Dual-Arm Data Collection System

Wuji Hand 2 and Wuji Glove on a force-controlled dual-arm robot. Start collecting bimanual data the day it arrives.

One system, two roles.

Collect real-robot data as it happens.

• Reproduce skilled hand and arm motions on the real robot
• Record arms, hands and 5 cameras on a synchronized timeline
• Export to LeRobot / HDF5 and start training right away

Validate and deploy trained models on the real robot.

• Evaluate trained models directly on the real robot (Sim2Real)
• Run models on the robot with onboard Jetson
• Trial operation in real environments, from lab to field

Collect → train → validate on the same system, for a faster improvement cycle.

System Configuration

Operator

Robot

• 1 Multi-camera 5 cameras on head and wrists, synchronized multi-view recording
• 2 7-DoF force-controlled arm ×2 Arm: Marvin Pro
• 3 Wuji Hand 2 ×2 20-DoF dexterous hand Product page →
• 4 Control & recording PC NVIDIA Jetson AGX Orin
• 5 Stand With casters and leveling feet

* The diagram is for illustration. Number of cameras, stand, etc. can be changed on request.

Put on VR, and move both arms as you are.

Watch the robot’s head-camera view in VR and simply move your hands and arms. Skilled motions are sent to the robot in real time and recorded directly as training data.

* Image for illustration. The standard configuration uses Wuji Glove for finger input and Wuji Hand 2 as end effectors.

Start collecting data in 4 steps, the day it arrives

Delivered assembled and calibrated. Just power on, put it on, and move.

• 01 Put on and operate Wear Wuji Glove and VR, then simply move your hands and arms
• 02 The robot reproduces The dual arms and Wuji Hand 2 perform the same motion in real time
• 03 One-click recording Arms, hands and 5 cameras recorded automatically on one timeline
• 04 Straight to training Export to LeRobot / HDF5 for imitation learning and VLA

Collect data for tasks like these

The ideal platform for your own development

Open hardware, software and data. Use it directly to develop your own algorithms and applications.