- Collect Without the Robot The handheld gripper records demonstrations at a bench, in a kitchen or in a warehouse aisle.
- Same View for Human and Robot Handheld and robot-mounted grippers share the camera viewpoint and finger design.
- 5.3K Wide-Angle Video A GoPro Hero 12 records at 60 fps with a 155° field of view and a built-in IMU.
- 6D Pose Tracking Visual-inertial SLAM based on ORB-SLAM3 tracks the gripper and annotates each session.
- Data to Deployment The software runs collection, calibration, training and deployment, with an SDK and Docker environment.
- Up to 4,000 Demonstrations a Day Dobot rates that throughput for a fast workflow, depending on the task.
Description
The Dobot PocketGo collects robot training data without the robot. An operator performs a task with a handheld gripper that carries a camera, wherever the task happens, and the system records what the gripper saw and exactly how it moved. A matching gripper on a Dobot arm then uses that data to learn and repeat the skill, following the behaviour cloning approach of research systems such as UMI.
The kit holds two gripper sets, one handheld for recording and one robot-mounted for running policies, with the same camera viewpoint and finger design so that what a model learns from people carries over to the robot. Recording runs through a GoPro Hero 12 at 5.3K and 60 fps, with a 155° field of view, HyperSmooth 6.0 stabilisation and a built-in IMU. Visual-inertial SLAM based on ORB-SLAM3 tracks the gripper's 6D pose and annotates each session automatically.
The software covers the full pipeline: collection, calibration, SLAM trajectory processing, training and deployment, with a secondary development SDK and a Docker environment for custom research. Dobot rates throughput at up to 4,000 demonstrations a day, depending on task length and workflow, and the gripper handles objects up to 2 kg.
Deployment needs a Dobot Nova or CR series arm, sold separately. Training needs a GPU: the optional mini AI computer (Intel i9 with an RTX GPU of 11 GB VRAM or more) or your own workstation. High-resolution video from thousands of demonstrations also needs plenty of storage.
Key Capabilities
- Collect Without the Robot: The handheld gripper records demonstrations at a bench, in a kitchen or in a warehouse aisle.
- Same View for Human and Robot: Handheld and robot-mounted grippers share the camera viewpoint and finger design.
- 5.3K Wide-Angle Video: A GoPro Hero 12 records at 60 fps with a 155° field of view and a built-in IMU.
- 6D Pose Tracking: Visual-inertial SLAM based on ORB-SLAM3 tracks the gripper and annotates each session.
- Data to Deployment: The software runs collection, calibration, training and deployment, with an SDK and Docker environment.
- Up to 4,000 Demonstrations a Day: Dobot rates that throughput for a fast workflow, depending on the task.
Buying through Robotix Market
This unit is sourced new through established distribution channels and ships worldwide with the manufacturer warranty. Technical support is coordinated with the manufacturer's own service network. Deployment scope and configuration are agreed before dispatch, and on-site installation is not included in the product price.
Features
Specs
| Brand | Dobot |
|---|---|
| Model | PocketGo |
| Part Number | 1200000404 |
| Product Type | Portable single-arm demonstration data collection platform |
| Gripper Kits | Handheld kit and robot-mounted end-effector kit |
| Gripper Payload | Up to 2 kg |
| Camera | GoPro Hero 12, 5.3K at 60 fps, 155° field of view, HyperSmooth 6.0, built-in IMU |
| Pose Tracking | Visual-inertial SLAM (ORB-SLAM3), 6D pose |
| Software | Collection, calibration, SLAM trajectory processing, training and deployment |
| Development | Secondary development SDK, Docker environment |
| Throughput | Up to 4,000 demonstrations per day, depending on task length and workflow |
| Compatible Robots | Dobot Nova and CR series (not included) |
| Optional Computer | Mini AI computer, Intel i9 with RTX GPU (11 GB VRAM or more), not included |
| Warranty | 12 months manufacturer warranty |
| What's in the Box | Handheld gripper kit, robot-mounted gripper kit, GoPro camera, software stack, SDK and Docker environment |
Industries
Frequently Asked Questions
No. Data is collected with the handheld gripper. A robot is needed to deploy the trained policy.
Dobot Nova and CR series arms, sold separately.
Visual-inertial SLAM based on ORB-SLAM3 tracks the gripper's 6D pose from camera and IMU data.
No. A mini AI computer with an Intel i9 and an RTX GPU is optional, or your own GPU workstation can train the models.
Dobot rates it at up to 4,000 demonstrations a day, depending on task length and workflow.
The Dobot PocketGo Data Collection Platform is priced at SAR 26,100 in your market. For volume or project orders, contact us for a business quote.
Recently Viewed
You haven't viewed any products yet.
