PettingZoo
E95191
PettingZoo is a Python library that provides a standardized interface and tools for developing, running, and benchmarking multi-agent reinforcement learning environments.
Statements (47)
| Predicate | Object |
|---|---|
| instanceOf |
Python library
→
multi-agent reinforcement learning framework → |
| aimsTo |
improve reproducibility in multi-agent reinforcement learning research
→
simplify comparison of multi-agent RL algorithms → |
| compatibleWith |
RLlib
→
Stable Baselines →
surface form: "Stable-Baselines3 (via wrappers)"
Tianshou → |
| documentationURL | https://pettingzoo.farama.org → |
| domain | artificial intelligence → |
| focusesOn | standardization of multi-agent RL interfaces → |
| hasAPIStyle |
AEC API
→
classic API → parallel API → |
| hasEnvironmentType |
Arcade Learning Environment
→
surface form: "Atari multi-agent environments"
MPE (Multi-Agent Particle Environments) → board games → butterfly environments → classic control → |
| hasFeature |
environment versioning
→
seeding for reproducible experiments → support for vectorized environments → wrappers for environment preprocessing → |
| hostedOn | GitHub → |
| inspiredBy | OpenAI Gym NERFINISHED → |
| license | MIT License NERFINISHED → |
| maintainer | Farama Foundation → |
| partOf | Farama Foundation ecosystem → |
| programmingLanguage | Python → |
| provides |
standardized interface for multi-agent environments
→
tools for benchmarking multi-agent environments → tools for developing multi-agent environments → tools for running multi-agent environments → |
| relatedTo |
Gymnasium
→
SuperSuit → |
| subdomain |
multi-agent systems
→
reinforcement learning → |
| supports | multi-agent reinforcement learning environments → |
| supportsActionSpace |
continuous action spaces
→
discrete action spaces → |
| supportsAgents |
competitive agents
→
cooperative agents → mixed cooperative-competitive agents → |
| supportsObservationSpace |
continuous observation spaces
→
discrete observation spaces → |
| usedFor |
benchmarking multi-agent RL algorithms
→
education in reinforcement learning → research in multi-agent reinforcement learning → |
Referenced by (2)
Full triples — surface form annotated when it differs from this entity's canonical label.