DOBOT Pocket Go Data Collection Platform for Embodied AI
Shipping from China within 5–10 working days
- During check out, you can choose direct shipping from China, or let us import the order to our EU/US office first and ship it to you locally.
- Import customs clearance and VAT are fully handled by us.
Need a larger quantity?
Need bulk pricing?
Get a formal quotation with lead time and volume pricing.
For universities, companies, distributors, and project orders.
DOBOT Pocket Go Data Collection Platform for Embodied AI
DOBOT Pocket Go is a portable data collection platform for embodied AI and imitation-learning workflows. It pairs a handheld gripper kit with a robot-mounted end-effector gripper kit, synchronized GoPro cameras, visual-inertial pose tracking, and an end-to-end software stack so an operator can capture high-quality robot manipulation demonstrations for compatible DOBOT Nova and CR robot arms. Pocket Go is designed to shrink the setup burden of collecting repeatable demonstrations and to move the data directly into training and deployment.
Overview
Pocket Go turns human demonstration into machine-ready training data. An operator performs a task with the handheld gripper while synchronized cameras and onboard pose tracking record every movement. The same end-effector gripper kit then attaches to a compatible DOBOT robot arm so the recorded policy can run autonomously without hardware modification. The result is a closed loop from teleoperation to imitation learning that research labs, universities, and industrial AI teams can repeat at scale.
Key Features
- Handheld and robot-mounted collection kits — The system includes a handheld gripper kit and an end-effector gripper kit for paired demonstration capture, each with its own GoPro camera.
- Camera-based observation — GoPro Hero 12 cameras with 5.3K capture, 155-degree field of view, and HyperSmooth 6.0 stabilization record every demonstration.
- Visual-inertial 6D pose tracking — ORB-SLAM3 fuses camera and IMU data to generate precise 6D trajectory annotations automatically, without manual labeling.
- High-throughput capture workflow — The platform is listed for up to 4,000 demonstrations per day; real throughput depends on task design and operator process.
- Robot-arm compatibility — Built for compatible DOBOT Nova and CR robot arms (Nova 2, Nova 5, CR3, CR5, CR7), which are separate hardware and are not included in this listing.
- Portable operating format — A mobile stand and software platform help move between collection scenarios and laboratory setups.
- Full AI training pipeline — The included software stack covers data ingestion, SLAM trajectory processing, model training, and deployment, with a secondary-development SDK and Docker-based environment for custom research integrations.
How the Data Collection Pipeline Works
Pocket Go is more than a gripper — it is an end-to-end embodied-AI workflow. The documented pipeline is:
- Demonstrate — Operate the handheld gripper kit to record a manipulation task; synchronized GoPro cameras capture the scene.
- Track — ORB-SLAM3 visual-inertial SLAM generates 6D pose trajectories automatically from each recorded session.
- Collect — Repeat to build a labeled demonstration dataset (up to 4,000 demos/day under the stated workflow).
- Train — Use the software stack to process SLAM trajectories and train an imitation-learning or diffusion policy.
- Deploy — Attach the end-effector gripper kit to a compatible DOBOT Nova or CR arm and run the trained policy for autonomous inference.
This closed loop lets the same recorded data flow from human teleoperation into a deployable robot policy with minimal reconfiguration.
