Habitat Lab
NormalDescription
A modular high-level library from Meta to train embodied AI agents across a variety of tasks and environments.
Key Features
- Flexible task definitions for single and multi-agent embodied AI tasks
- Diverse embodied agent configurations including commercial robots and humanoids
- Training via imitation learning, reinforcement learning, or SensePlanAct pipelines
- Human-in-the-loop interaction for data collection and agent evaluation
- Integration with Habitat-Sim as the core physics simulator
- Standardized benchmarking metrics for task performance evaluation
Use Cases
💡 Training indoor navigation agents in simulated home environments
💡 Developing robotic rearrangement and manipulation skills
💡 Research on embodied question answering and instruction following
💡 Building human-robot interaction systems for household assistance
💡 Benchmarking reinforcement learning algorithms on embodied tasks
Categories
Quick Start
1. Create conda environment: conda create -n habitat python=3.9 cmake=3.14.0 && conda activate habitat. 2. Install habitat-sim: conda install habitat-sim withbullet -c conda-forge -c aihabitat. 3. Install habitat-lab: git clone --branch stable https://github.com/facebookresearch/habitat-lab.git && pip install -e habitat-lab. 4. Download test scenes: python -m habitat_sim.utils.datasets_download --uids habitat_test_scenes. 5. Run example: python examples/example.py.