic Policy Gradient (Lowe, R., Wu, Y., Tamar, A., Harb, J., Abbeel, O. P., & Mordatch, I. (2017). Multi-agent actor-critic for mixed cooperative-competitive environments. In Advances in Neural Information Processing Systems (pp. 6379-6390)).

Some results

After 1400 episodes of training.

DDQN 2vs2	MADDPG 2vs2	DDQN 2v1 Magic Switch

Environment

Blue dots represent preys and orange dots are predators.

Action space

The action space is discrete. Every agent can do one of none, left, right, top, bottom.

State space

The state is perfectly known by all the agents.

The state is the 3D coordinates (x, y, z) for every agent.

marl-patrolling-agents's Issues

there is no code for two predator and two prey ?

Example Config File for 3D Environment

Do you have an example config file that includes a 3D world with obstacles?
Thanks!

How to run the test agent?

Hi, I tested the agent under the test folder but I have error saying :

RuntimeError: size mismatch, m1: [128 x 4], m2: [72 x 512]
Thanks