Skip to content
Log in

Physical-Intelligence/openpi

Embedl's build embedl/pi05-droid of the original model Physical-Intelligence/openpi

π0.5 DROID
embedl/pi05-droid architecture graph
embedl/pi05-droid architecture graph
Labels
  • Text
  • VLA
  • Robot Policy
Available on
NVIDIA Jetson AGX Thor, NVIDIA H200

π0.5 is Physical Intelligence's vision-language-action policy: it takes camera images, the arm's joint and gripper positions and a task instruction, and outputs robot actions. Embedl serves its DROID checkpoint on joint positions on NVIDIA H200 and Jetson AGX Thor, driving a Franka arm in the RoboLab simulator. Pick a device (NVIDIA Jetson AGX Thor, NVIDIA H200) for its measured speed, accuracy and how to run it.