Dobot Dobot X-Trainer (Dobot X-Trainer)

Most embodied AI projects stall long before the model does. The scarce resource is rarely compute; it is clean, repeatable demonstrations of a person doing a physical task well. The Dobot Dobot X-Trainer (Dobot X-Trainer) tackles that bottleneck directly. An operator guides two lightweight master hands, two Dobot Nova2 arms mirror every movement, and cameras on each wrist and above the workspace record the result, so that a session of folding clothes or rinsing dishes becomes a dataset ready for training. It is an AI data collection and training robotic system designed for AGI research and practical AI projects, and it treats the human demonstrator as the most important instrument on the bench.

Robots Europa, based in Dublin, supplies the X-Trainer to universities, research institutes and corporate AI teams throughout the European Union and the DACH region. We prepare the quotation, coordinate delivery to your laboratory and remain your point of contact for support once the system is in use, so that procurement, import and after-sales questions are handled by one European partner. The wider range we carry from this manufacturer is listed in our Dobot category.

Background

DOBOT and the move into embodied AI

DOBOT built its reputation on educational robot arms and collaborative robots, and the Nova series from its cobot range provides the two slave arms at the heart of this system. In April 2024 the company released the X-Trainer AI Data Collection and Training Robotics System, presenting it as a tool to empower embodied AI agents. The launch reflects a wider shift across the robotics sector: as learning-based control has matured, the hardware that produces training data has become as significant as the robots that eventually run the trained policies. A research group can no longer rely on hand-written motion scripts alone; it needs a dependable way to capture how people actually handle objects.

Teleoperation as a data pipeline

The X-Trainer is organised around master-slave teleoperation. The operator moves the master hands, the Nova2 arms follow, and the system records joint angles, gripper state and camera images in step with one another. This is the foundation of imitation learning, in which a model learns a task by observing many demonstrations rather than by being programmed explicitly. The X-Trainer generates data formats suitable for Mobile ALOHA training, an open research approach to learning bimanual manipulation from human demonstration that has become a common reference point in the field. Around the hardware, DOBOT provides a complete SDK covering the full workflow of data collection, model training and autonomous inference, with detailed documentation that guides users through the entire data processing pipeline.

Technical Specifications

Specifications as published by the manufacturer; confirmed against the datasheet at quotation.

DOBOT has not published the dimensions, weight, power supply, power consumption, camera resolution, gripper type and stroke, operating temperature or host computer requirements; Robots Europa confirms these details with you at quotation.

How does the X-Trainer turn demonstrations into a trained model?

The value of a teleoperation rig lies in how faithfully the slave arms reproduce what the operator intends, and how cleanly that intention is recorded. The X-Trainer addresses fidelity first: the Nova2 arms achieve ±0.05mm repeat positioning accuracy, which suits precise experimental operations and sensitive data collection tasks, and the 25Hz end-to-end motion interface improves response speed by 150% compared with similar products while running more stably. With 625mm of reach per arm, extending to 1200mm in dual-arm mode, and 6-axis flexibility, the workspace is broad enough for tabletop household and handling tasks.

Recording is placed in the operator's hands, literally. Each master hand carries an unlock button, which locks or releases the master hand pose so that the demonstrator can pause without holding an awkward posture, and a record button that starts and stops data capture without anyone reaching for a keyboard. The end cameras on both arms and the global camera on the mounting platform supply the visual context that learning-based policies depend on.

Once the demonstrations are captured, the SDK carries them through model training and on to autonomous inference, where the trained policy drives the Nova2 arms without an operator. Routine scenario training, such as dish-washing or folding clothes, can be completed in 2H, a reduction of 70%. For teams that prefer their own tooling, open APIs expose master and slave joint angles, gripper control and camera RGB and depth data, so custom AI models can be integrated without working around a closed system.

What makes the X-Trainer suitable for long data collection sessions?

A dataset of useful size demands hours of demonstration, and two things decide whether those hours are productive: operator comfort and equipment safety. On comfort, the master hands keep hand loads under 400g and use 3D printed monolithic joints that reduce the number of parts and keep the experience light. The unlock function means the operator can rest mid-task without losing the pose.

