plureto

Giving robots access to the physical world.

The Plureto Glove records what a human hand feels. Normal and shear force on every fingertip, 1,000 times a second, in step with every camera on the rig.

Watch the film Talk to us
The Plureto Glove on a hand form, palm toward the camera: the fingertip keycap arrays, the cuff module and the wrist camera.

1 min 20 s · Renders of the simulated glove

01 · The numbers

Shear and normal force. Every fingertip. Every millisecond.

Layout revision 3.1. Design values, and the targets each unit is tested to.

772

Taxels across the fingertips, the phalanges and the palm.

115

Three-axis cells. 23 on each fingertip.

1,000fps

Tactile frame rate. Every taxel, every frame.

Target

±10N

Shear force per fingertip. Held shear does not decay.

Target

0.02–20N

Normal force per fingertip. Survives 120 N.

0.1ms

Alignment between the glove, the cameras and the IMU.

60–120Hz

Hand pose from electromagnetic finger tracking and an IMU.

2.8mm

Keycap pitch. Taxel pitch 1.25 mm.

Target

≤250g

Glove and cuff module together.

Values without a tag follow from the design. Targets are the limits each unit is tested to. First calibrated units are planned for December 2026.

02 · Shear

Shear, not only pressure.

Most tactile gloves read pressure. A manipulation policy also needs the sideways force: the slip, the twist, the drag of an object in the hand.

Close view of the index fingertip of the glove: 23 silicone keycaps in rows over the pad.
Index fingertip · 23 keycaps · 2.8 mm pitch

1

A silicone keycap sits on four taxels that it does not share.

2

The sum of the four readings is the normal force.

3

The differences are the shear in x and in y. The readout works at DC, so a held shear does not fade.

03 · The rig

Glove, cameras and edge computer on one clock.

Put it on. Press start. Every stream carries a stamp from the same clock.

The capture kit laid out from above: the glove with its cuff module, the slide-on wrist camera, the head stereo camera on its band, the edge computer and its cable.
  • Glove and cuff module772 taxels, electromagnetic finger tracking, wrist IMU. One USB-C cable to the edge computer.
  • Wrist cameraA slide-on module on the cuff. It sees the object and the fingers around it.
  • Head stereo cameraThe scene from the worker's point of view.
  • Edge computerStamps every stream within 0.1 ms, records in 60 s segments and encrypts before upload.

04 · The data

What you receive.

Raw signals, calibrated forces and the files your training stack already reads.

Every frame

  • Raw count of each taxel, in sensel-map order
  • Device stamp and sequence number
  • IMU, magnetometer and EMF samples with their own stamps
  • Sync edges and status

Every fingertip

  • Six-axis wrench: Fx, Fy, Fz, Mx, My, Mz
  • 23 keycap force vectors
  • Taxel forces in newtons
  • Hand pose at 60 to 120 Hz

Delivery

  • LeRobot v3 with GR00T modality.json
  • ROS 2 bag (MCAP)
  • Full-rate tactile sidecar and sensel map
  • Calibration file, sync report, QC report

05 · Collection

200+

Partner factories

Data from the factory floor.

We collect bespoke datasets. You define the tasks, the objects and the volume. We run them with our partner factories and deliver the data in the formats above.

Task types from assembly and packing to sorting, inspection and tool use.

Build your policy on touch.

Tell us the tasks, the robot and the volume you need.

Talk to us