All Tools
H
AgentsFreeOpen Source
HABITAT-LAB
Train embodied AI agents across indoor tasks and robots
MIT
ABOUT
Embodied agents need task definitions, robot configs, and training loops on top of a simulator. Habitat-Lab supplies navigation and rearrangement benchmarks, PPO baselines, and Gym wrappers so teams train indoor agents without writing that stack from scratch.
INSTALL
pip install habitat-labINTEGRATION GUIDE
1. Train point-goal and object-goal navigation agents on indoor scans
2. Benchmark rearrangement skills like pick, place, and open on ReplicaCAD
3. Run multi-agent or human-in-the-loop Habitat 3.0 co-habitation tasks
4. Fine-tune PPO baselines from habitat-baselines on custom configs
TAGS
pythonagentsembodied-aireinforcement-learningroboticshabitatfacebook