ALOHA webcam teleop — simulation

Your hands drive the mocap targets of the same MuJoCo scene the policies train in. Hold space while you want the arms to follow · pinch thumb–index to close a gripper · your video stays in the browser.
simulation stream
SPACE
Connecting…
Hold space (or press and hold this bar) while the arms should follow your hands. Release to move your hand back, hold again to continue.
connecting… L disengaged R disengaged L grip — R grip — auto-record — episodes 0 · handoffs 0 net — session —
Instrument models: 3D-COSI, Zenodo (CC-BY 4.0)
How to drive
  1. Allow the webcam and show one or two hands. Your right hand drives the right arm, left hand the left arm.
  2. Hold space — don't tap it. The arms only follow while it's held (the bar above turns green). Release to bring your hand back to a comfortable spot, hold again to continue from where the arm is — like a clutch.
  3. Left/right and up/down follow your hand; move toward or away from the camera for depth (into / out of the scene).
  4. Pinch thumb and index finger to close that gripper; open your hand to release.
  5. Pick the object up with the right arm, carry it above the table to the left gripper, close the left, then open the right — that's a complete handoff.
Recording is automatic once enabled: each stretch of driving is saved when you pause, reset, or leave — idle time is never recorded. Every episode is re-scored later by replaying it.

No webcam? Autopilot runs the scripted expert through the same pipe (not recorded).