XLeRobot: A $660 Dual-Arm Mobile Home Robot for Embodied AI Research

What XLeRobot Is and Who It Is For

XLeRobot is a dual-arm mobile home robot designed for affordability and community research. The basic configuration costs approximately $660 USD in parts for US buyers, using a personal laptop and a single RGB head camera. Assembly is stated to take under four hours. The cost table in the README excludes 3D printing materials, tools, shipping, and taxes, so the real entry cost is higher depending on a builder's access to a printer.

The project was created by Gaotian (Vector) Wang, a CS graduate student at Rice University's RobotPi Lab, as a personal research platform to test theories from his work on robust object manipulation and sim-to-real transfer. XLeRobot builds on top of several prior open-source projects: LeRobot from Hugging Face, SO-100 and SO-101 from TheRobotStudio, Lekiwi from SIGRobotics-UIUC, and Bambot.

The target users are researchers, students, and hobbyists who want a physical platform for embodied AI experiments, particularly for tasks like household manipulation. The README notes that users new to programming should spend at least a day getting familiar with Python, Ubuntu, and GitHub before starting.

Cost Breakdown and Purchase Options Across Regions

The README publishes a cost table for the base configuration and optional upgrades across four regions. The basic configuration, which requires using your own laptop and includes a single RGB head camera, costs approximately $660 in the US, €680 in the EU, ¥3999 in China, and ₹87000 in India.

Optional upgrades stack on top of the base price. A stereo dual-eye RGB head camera adds $30 (or €30 / ¥199 / ₹6550). Adding a Raspberry Pi adds $79 (or €79 / ¥399 / ₹7999). Upgrading to a RealSense RGBD depth camera adds $220 (or €230 / ¥1499 / ₹35726).

For developers who prefer not to source individual components, a pre-packaged developer assembly kit is available from Wowrobo. As of September 9, 2025, the kit was listed at ¥3699 on Taobao for China buyers and $579 worldwide. The kit excludes battery and the IKEA cart used as the mobile base. The README notes that this is a developer kit only and that builders should review the documentation and repository before purchasing.

The dual-wheel version released December 2, 2025, provides a more stable differential-wheel base at a comparable or lower price compared to the original base design.

Hardware Assembly: BOM, 3D Printing, and the Build Sequence

The README outlines four steps for new builders. First, buy the parts using the Bill of Materials (BOM) linked at xlerobot.readthedocs.io/en/latest/hardware/getting_started/material.html. Second, 3D-print the required structural components using files documented at the 3D printing page. Third, assemble the robot following the assembly guide. A hardware assembly video tutorial from WOWROBO was released on September 22, 2025, available on both YouTube and Bilibili. Fourth, install the software to get the robot moving.

The repository top-level organizes content into hardware/, simulation/, software/, web_control/, XLeVR/, and docs/ directories. Hardware-specific files and configurations reside in hardware/. The docs/ directory is the primary reference for assembly instructions, with the canonical version hosted at xlerobot.readthedocs.io.

The README warns that some documentation may be evolving and that builders should check the documentation website before beginning, since the assembly guide and BOM can change as the project advances.

Software Stack: Simulation, Bluetooth Control, and VLA Training

The software layer covers three main modes: simulation, physical control, and machine learning training.

For simulation, the repository provides a MuJoCo environment with updated URDFs and control scripts. The documentation states that users can get started with simulation in about 15 minutes. Input methods include Quest3 VR headset, keyboard, Xbox controller, and Switch joycon. A browser-based MuJoCo plus 3DGS simulation environment launched on January 29, 2026, at vector-wangel.github.io/MuJoCo-GS-Web/, allowing browser-based interaction without local setup.

For physical control, XLeRobot supports keyboard, Xbox controller, and Switch joycon over Bluetooth. The documentation states that this requires no Wi-Fi and has zero latency. As of July 30, 2025, the documentation covers controlling the robot remotely anywhere via Bluetooth.

For machine learning, a Vision-Language-Action (VLA) training tutorial was added on November 11, 2025. The README describes it as a temporary version, with an official version stated as coming soon. This means the VLA training workflow is documented but not yet finalized.

The software/ directory holds the control scripts, and the simulation/ directory holds simulation-specific assets. The README notes that the codebase requires basic familiarity with Python and Ubuntu to work with.

Where XLeRobot Falls Short

Several real limitations affect XLeRobot as a platform. The README itself warns new users that some assembly and documentation may be in flux. The VLA training tutorial is described as temporary. The cost table explicitly excludes 3D printing, tools, shipping, and taxes, meaning the stated $660 starting point understates total cost for most builders.

The project targets Ubuntu as the primary operating system for the software stack. Windows native support for running the control software is not documented. Users who do not have access to a Linux environment will face an additional setup burden.

The hardware documentation does not cover repair or replacement for individual servo motors or structural parts after wear. The README notes that not all parts of the design may be finalized; the dual-wheel base variant released in December 2025 represents a significant change to the base design from earlier versions.

For users who want a reliable, production-quality household robot with support contracts and guaranteed uptime, XLeRobot is not designed for that use case. It is explicitly a developer and research platform.

XLeRobot vs. LeRobot from Hugging Face

XLeRobot builds directly on LeRobot (huggingface/lerobot), and the two serve different purposes despite the shared lineage. LeRobot is a software framework: it provides datasets, pretrained models, and simulation environments for robot learning, but it does not specify a hardware design at a particular price point. A researcher using LeRobot alone would need to separately source or design their own robot hardware.

XLeRobot takes the LeRobot software stack and pairs it with a specific, community-documented hardware design that builders can source from off-the-shelf components. The documentation provides a BOM, 3D printing files, and an assembly guide. The $660 price point and the under-four-hour assembly claim are specific to the XLeRobot hardware design, not to LeRobot in general.

Teams that want only the learning and simulation software without a specific hardware design can use LeRobot directly. Teams who want a documented hardware build with community support and a concrete parts list would use XLeRobot as the hardware companion to LeRobot's software tools.

Editorial conclusion

XLeRobot suits researchers and hobbyists who want a low-cost dual-arm robot platform and are comfortable with Ubuntu and Python. It is the wrong choice for production deployments or teams expecting polished documentation across all features. Before starting, check the current BOM at xlerobot.readthedocs.io to confirm part availability and pricing in your region, since the stated costs exclude 3D printing materials, tools, and shipping.

Frequently asked questions

Official sources

1. Issues
2. License: Apache-2.0
3. Project website
4. README
5. Vector-Wangel/XLeRobot on GitHub