Gurvan/ssbm_gym
📦 Open Source ProjectGurvan
A specialized OpenAI Gym environment for training AI agents to play Super Smash Bros. Melee.
The ssbm_gym project serves as a bridge between the competitive gaming scene and reinforcement learning research. It wraps the Dolphin emulator, a popular GameCube and Wii emulator, into a standard Gym environment, enabling researchers to apply various RL algorithms directly to Super Smash Bros. Melee.
Key technical features include the integration of game state observation and action space mapping, allowing agents to receive visual or memory-based inputs and execute controller inputs. This environment is particularly useful for studying frame-perfect inputs, complex movement patterns, and adaptive strategies in a high-speed, non-deterministic environment. It provides the necessary hooks for researchers to define custom reward functions, making it a versatile tool for both novice and advanced AI practitioners looking to tackle the challenges of fighting games.
💡Highlights
- ├─OpenAI Gym compatible interface
- ├─Dolphin emulator integration
- └─Customizable reward functions
🎯For
- ├─Reinforcement Learning Researchers
- └─Game AI Developers