| Definition | Slime Volleyball is a gym environment for single and multi-agent reinforcement learning, where agents compete to score by getting a ball to land on their opponent’s side. |
| No of environments | 14 |
Slime Volleyball is a lightweight gym environment designed for single and multi-agent reinforcement learning, where agents compete by getting the ball to land on their opponent’s side. Each agent starts with five lives, and the game ends when one loses all lives or after 3000 timesteps. The simple game design, fast iteration speeds, and minimal dependencies make it a flexible tool for testing various RL methods, including self-play, multi-agent learning, and continual learning.
It features pixel observation modes, state-space observations, and options for custom agent policies, making it suitable for educational purposes and advanced research alike. Additionally, the environment can be used as a benchmark for comparing RL agents across a variety of training methods, enhancing its versatility in reinforcement learning experimentation.
| Example observations | |
| Example actions | |
| Example reward |