Connecting to sources…

How to build XLeRobot: the whole process

77 steps to build XLeRobot, from buying parts, printing and assembly to software, calibration and training, with the bill of materials, shop links and the official sourc…

Basic build (laptop plus single RGB head camera) about $660 or CN¥3,999, excluding printing, tools, shipping and tax. Options: stereo camera +$30, Raspberry Pi +$79, RealSense +$220

Buy parts

  • Compare regional material costs

    Review the US, EU, CN, and IN prices in the bill of materials. The total excludes 3D printing, tools, shipping, and taxes.

  • Buy printed parts or assembled kit

    Wowrobo sells 3D-printed parts or pre-assembled kits for SO101 arms.

  • Check BOM against existing arms

    If you already have SO101 arms or Lekiwi, use this BOM selectively. Check the SO100 and Lekiwi repositories for potentially better-priced options.

  • Prepare missing assembly tools

    Purchase the assembly tools listed in the table if you do not already have them. Skip this section if you already have all of them.

  • Obtain the feature parts

    Select and purchase the required feature parts and quantities from the feature-parts table.

  • Obtain the electronic components

    Select and purchase the required electronic components and quantities from the electronics table.

  • Get the servo version or buy parts

    You can purchase this version directly from wowrobo, or buy the required parts yourself besides the 3D printed parts and servo motors.

  • Start with Bracket Bot

    Use the Bracket Bot robot as the basis for the brushless-motor version.

  • Buy a pre-assembled kit

    Pre-assembled kits are also available for SO101 arms.

  • Obtain an IKEA cart

    Obtain an IKEA cart to use as the robot’s cart base.

  • Purchase the bill of materials

    Purchase the parts listed in the BOM.md bill of materials. The total cost excludes 3D printing, tools, shipping, and taxes.

  • Buy the robot parts

    Follow the linked Bill of Materials to buy the parts.

Print and make

  • Download the XLeRobot 0.3.0 file

    Download the XLeRobot 0.3.0 .3mf file for 3D printing.

  • Position models and add supports

    Place and orient the models yourself, and add supports to ensure the best printing quality.

  • Print soft Finray fingers in TPU95A

    The .3mf file contains soft Finray fingers that must be printed with TPU95A.

  • Choose the original SO101 finger file

    If you want the original SO101 finger, download the original SO101 .stl file.

  • Choose the head camera connector version

    The .3mf contains two versions of the single RGB head camera connector and shell: one for a webcam and RealSense, and one for the cheaper camera used with the hand camera.

  • Print the base support washers

    No adjustment to the base connector is needed this time; print the washers.

  • Choose the D435 gimbal pitch holder

    Use Gimbal_Pitch_Holder_d435.stl for an Intel RealSense D435 or similar RGBD camera.

  • Choose the D455 gimbal pitch holder

    Use Gimbal_Pitch_Holder_d455.stl for an Intel RealSense D455 or similar RGBD camera.

  • Download the 3D Models

    Open the XLeRobot hardware directory and download the 3D models needed for printing.

  • Prepare and Slice Models

    Place and orient the models yourself in the slicing software and add supports; adjust infill and print speed for the desired balance of strength, efficiency, and detail.

  • Print the Arm Base

    Print the arm base parts using STL files and PLA filament.

  • Print the Neck and Head

    Print the neck and head parts using STL files and PLA filament.

  • Print the Wheel Base

    Print the wheel base parts using STL files and PLA filament.

  • Download and print the STL file

    Download XLeRobot_0_4_0_extra.stl from the linked file and use it to make the required 3D printed parts.

  • Get the IKEA cart connector

    Obtain the STL file for the connector that attaches to the IKEA cart from the provided link.

  • Print the robot parts

    Print the required parts following the 3Dprint.md guide.

  • Print the robot components

    Follow the linked 3D printing guide to print the robot components.

  • Modify and print the connectors

    If the fit is incorrect, slightly adjust the connectors’ 3D model’s z-axis scale in the slicer, keep the xy-axis scale unchanged, and print them again.

Assemble

  • Assemble the robot from scratch

    Assembly takes an estimated 2–4 hours from scratch, or 1–2 hours with assembled SO100/SO101 arms. Wowrobo’s official assembly videos and Greg’s Tech tips are provided.

  • Check the SO101 arms

    If you already have two assembled SO101 arms with configured motors, skip this section.

  • Prepare the Lekiwi base plate

    If you already have a Lekiwi base, detach the battery, servo mounts, and similar items. Keep the three mounted motors with wheels and their wiring on the base plate.

  • Select the correct base plate

    Choose the correct plate because each plate has its own fixed component order.

  • Flip the cart for assembly

    Flip the cart over before performing the base assembly.

  • Install the twelve M3 screws

    This new hardware version fits the IKEA cart metal mesh, and all twelve M3 screws should fit easily.

  • Assemble the Top Base

    Complete the top base assembly according to the assembly image in the document.

  • Assemble the Robot Head

    Complete the head assembly according to the assembly image in the document.

  • Complete Top Base Wiring

    Complete all wiring and cable management for the top base, and put the Raspberry Pi in its case before clamping the top base into the IKEA cart.

  • Place the Battery and Connect Power Last

    Place the battery. To protect the motor control board, connect the power cables last, and always disconnect the power cables when plugging or unplugging other cables.

  • Fit the Base into the Cart

    After completing the top base wiring and cable management and putting the Raspberry Pi in its case, clamp the top base into the IKEA cart.

  • Assemble the dual-wheel version

    Assemble the dual-wheel version using the assembly illustrations and official assembly video.

  • Secure the cart connector

    Use the required M8x20 screws to secure the connector to the IKEA cart.

