hamishs / MultiAgentAI

Multi-agent artificial intelligence - coursework 2

MultiAgentAI

Implementations of some multi-agent deep refinforcement learning algorithms for the games Combat and Snake.

Algorithms

PPO: Proximal Policy Optimisation. Implementation following OpenAI. https://arxiv.org/abs/1909.07528.

COMA: Counterfactual multi-agent policy gradients. https://arxiv.org/pdf/1705.08926.pdf.

Environments

Combat

Multi-agent battle environment. Implementations of PPO and COMA.

MA-Snake

Multi-agent snake on Jidi. Implementation using PPO with self-play.