UMI GRIPPER
UMI gripper teaching: how GB / GT fit capture
UMI-gripper records human manipulation with a parallel gripper. GB covers video, inertial, and jaw width; GT adds fingertip tactile. Write locally first, then archive; capture and deployment stay separate.
- UMI gripper
- UMI-GB
- UMI-GT
- Manipulation teaching
Two versions, one workflow
UMI-GB records video, inertial, and jaw width for basic ops and early POCs; UMI-GT adds normal/shear fingertip tactile for contact, force, slip, and soft-object work.
Clips enter the sync queue only after a full SD write. Configured units can keep capturing offline until storage fills; they archive to NAS when online.
From teaching to execution validation
Capture mode and robot deployment mode are separate. Policy transfer still requires calibration, action adaptation and real-task evaluation; data capture alone does not promise training gains.
GT packing lists, battery life, and tactile ranges follow the configuration checklist — confirm versions and quantities with sales.
Related FAQ
- What is the difference between UMI-GB and UMI-GT?
- They share one capture workflow. UMI-GB records video, inertial, and jaw width; UMI-GT adds fingertip normal/shear tactile for contact and force-fusion experiments.
- Can capture continue offline?
- Yes. Configured devices still write to SD when Wi-Fi or NAS is down until storage fills; pending clips archive automatically once online.
- Does capture data guarantee robot training success?
- No. Capture and deployment modes are separate. Policy transfer still needs calibration, motion adaptation, and real-task evaluation.
Drawn from public product and solution copy. No customer logos or performance promises. Configuration and acceptance follow the project checklist.