TheRobotStudio/SO-ARM100
Standard Open Arm 100
🤖 TheRobotStudio / SO‑ARM100 – Open‑source 2‑joint robot arm (SO‑100 / SO‑101)
What it is – A low‑cost, 3‑D‑printable robot arm that can be built from off‑the‑shelf servos and a small control board. The design is the successor to the original SO‑100 and adds a leader arm (the “101” version) with improved wiring and motors. The hardware is meant to work out‑of‑the‑box with the 🤗 LeRobot library, letting you run tele‑operation or reinforcement‑learning experiments on a physical platform.
📦 What you get
| Item | What it does |
|---|---|
| STS3215 servos (7 × 7.4 V) | Provide the joint actuation. Different gear ratios (1/345, 1/191, 1/147) give the required torque and speed for the follower and leader arms. |
| Motor control board (Waveshare) | A tiny USB‑C powered board that converts host commands into PWM signals for the servos. |
| Power supplies (5 V / 5 A or 12 V / 5 A for the optional higher‑torque motors) | Powers the servos and board. |
| 3‑D‑printed structural parts (STL files in the repo) | Base, motor mounts, arm links, clamps, and optional accessories (raised base, camera mounts). |
| Table clamps & screwdriver set (recommended tools) | Secure the arm to a work surface and assemble the screws. |
The README even lists a complete bill‑of‑materials with regional links (US, EU, China, Japan) and a rough cost of ~ $230 USD for a full two‑arm (follower + leader) setup.
🛠️ Building the arm
- Source or 3‑D‑print the parts – STL files are provided for common printer bed sizes (Ender‑3, Prusa Mini+, BambuLab, etc.). The guide includes printer settings (PLA+, 0.2 mm layer height, 15 % infill) and a gauge‑printing step to verify dimensional accuracy.
- Assemble – Follow the step‑by‑step assembly guide hosted on Hugging Face (
https://huggingface.co/docs/lerobot/so101). It walks you through installing the servos, mounting the control board, wiring the power supply, and tightening the clamps. - Optional hardware – The repo contains extra CAD files for things like a raised leader base or different camera mounts, letting you customise the platform for vision‑based tasks.
📚 Software integration
- LeRobot 🤗 – The arm is a first‑class device in the open‑source LeRobot library (the same library used for the popular LoCoBot and other research arms). After assembly you run the LeRobot SO‑101 tutorial which:
- Installs the Python package.
- Detects the USB‑C control board.
- Streams joint states and accepts position/velocity commands.
- Provides a ready‑made tele‑operation interface (e.g., joystick or VR controller) and example reinforcement‑learning scripts.
- End‑to‑end AI – Because the hardware speaks the same API as simulated arms, you can train policies in simulation and directly transfer them to the physical SO‑101, facilitating research on sim‑to‑real transfer, imitation learning, or embodied language agents.
🌐 Community & support
- Discord – A dedicated server (
https://discord.gg/ggrqhPTsMe) where builders share prints, troubleshoot wiring, and discuss AI experiments. - Vendors & kits – Multiple commercial partners (Robonine, PartaBot, ForgeMotion Labs, Seeed Studio, WowRobo, etc.) sell either parts‑only kits (you 3‑D‑print the frames) or fully assembled kits for those who prefer a ready‑made robot.
- Open contributions – The repo encourages pull‑requests for new vendor links, additional optional hardware, or improvements to the STL files.
🚀 Quick‑start checklist
- Pick a route – Build yourself from the BOM or order a kit.
- Print the STL files using the recommended printer settings.
- Assemble following the Hugging Face guide.
- Install LeRobot (
pip install lerobot) and run the SO‑101 tutorial. - Start experimenting – try the tele‑op demo, then move on to RL or vision‑based policies.
TL;DR
The SO‑ARM100 / SO‑101 repo is a genuine open‑source robotics project that delivers a cheap, 3‑D‑printable robot arm ready for AI research. It bridges hardware and software by providing all CAD files, a detailed BOM, assembly instructions, and seamless integration with the LeRobot library for tele‑operation and learning experiments.
Related
- Project
- Project
- Project
- Project
- Project