Trillion UMI is the best handheld device for capturing robotics demonstration data. Collect two precise 6-DoF poses, three camera streams and gripper states data to train next-generation imitation learning models for physical AI
Three synchronized cameras—left-gripper, right-gripper, and head—up to 120 fps. High field of view, global shutter, and low motion blur for crisp, consistent capture.
Full position and rotation tracking with millimeter precision. No external camera required—easy to move around and gather data on real sites.
Continuous monitoring of the gripper opening width, calibrated to provide ground-truth action labels for learning.
Capture demonstrations at natural human pace. No artificially slowed motions like in teleop—data that matches how tasks are really performed.
Take Trillion UMI into the field: warehouses, kitchens, workshops. No need to stay in a lab or replicate environments like with teleop rigs.
Way less expensive than teleop systems. Scale data collection without the cost of dedicated robot hardware and operator setups.
| Spec | Details |
|---|---|
| Cameras | 3 synchronized (left-gripper, right-gripper, head) |
| Frame rate | Up to 120 fps |
| Field of view | 118° (gripper cameras), 150° (head camera) |
| Shutter | Global shutter |
| Motion blur | Low |
| Tracking | 6-DoF (position + rotation) |
| Precision | Millimeter |
| External camera | Not required |
| Gripper | Continuous monitoring, calibrated opening width for ground-truth action labels |
| Deployment | Portable; collect data on real sites (warehouses, kitchens, workshops) |
Get in touch to learn more about Trillion UMI.