On safety, the system automatically de-synchronises the master and slave arms on power loss or alarms, detects when a master hand has been dropped, and uses low-speed master-slave synchronisation to prevent collisions. The Nova2 cobot itself offers 5-level collision detection and ISO15066 certification for safe human-robot interaction. The workspace is padded with flexible EVA foam, which cushions accidental drops of the end-effector and minimises the risk of damage to the gripper and arm. Together these measures allow inexperienced students and seasoned researchers alike to work close to the arms with confidence.

Who is the Dobot X-Trainer for?

The clearest fit is the AGI and embodied AI research group that needs its own demonstration data rather than relying on public datasets. Such teams benefit from the complete pipeline from capture to autonomous inference and from data formats suitable for Mobile ALOHA training, which shortens the path from a new idea to a first trained policy.

University departments teaching robotics, machine learning or human-robot interaction form a second segment. The documented SDK guides students through each stage of the data processing pipeline, and the safety design lets them operate the system hands-on rather than watching from a distance.

Corporate research and development teams exploring service, household or light handling tasks are a third audience. Here the open APIs matter most: they let engineers plug proprietary models into the system and use the X-Trainer as a testbed before committing to production hardware.

The system is offered in two configurations, the DOBOT X-Trainer and the DOBOT X-Trainer (Lightweight Base), and the Robots Europa team will help you choose the one that best suits your laboratory and its space.

What is the Dobot X-Trainer used for?

1. Bimanual imitation learning — recording two-handed demonstrations that teach a policy to coordinate both Nova2 arms.
2. Household task research — capturing and training routine scenarios such as dish-washing or folding clothes.
3. Dataset creation for Mobile ALOHA training — producing data in formats suitable for this widely used research approach.
4. Custom model integration — connecting in-house AI models through open APIs for joint angles, gripper control and camera data.
5. Vision-based manipulation studies — combining RGB and depth images from the end and global cameras with arm motion data.
6. Teaching embodied AI — giving students a guided, safe path from demonstration through training to autonomous inference.

Pricing and Availability

Robots Europa supplies the Dobot X-Trainer on quotation. Tell us about your research objectives, the configuration you are considering and your delivery location, and we will prepare an offer matched to your project. Delivery, customs and support are arranged with the quotation, so your institution deals with a single contact from order to commissioning. To begin, use the quotation request available on this page.

Frequently Asked Questions

What is included with the Dobot X-Trainer?

The core components are Teleoperated Master Hand *2, Dobot Nova2 *2 (including end gripper *2, end camera *2) and a removable mounting platform *1 with global camera *1. A complete SDK covering data collection, model training and autonomous inference is also provided.

How precise are the arms?

The Nova2 arms achieve ±0.05mm repeat positioning accuracy. Each arm reaches 625mm, extending to 1200mm in dual-arm mode, with 6-axis flexibility.

Can I use my own AI models with the system?

Yes. Open APIs give access to master and slave joint angles, gripper control and camera RGB and depth data, enabling integration with custom AI models and research projects.

How quickly can a routine task be trained?

Routine scenario training, such as dish-washing or folding clothes, can be completed in 2H. The training time is shortened by 70%.

What safety features does the X-Trainer offer?

It de-synchronises the master and slave arms on power loss or alarms, detects a dropped master hand and uses low-speed synchronisation to prevent collisions. The Nova2 cobot adds 5-level collision detection and ISO15066 certification, and the workspace is padded with EVA foam.

Which cameras does the system use?

Each Nova2 arm carries an end camera, and the removable mounting platform holds a global camera. The open APIs provide access to camera RGB and depth data.

Where can I buy the Dobot X-Trainer and what does it cost?

You can buy the Dobot X-Trainer from Robots Europa, which supplies customers across the EU and the DACH region from Dublin. Pricing is provided by individual quotation, which you can request on this page.

Summary

The Dobot X-Trainer starts from a simple observation: in embodied AI, the quality of the demonstrations sets the ceiling for the model. By pairing light master hands with two precise Nova2 arms, wrist and global cameras, layered safety measures and a complete SDK running from capture to autonomous inference, it turns a skilled operator's time into training data with little friction. Open APIs keep the door open for custom models, and data formats suitable for Mobile ALOHA training connect it to established research practice. For laboratories and AI teams in the EU and the DACH region looking to build their own demonstration datasets with Robots Europa as their supplier, this page is the starting point.