Monopoly AI — MARL, Self-Play with PPO

This repository contains a multi-agent, self-play reinforcement learning setup for Monopoly using Ray RLlib (PPO) and a custom Gymnasium environment.

What’s included

• monopolyEnv.py: Custom Monopoly environment used for training and inference.
• train.py: Trains PPO agents via self-play; saves checkpoints to the training/ folder.
• singleAction.py: Runs a small HTTP server that returns the agent’s next action given an observation (deterministic inference).
• singleActionNonDeterministic.py: Non-deterministic inference variant.
• singleRandomAction.py: random action server.
• test.py / testEnv.py: Simple test/diagnostic scripts.
• training/: Saved checkpoints and artifacts (some examples included).
• requirements.txt and install.py: Dependencies and optional setup helper.

Quick start

1. Install dependencies
   • Python 3.10+ recommended
   • python install.py
2. Train
   • python train.py
   • Checkpoints are written under training/
3. Run the AI
   • python singleAction.py

Notes

• The included training/ subfolders contain example checkpoints.
• Ray may try to use a GPU if available; adjust config in singleAction.py or train.py as needed.