Technical Specifications
| Parameter | Value |
|---|---|
| Product type | Portable embodied AI data collection platform |
| Capture kits | Handheld gripper kit and robot end-effector gripper kit (each with a GoPro camera) |
| Listed camera | GoPro Hero 12 (1/1.9" CMOS, 24.69 MP, 155-degree FOV, HyperSmooth 6.0) |
| Video resolution | 5.3K (8:7) @ 30fps; 5.3K (16:9) @ 60fps; 4K (8:7) @ 60fps; 4K (9:16) @ 60fps |
| Pose tracking | Visual-inertial 6D tracking based on ORB-SLAM3 |
| Listed collection capacity | Up to 4,000 demonstrations per day |
| Handheld gripper range | Opening range 73.5 ± 1.5 mm, gripping depth 100 mm, max gripping weight ≤ 2 kg, net weight 0.93 kg |
| End-effector gripper range | Opening range 73.5 ± 1.5 mm, object size 5–75 mm, gripping depth 100 mm, max gripping weight ≤ 2 kg, net weight 1.58 kg |
| Maximum gripper payload | 2 kg |
| Mobile stand | 680 × 600 × 664 mm, 4 swivel casters, fits Nova 2 / Nova 5 / CR3 / CR5 / CR7 |
| Supported robot arms | Compatible DOBOT Nova and CR robot arms (Nova 2, Nova 5, CR3, CR5, CR7), sold separately |
| Software environment | Ubuntu 20.04 and Ubuntu 23.10 (Linux only), listed in the product materials |
| Training GPU (recommended) | ≥11 GB VRAM (NVIDIA RTX 4090 recommended); inference GPU ≥2 GB VRAM |
| Recommended training data | ≥200 task demonstrations for reliable policy performance |
| Optional computer | Mini AI reasoning computer is optional and not included |
Software & System Requirements
The Pocket Go software platform runs on Linux only. The product materials list Ubuntu 20.04 and Ubuntu 23.10 as the supported operating systems. Pose tracking uses the ORB-SLAM3 visual-inertial monocular SLAM algorithm for 6D pose estimation. For local model training and inference, DOBOT recommends an NVIDIA GPU with at least 11 GB VRAM (RTX 4090 class); an optional Mini AI reasoning computer (Intel Core i9, NVIDIA RTX ≥11 GB GDDR6, 64 GB DDR5, 2 TB SSD) is available separately for edge training and inference.
Compatible Robot Arms
Pocket Go is built for the full DOBOT Nova and CR collaborative series. The mobile stand accommodates Nova 2, Nova 5, CR3, CR5, and CR7. CR10 and larger arms are not compatible with the standard stand. A compatible robot arm is required for autonomous inference and must be ordered separately.
Configuration & Packing List
The standard product materials list a handheld gripper kit with camera, a robot end-effector gripper kit with camera, a software platform with detailed user manual and SDK, and a mobile stand. Compatible DOBOT robot arms and the optional Mini AI reasoning computer are not included.
| Item | Quantity |
|---|---|
| Handheld gripper kit with GoPro camera | x1 |
| Robot end-effector gripper kit with GoPro camera | x1 |
| Software platform with user manual and SDK | x1 set |
| Mobile stand | x1 |
| Compatible DOBOT robot arm | Not included |
| Mini AI reasoning computer | Optional, not included |
Applications
- Embodied AI demonstration recording
- Imitation learning and policy-development datasets
- Robot manipulation research with compatible DOBOT arms
- Portable teleoperation and task-demonstration capture
- University and industrial AI-robotics experimentation
Developer Resources
Pocket Go ships with a complete software platform — data collection, SLAM trajectory processing, model training, and deployment — plus a secondary-development SDK and a Docker-based environment for custom research integrations. Because collected data and trained policies are deployed to DOBOT Nova and CR arms, the robot-side SDKs below are the standard tooling for turning demonstrations into running robot code.
DOBOT Official Developer Channels
- DOBOT Developer Portal — official SDKs, API references, and documentation hub.
- DOBOT GitHub (Dobot-Arm) — official open-source SDKs, demos, and protocol implementations.
- DOBOT Download Center — user manuals, software, and technical documents.
- Technical support: support@dobot.cc
Nova / CR Robot Arm SDKs (Deploy Trained Policies)
- TCP-IP-Python-V4 — Python SDK for V4-series Nova/CR arms over TCP/IP (motion, IO, feedback).
- TCP-IP-CR-CPP-V4 — C++ SDK for V4-series Nova/CR arms.
- DOBOT_6Axis_ROS2_V4 — ROS 2 package for V4-series arms.
- DOBOT_6Axis_ROS2_V3 — ROS 2 package for CR/Nova V3 arms.
- TCP-IP-ROS-6AXis — ROS integration for CRA/CR/E6/Nova arms.
Data Collection & Teleoperation Toolkits
- Embodylink (DobotRX) — toolkit for robot data collection, annotation, model-training management, and teleoperation.
- dobot_xtrainer — teleoperation and data-collection control interface with minimal configuration.
