Practical overview
When it fits
Use Habitat Lab when a research workflow needs configurable indoor simulation tasks and tools for training or evaluating embodied agents. The documented setup pairs it with Habitat-Sim; install Habitat Baselines separately when using the included baseline algorithms.
Who it is for
- Embodied AI and robotics researchers building agents for simulated indoor environments.
- Developers defining navigation, rearrangement, instruction-following, or other embodied tasks.
- Teams evaluating agents with simulation benchmarks or collecting interaction data.
Documented uses
- Define single-agent or multi-agent embodied tasks, including navigation and rearrangement.
- Configure agents, sensors, and capabilities for simulated tasks.
- Train agents using imitation learning or reinforcement learning, or develop non-learning SensePlanAct pipelines.
- Benchmark agent performance with standard task metrics.
- Enable human interaction with the simulator or trained agents.
Requirements & limits
The repository is public and displays an MIT license.
- Habitat-Sim is a required part of the documented setup and serves as the core simulator.
- The README specifies Python 3.9 or later and CMake 3.14 or later for its Conda setup.
- Habitat Baselines is a separate installation from the core Habitat Lab package.
- The repository README warns that official active development and maintenance by Meta internal teams ended beyond v0.3.4.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Habitat Lab from the stable branch
Prepare the documented Conda environment and install Habitat-Sim and Habitat Lab.
Before you start
- Conda
- Python 3.9 or later
- CMake 3.14 or later
Steps
Create and activate a Conda environment with Python 3.9 and CMake 3.14.0.
Install Habitat-Sim with Bullet physics using the documented Conda channels.
Clone the stable branch of the Habitat Lab repository and install habitat_lab in editable mode.
If baseline algorithms are needed, install habitat_baselines separately in editable mode.
Expected outcome
The documented core Habitat Lab package is installed; the separate Baselines package can be added for training baselines.
Guide limitations
- The core install does not include habitat_baselines.
- The README says newer Habitat-Sim features may require the ai habitat nightly channel.
- The repository warns that official active development and maintenance by Meta internal teams ended beyond v0.3.4.
Run the documented pick-task example
Use the README's sample setup and example script to exercise a configured pick task.
Before you start
- Habitat Lab and Habitat-Sim installed
- 3D test scenes and point-goal navigation episodes downloaded using the Habitat-Sim dataset utility
Steps
Download the documented Habitat test scenes to a data directory.
Download the point-goal navigation episodes for the test scenes to the same data directory.
Run the documented example script from the repository.
Expected outcome
The README describes the example as running a pick task configured by the included YAML task and agent settings.
Guide limitations
- The README notes that its testing scenes do not include semantic annotations.
- The documented sample is a testing example and does not by itself establish results for other tasks or environments.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- facebookresearch/habitat-lab · Official homepage title · retained in its source language