RL Mario
PPO agent on gym_super_mario_bros, clears World 1-1 and 1-2
Overview
A reinforcement learning project training a Super Mario agent with Proximal Policy Optimization (PPO) on gym_super_mario_bros. The agent learns from pixel observations to navigate, avoid enemies, and complete levels.
Highlights
- Algorithm: PPO (stable policy gradients, clipped objective)
- Env:
gym_super_mario_broswith custom observation/action wrappers - Results: Successfully clears World 1-1 and 1-2
Tech
- Python, PyTorch
- OpenAI Gym interface,
gym_super_mario_bros
Repo
- Code and logs: github.com/QingyuanYuu/RL_Mario
This project using PPO algorithm training agent based on gym_super_mario_bros, and successfully passed 1-1 and 1-2.