All Tools
M
AgentsFreeOpen Source
MINIWORLD
Lightweight 3D environments for reinforcement learning
Apache-2.0
ABOUT
Full 3D simulators are slow to configure when you only need rooms, hallways, and objects for visual RL. MiniWorld provides FPS-like Gymnasium environments that are easy to customize so agents can learn navigation without standing up Habitat or Unity.
INSTALL
pip install miniworldINTEGRATION GUIDE
1. Train a visual navigation policy in a configurable MiniWorld maze
2. Prototype a custom 3D room layout without a full robotics simulator
3. Benchmark sample-efficient RL on lightweight first-person tasks
4. Teach an agent to pick up and carry objects in a simple indoor world
TAGS
pythonreinforcement-learning3dgymnasiumfaramanavigation