  • Assemble XLeRobot

    Complete the XLeRobot assembly following the Assembly.md guide.

  • Insert washers into the IKEA cart

    Insert the printed washers into the IKEA cart for the best base support.

  • Assemble the robot

    Follow the linked assembly guide to assemble the robot.

Software

  • Install LeRobot

    Follow the official Installation Guide to install LeRobot.

  • Clone the XLeRobot repository

    Clone this repo, then change into the cloned directory.

  • Install the shared module

    Install the shared module with the SO101 kinematics helpers.

  • Install 3-omniwheel base plugin

    Install the robot plugin for the 3-omniwheel base.

  • Install 2-wheel differential drive plugin

    Install the robot plugin for the 2-wheel differential drive base.

  • Install mecanum-wheel base plugin

    Install the robot plugin for the mecanum-wheel base.

  • Install VR teleoperation plugin

    Install the VR teleoperation plugin.

  • Enable host and client for Raspberry Pi

    If you build based on RaspberryPi, uncomment xlerobot_host and xlerobot_client in the plugin’s __init__.py under software/src/robots/.

  • Install LeRobot plugins

    Install `src/robots/` and `src/model/` into LeRobot as plugins.

  • Run keyboard joint control

    Use keyboard joint control. The control codes should be compatible with both SO100 and SO101, even when the code name still uses so100.

  • Run keyboard end-effector control

    Control the end effector with the keyboard using analytical inverse kinematics.

  • Run dual-arm keyboard end-effector control

    Control the end effectors with the keyboard using analytical inverse kinematics in a dual-arm setup.

  • Learn the basic development environment

    Spend at least one day becoming familiar with basic Python, Ubuntu, and GitHub.

  • Install and run robot software

    Install the software and get the robot running using the Software.md guide.

  • Reproduce the Remote-Control Demo

    Connect one arm to the Raspberry Pi and the other to the desktop to reproduce Demo 0.1.0 with remote control.

  • Install dependencies

    Run the requirements.txt dependency installation command in the XLeVR directory; pip install -e is not needed.

  • Run the script from another folder

    Copy vr_monitor.py to any directory, make sure the XLEVR_PATH variable at the top of the script points to the correct XLeVR main directory, and install the dependencies as described above.

  • Install and run the software

    Follow the linked software guide to get the robot moving.

  • Grant Linux serial permissions

    On Linux, run sudo chmod 666 /dev/ttyACM0 first; Windows and Mac do not need this step.

Calibrate and test

  • Configure motors with Bambot

    Use Bambot because the official lerobot codebase currently does not support motor configuration beyond the arm.

  • Test the base-to-cart fit

    Place the Lekiwi base with its connectors under the cart and check that it provides enough pressure while all four cart wheels still touch the ground.

Data and training

  • Run the VR monitor directly

    From the XLeVR directory, run vr_monitor.py, open the HTTPS address shown in the terminal in the VR headset browser, and move the VR controllers to print real-time control data in the terminal.

  • Access VR data in code

    Create a VRMonitor instance and call initialize(), then use get_left_goal_nowait() and get_right_goal_nowait() in an event loop or callback to read the left and right control goals. The returned ControlGoal objects contain position, orientation, and other information.

  • Inspect ControlGoal data fields

    The main ControlGoal fields are arm, mode, target_position, wrist_roll_deg, wrist_flex_deg, gripper_closed, and metadata, covering device type, control mode, 3D position, wrist angles, gripper state, and additional controller information.

  • Run YOLO hand-camera tracking

    Run `simple_so100_yolo_ee_control.py` to use YOLO for object detection and segmentation and for real-time tracking with the hand camera.

Use it

  • Ensure the server is running and reachable

    Before opening the UI, ensure the server is running and reachable.

  • Serve the client dev server or production build

    Serve the client development server or production build for browser access.

  • Open the default dev URL and verify the dashboard

    Open the default development URL, http://localhost:5173, and verify that the dashboard loads.

  • Monitor terminal logs for telemetry and errors

    Monitor terminal logs for telemetry, errors, or reconnect issues to help with troubleshooting.

  • Add Optional Visual Upgrades

    If you want a better visual outlook, you can install these parts on XLeRobot.

  • Use plugin types in LeRobot commands

    --robot.type=xlerobot / xlerobot_2wheels / xlerobot_mecanum and --teleop.type=xlerobot_vr now work on any standard LeRobot CLI command (lerobot-teleoperate, lerobot-record, …).

Common problems

  • Motor IDs, scanning, and firmware updates prevent arm configuration #85 #38 #70
  • Hardware fit, retention, and power prevent reliable robot setup #45 #25 #105 #140
  • Repository download, instructions, and model assets are missing or invalid #30 #56 #131
  • Example scripts fail from missing modules and stale simulation references #89 #108 #80
  • YOLO examples fail to track objects or load models correctly #72 #76
  • Keyboard, VR, and web controls fail to connect #52 #87 #95
  • RealSense camera setup fails during device testing #17
  • VR teleoperation data collection crashes after starting successfully #119

Datasets

Parts

  • Two SO‑101 arms and a LeKiwi-style base (built on SO‑101, LeKiwi and Bambot)
  • IKEA cart and battery (not included in the developer kit)
  • RGB head camera; stereo or RealSense depth optional
  • A laptop or Raspberry Pi as the controller

Tools

  • 3D printer
  • Tool list not stated

Skills

  • Python, Ubuntu, Git and pip
  • Optional: MuJoCo simulation and VLA training

Software

  • LeRobot family
  • Keyboard, Xbox, Joy‑Con and Quest 3 teleoperation
  • MuJoCo simulation and RL environment

Documents