Imagine being able to beat Metal Slug, the classic arcade game, without even lifting a finger. That’s exactly what’s possible with the latest breakthrough in deep reinforcement learning. A Reddit user, AgeOfEmpires4AOE4, has successfully trained an AI agent to play Metal Slug using Stable-Baselines3 (PPO) and Stable-Retro.
The agent, which receives pixel-based observations, was trained specifically on Mission 1, where it faced a tough challenge: dodging missiles from a non-boss helicopter. Despite the difficulty, the agent started to show decent policy learning, prioritizing movement and avoiding close-range enemies.
What’s impressive is that the agent was able to generalize its learning to Mission 2, demonstrating its ability to adapt to new situations. The goal of this experiment was to explore how well PPO handles sparse and delayed rewards in a fast-paced, chaotic environment with hard-to-learn survival strategies.
This achievement has significant implications for the field of AI research, and it’s exciting to think about the possibilities of applying deep reinforcement learning to other complex tasks.
If you’re interested in learning more about the technical details of this project, be sure to check out the GitHub repository and the accompanying video.