TACTILE / ELECTROMAGNETIC TRACKING
It records not just how the hand moves, but when and where contact happens. A palm-side tactile array combined with electromagnetic fingertip poses provides multimodal research input for dexterous manipulation, teleoperation mapping and visuo-tactile fusion.
Data outputs, interface adaptation and delivery scope are confirmed per task.
A fabric piezoresistive array covers the main palm-side contact areas, with 360 points standard and 600 optional. Raw ADC values and point mapping are recorded; converting to force or pressure requires independent calibration.
Five electromagnetic sensors provide 6-DoF tracking of each fingertip, and an algorithm reconstructs a 25-DoF hand pose. The reconstruction is not 25 independently measured joints, nor ground-truth robot actions.
Data is accessed over USB with a licensed SDK; wireless is optional. In joint experiments with Ego, UMI or robots, calibration, cross-device timing and end-effector mapping must be validated separately.
TWO STREAMS. ONE EXPERIMENT.
Touch and pose are recorded separately and, after calibration and time alignment, become traceable research data.
200 Hz standard · rated
Preserves contact location and relative change. Single-point raw readings can only be interpreted as physical quantities under agreed conditions after dedicated calibration.
Tactile specs180 Hz tracking output · rated
Electromagnetic tracking captures fingertip poses, from which an algorithm reconstructs a 25-DoF hand pose. Observations, reconstruction and robot action mapping are stored in separate layers.
Tracking limitsJoint capture plan: touch + pose + Ego video + task events. The two streams run at different sampling rates; cross-device hardware sync and robot adaptation are not built in and need project validation.
THE DETAILS MATTER
Palm tactile array and fingertip electromagnetic tracking for dexterous manipulation and teleoperation research.
| Item | Standard | Optional / future |
|---|---|---|
| Tactile points | 360 points | 600 points optional; point layout confirmed separately |
| Tactile sampling | 200 Hz | 400 Hz is a future upgrade path, not a current delivery commitment |
| Connection | Wired USB | Optional Bluetooth / wireless, requires receiver and battery module |
| Power | USB | Optional battery module; capacity and battery life confirmed per setup |
| Appearance | Confirmed by style and hand | Color, outer material and protection customizable |
| Base package | 1 glove + 1 data cable | Two-hand sets and tracking equipment listed separately |
The number of gloves shown is not the standard set. Inclusion of the electromagnetic field generator / reference source, receiver, SDK license and related accessories must be confirmed separately.
| Parameter | Spec / configuration |
|---|---|
| Sensing principle | Flexible fabric piezoresistive tactile array |
| Standard point layout | 24 points per finger (16 fingertip + 8 mid-finger); palm 16 × 15 = 240 points; 360 points in total |
| Coverage | Main palm-side areas, excluding the base of the four fingers; final coverage per the point-mapping table |
| Point density | Rated 9 points/cm² |
| Single-point trigger / max range | 20 g / 1500 g |
| Standard sampling rate | 200 Hz |
| Raw output | Per-point ADC values / total ADC; not directly N or Pa |
| Repeatability / consistency | Both ±15% FS (full scale), not ±15% of the current reading |
Contact area, pre-stretch, fit tightness and ambient temperature and humidity all affect readings. Across people, days or re-wearing, record a baseline and validate calibration.
| Parameter | Spec / configuration |
|---|---|
| Sensors | 5 electromagnetic tracking sensors, each providing 6-DoF tracking |
| Tracking output | 3-axis position + 3-axis rotation per fingertip; rated 180 Hz |
| Hand reconstruction | Algorithm-reconstructed 25-DoF pose, not 25 independently measured joints |
| 0.2 mm figure | Tracking accuracy 0.2 mm / resolution <0.2 mm |
| Data latency | Table rates 10 ms wired / 30 ms wireless, and the text says below these values; actual latency must be stated with measurement endpoints and test conditions |
| Deployment and integration | Tracking volume, reference frame, field generator specs and metal-interference conditions |
Documented tracking figures are not treated as measured accuracy, and sensor link latency is not robot end-to-end control latency. Different sampling rates do not mean the two streams are hardware-synchronized.
| Parameter | Spec / configuration |
|---|---|
| Battery | 1200 mAh |
| Battery life / charging | Rated 2.5 hours / 2 hours; load and test conditions not stated |
| Operating temperature | -15 °C to 60 °C, rated |
| Operating humidity | 0–90% RH |
| Reliability | >500,000 impacts at 10 kg (rated) |
| Protection and cleaning | No water-resistance claim for the electronics; avoid liquids, sharp objects and pulling on cables; clean with a soft dry cloth |
| Parameter | Spec / configuration |
|---|---|
| Wired interface | USB communication; custom PID confirmed per project |
| Wireless interface | Optional Bluetooth / wireless receiver and battery module; protocol version and compatibility to be confirmed |
| SDK | Documentation provides Windows / macOS / Linux SDK licenses; license scope per agreement, not open source |
| Custom development | Data frame protocol, point-mapping table and development versions provided as agreed |
| Tactile data | Per-point ADC / total ADC, tied to hand, point table and calibration version |
| Pose data | Position and rotation of each fingertip; reconstructed pose and algorithm version stored separately |
| Recommended archive | Raw frames, device version, hand, anonymous participant ID, wearing baseline, task events, timestamps and clock mapping |
| Platform adaptation | Pairing planned with the Ego Thor series, UMI, Atlas and FMG + IMU; export format and cross-device sync error need project validation |
Raw ADC, calibrated force values, estimated hand pose and mapped robot actions should be stored separately. Data alignment and platform adaptation are part of the integration scope, not a default promise of current device capability.
FROM CAPTURE TO ARCHIVE
Check left/right hand, tactile points, tracking equipment and connection; inspect cables and contact areas.
After putting it on, record the pre-stretch baseline, confirm the electromagnetic reference source, coordinate frame and metal-interference conditions, and complete the calibration the task needs.
Check the actual rate, dropped frames and latency of the 200 Hz tactile and 180 Hz pose outputs separately, saving raw data and versions.
Link to vision or real-robot data, validate time mapping, then generate contact events, reconstructed poses and training data, keeping all processing records.
BUILD YOUR NEXT SETUP
MODULAR / ROBOT LEARNING
Modular platform
Ego Thor SeriesVISION / INERTIAL
Multiple configurations
UMI GripperMANIPULATION / TACTILE
Multiple configurationsSTART WITH YOUR TASK
From device selection to data and training workflows, we'll help you find the right place to start.