Python

Habitat

Train agents to act in photorealistic homes

by facebookresearch · MIT

From the README

Install

git clone --branch stable https://github.com/facebookresearch/habitat-lab.git
      cd habitat-lab
      pip install -e habitat-lab  # install habitat_lab

Excerpts from the project README on GitHub. Copyright and licensing remain with the respective authors.

What is Habitat?

Habitat simulates indoor environments at thousands of frames per second so agents can learn navigation and object manipulation from vision.

Why people choose Habitat

• Extremely fast photorealistic simulation
• Large library of scanned real homes
• The standard benchmark for embodied navigation

Habitat at a glance

Open-source alternatives to Habitat

Other open-source robotics & embodied ai tools worth comparing:

GenesisGenerate robotic worlds from a text promptLeRobotTrain real robots with the Hugging Face stackArduPilotAutopilot for drones, rovers and boatsMuJoCoThe physics engine most robotics research runs onStable-Baselines3Reliable RL algorithms you can actually trustopenpi (π0)Open weights for robot foundation modelsGymnasiumThe standard interface for reinforcement learningPX4The autopilot behind most commercial dronesAutowareA complete open-source self-driving stackIsaac LabMassively parallel robot training on NVIDIA GPUsROS 2The operating system layer of modern roboticsNav2Make a mobile robot navigate on its ownDiffusion PolicyTeach a robot by showing it, using diffusionMoveIt 2Motion planning and manipulation for robot armsGazeboSimulate a whole robot, sensors included

Habitat head-to-head

FAQ