Wuji Hand Setup Guide

From unboxing to your first teleoperation session. Plan for about 20–30 minutes.

1

Unboxing & USB Connect

Remove the Wuji hand from packaging and inspect the finger joints, cable connections, and USB connector for any shipping damage. All 5 finger joints should move freely.

Plug the Wuji hand into your computer via USB. Confirm the device appears:

# Linux — confirm USB device at VID 0x0483
lsusb | grep 0483

If the device appears in lsusb but you cannot open it, add yourself to the dialout group (or add a udev rule for the VID/PID):

sudo usermod -aG dialout $USER
# Log out and back in for the group change to take effect
2

Software Install

Install the Wuji SDK Python package and numpy. The wujihandpy package communicates directly with the hardware over USB and provides the joint position and sensor reading API.

pip install wujihandpy numpy

Verify the install:

python3 -c "import wujihandpy; print(wujihandpy.__version__)"

Optionally install Wuji Studio for a GUI visualization of joint positions and pressure maps without the platform pipeline — useful for hardware bring-up and calibration:

# Linux amd64
wget https://studio.wuji.ai/releases/wuji-studio_0.7.0_amd64.deb
sudo dpkg -i wuji-studio_0.7.0_amd64.deb
3

Start JSONL Stream

Start the JSONL bridge script. This reads from wujihandpy and emits one JSON object per line to stdout at 30 Hz. You should see frames containing bends, joint_actual_position_5x4, pressure_map_24x32, imu, and emf fields.

python3 wuji_hand_sdk_stream.py --hand-side right --hz 30
No hardware available? Use the mock stream to generate synthetic JSONL frames with animated finger bends, a Gaussian pressure blob, IMU data, and EMF readings: python3 mock_wuji_stream.py --hz 30 --hand-side right --seed 42
4

Pair with Robot & Calibrate

Launch the wuji_glove_agent. This wraps the stream script as a subprocess, parses the JSONL output, and forwards decoded sensor data to the Fearless Platform via WebSocket. You'll need a session ID from the platform.

python3 wuji_glove_agent.py \
  --session YOUR_SESSION_ID \
  --node-id wuji-right \
  --glove right

For bilateral (left + right) recording, run two agent processes connected to the same session:

# Terminal 1 — right hand
python3 wuji_glove_agent.py \
  --session RC-XXXX-XXXX --node-id wuji-right --glove right \
  --wuji-cmd "python3 wuji_hand_sdk_stream.py --hand-side right --serial-number SN001"

# Terminal 2 — left hand
python3 wuji_glove_agent.py \
  --session RC-XXXX-XXXX --node-id wuji-left --glove left \
  --wuji-cmd "python3 wuji_hand_sdk_stream.py --hand-side left --serial-number SN002"

After the agent connects, the platform GloveWorkbench panel renders real-time bend bars for all 5 fingers and the 24×32 tactile heatmap. Verify the finger bend values move correctly when you flex each finger.

5

First Teleoperation Session

With the agent running and the platform panel showing live data, start recording a demonstration episode:

  • Navigate to the Episodes panel in the Fearless Platform and click Record.
  • Perform your manipulation task wearing the Wuji hand.
  • Click Stop when the episode is complete. The episode is uploaded immediately and appears in the episode browser.

Episodes are archived as JSONL (one frame per line) with the full joints, pressure_map_24x32, imu, and emf payload plus the session ID and millisecond timestamp. They can be downloaded, replayed, or fed directly into a training pipeline via the platform API.

Works best with OpenArm 101. The Wuji Hand is the recommended operator glove for OpenArm 101 teleoperation. Mount the OpenArm on a stable surface, pair the Wuji agent to the same session as the arm agent, and you have a complete bilateral teleoperation and data collection rig.

For full agent flag reference, JSONL frame format, and platform integration details, see the Wuji Hand wiki page.

Setup Complete?

Check the full specs or join the community for teleoperation tips.