Product Documentation
- DOBOT Pocket Go Product Introduction (PDF)
- DOBOT Pocket Go User Guide V1.0 (PDF)
- DOBOT PocketGo User Guide V1.0 (Wiki)
FAQ
What is DOBOT Pocket Go?
Pocket Go is a portable platform for collecting embodied AI demonstrations with handheld and robot-mounted gripper kits, cameras, pose tracking, and software.
Are robot arms included?
No. Compatible DOBOT Nova and CR robot arms are separate hardware and are not included in the listed Pocket Go package.
What cameras are listed?
The product materials list GoPro Hero 12 cameras (1/1.9" CMOS, 24.69 MP, 155-degree FOV, HyperSmooth 6.0) for the handheld and end-effector gripper kits.
Does Pocket Go include both a handheld and an end-effector kit?
Yes. The listed standard configuration includes one handheld gripper kit and one robot end-effector gripper kit, each with a camera.
What tracking method is listed?
The product materials list visual-inertial 6D pose tracking based on ORB-SLAM3.
What is the listed data-collection capacity?
The source lists up to 4,000 demonstrations per day. Actual throughput depends on the operator workflow, task, and compute environment.
What is the maximum gripper payload?
The listed maximum gripper payload is 2 kg. Confirm the task-specific safety margin before operation.
Is the Mini AI reasoning computer included?
No. The Mini AI reasoning computer is listed as optional.
Which operating systems are listed?
The product documentation lists Ubuntu 20.04 and Ubuntu 23.10 (Linux only). Review the user guide before preparing a deployment environment.
Does Pocket Go provide a developer SDK?
Yes. The Pocket Go software platform includes a secondary-development SDK and a Docker-based environment for custom research integrations. The Developer Resources section above links the DOBOT developer portal, official GitHub, and the Nova/CR robot SDKs used to deploy trained policies.
Which robot arms can run the collected data?
Collected data and trained models are compatible across the full DOBOT Nova and CR series (Nova 2, Nova 5, CR3, CR5, CR7). The arms are sold separately and are required for autonomous inference.
Where can I download the documentation?
The Product Documentation section links the individual Pocket Go product introduction and user guide PDFs, plus the OpenELAB wiki.
1. General Shipping Information
- We provide reliable shipping services with a tracking number for each order.
- Shipping addresses must be entered in English and should not contain special symbols, so that the courier system can recognize your location correctly.
- Please make sure your shipping address is accurate before placing your order. We ship strictly according to the address provided at checkout.
- In-stock orders are usually dispatched within 1 business day after order confirmation.
- Estimated delivery times refer to the period after dispatch and do not include order processing time, weekends, public holidays, customs inspection, or force majeure delays.
- If you need to cancel or modify your order, please contact us before the order is marked as “Shipped”. Once shipped, the order cannot be canceled or changed.
🚀 Need Faster Shipping?
If you require expedited shipping, please contact our customer support team at info@openelab.io for a customized quote based on your destination.
2. Shipping Rates & Options
Shipping rates are calculated based on order value, destination, and available shipping methods. The final available options will be displayed at checkout.
2.1 Germany Domestic Shipping
| Shipping Method | Order Value | Cost | Est. Delivery |
|---|---|---|---|
| Deutsche Post | €0.00 - €50.00 | €4.95 | 2-4 Business Days |
| Deutsche Post | Over €50.00 | Free | 2-4 Business Days |
|
DHL Paket (Faster Delivery) |
€0.00 - €50.00 | €6.95 | 1-3 Business Days |
|
DHL Paket (Faster Delivery) |
€50.00 - €100.00 | €2.00 | 1-3 Business Days |
|
DHL Paket (Faster Delivery) |
Over €100.00 | Free | 1-3 Business Days |
2.2 Selected EU Countries / Regions
Available EU shipping destinations are shown at checkout based on your shipping address. For orders shipped to selected EU countries or regions outside Germany, we use FedEx Regional Economy.
| Shipping Method | Order Value | Cost | Est. Delivery |
|---|---|---|---|
| FedEx Regional Economy | €0.00 - €100.00 | €7.95 | 3-5 Business Days |
| FedEx Regional Economy | Over €100.00 | Free | 3-5 Business Days |
EU orders may be fulfilled from our Munich warehouse or, when applicable, from our Shenzhen warehouse depending on inventory availability. For EU member states, OpenELAB covers applicable import duties and taxes under DDP service where required.
