Diffusion Policy generates robot actions with a diffusion model — the technique that made visuomotor imitation learning finally work reliably.
| Category | Robotics & embodied AI |
| Type | Imitation learning |
| License | MIT |
| Runs locally | Yes |
| Built with | Python |
| Skill level | Advanced |
| Best for | cloning a demonstrated skill rather than engineering a controller |
Other open-source robotics & embodied ai tools worth comparing:
GenesisGenerate robotic worlds from a text prompt
LeRobotTrain real robots with the Hugging Face stack
ArduPilotAutopilot for drones, rovers and boats
MuJoCoThe physics engine most robotics research runs on
Stable-Baselines3Reliable RL algorithms you can actually trust
openpi (π0)Open weights for robot foundation models
GymnasiumThe standard interface for reinforcement learning
PX4The autopilot behind most commercial drones
AutowareA complete open-source self-driving stack
Isaac LabMassively parallel robot training on NVIDIA GPUs
ROS 2The operating system layer of modern robotics
Nav2Make a mobile robot navigate on its own
HabitatTrain agents to act in photorealistic homes
MoveIt 2Motion planning and manipulation for robot arms
GazeboSimulate a whole robot, sensors includedDiffusion Policy is free and open-source (MIT license), so you can use, self-host and modify it at no cost.
Yes. Diffusion Policy is designed to run on your own machine or server, keeping your data private.
Popular open-source alternatives include Genesis, LeRobot, ArduPilot. See the comparisons above to choose.
Browse the full directory of open-source AI tools, models and projects — updated daily.
Browse all tools →