2.3 United States
| Region | Shipping Method | Order Value | Cost | Est. Delivery |
|---|---|---|---|---|
|
Continental U.S. (50 States) |
USPS Ground Advantage | €0.00 - €45.00 | €4.95 | 3-7 Business Days |
| USPS Ground Advantage | Over €45.00 | Free | 3-7 Business Days | |
| USPS Priority Mail | €0.00 - €45.00 | €16.95 | 1-4 Business Days | |
| USPS Priority Mail | Over €45.00 | €14.95 | 1-4 Business Days | |
|
Non-Continental U.S. (AK, HI, PR, etc.) |
USPS Ground Advantage | €0.00 - €60.00 | €6.95 | 5-9 Business Days |
| USPS Ground Advantage | Over €60.00 | Free | 5-9 Business Days |
Non-Continental U.S. regions include Alaska, American Samoa, Guam, Hawaii, the Marshall Islands, the Northern Mariana Islands, Palau, Puerto Rico, the U.S. Virgin Islands, and U.S. Armed Forces addresses.
2.4 International Destinations Outside the EU
For selected international destinations outside the EU, including Switzerland, the United Kingdom, and Norway, shipping rates are as follows:
| Order Amount | Shipping Cost |
|---|---|
| €0.00 - €300.00 | €19.95 |
| Over €300.00 | Free |
For non-EU destinations, import duties, taxes, and customs fees may be charged by the destination country and are the responsibility of the recipient.
3. Warehouses & Fulfillment
Our products may be stored in our Munich, Arlington, and Shenzhen warehouses. The actual shipping warehouse depends on product availability and destination.
3.1 Munich Warehouse
For products stored in our Munich warehouse, we use Deutsche Post or DHL Paket for domestic deliveries within Germany. For selected EU destinations outside Germany, we use FedEx Regional Economy.
3.2 Arlington Warehouse
For products stored in our Arlington warehouse, we use USPS or UPS for deliveries within the United States.
3.3 Shenzhen Warehouse
For pre-order items or products fulfilled directly from our Shenzhen warehouse, we arrange reliable international shipping based on destination and inventory status. For EU member states, OpenELAB covers applicable import duties and taxes under DDP service where required.
4. Inventory, Pre-orders & Split Shipments
- Inventory Status: Please check the product page for real-time stock information. If an item is out of stock in our local warehouses, it may be marked as “Pre-order”.
- Pre-order Fulfillment: Pre-order items may be shipped directly from Shenzhen or restocked to a local warehouse first before final delivery.
- Split Shipments: If your order contains both in-stock and pre-order items, we may ship them separately. No additional shipping fee will be charged for split shipments caused by our fulfillment arrangement.
5. Customs, Taxes & Delivery Issues
5.1 Customs and Taxes
For EU Member States:
Whether shipped from Germany or China, OpenELAB covers applicable import duties and taxes under DDP service where required. Customers in EU member states should not be charged additional import duties or VAT upon delivery.
For destinations outside the European Union, such as Switzerland, Norway, and the United Kingdom, import duties, taxes, and customs fees may be charged by local authorities upon delivery. These fees are the responsibility of the recipient.
5.2 Damaged, Delayed, or Lost Parcels
Please inspect your parcel upon delivery where possible. If you notice visible damage, please report it to the courier and contact us as soon as possible. This does not affect your statutory consumer rights.
If your parcel is delayed, lost, returned to sender, or the tracking information has not updated for an unusual period of time, please contact us at info@openelab.io. We will assist you in checking the shipment status with the courier.
5.3 Incorrect Address or Failed Delivery
If a parcel cannot be delivered due to an incorrect or incomplete address provided by the customer, refusal of delivery, or failure to collect the parcel, additional shipping or return costs may apply.
6. VAT
For orders within the EU, VAT is collected at checkout according to the applicable destination rules. No additional VAT should be collected upon delivery for EU member state orders covered under our DDP shipping arrangement.


