YOU CAN 3D-PRINT YOUR OWN WALKING STAR WARS BD-X DROID - FULLY OPEN-SOURCE
A 42 cm tall Star Wars BD-X droid that walks using a reinforcement-learning policy trained in simulation and running on-device, no cloud.
RoboticsAI
- difficulty
- ●●●●●
- time
- several weekends
- license
- Apache-2.0
- repo
- repo ACTIVE3,402 stars
●●●●● · several weekends · Apache-2.0 · 3,402 stars · repo ACTIVE
WHAT YOU’LL NEED
- 3D printer + filament — printable parts — files are in the repo
- Electronic parts — full list with part numbers in the repo BOM
- Dev board / microcontroller — runs the project firmware
- Soldering kit + basic tools — one-time buy, serves every future build
Partner
COMPAREE VERDICT
This is a genuinely impressive robotics project — a miniature BD-X that walks with learned gaits, runs a local language model, and is fully printable and reproducible — but it sits at the top end of difficulty and cost for home builds. The 400 dollars target is optimistic: servos dominate the budget and real-world builds report closer to 500-600 dollars once you include filament, bearings, screws, and the Raspberry Pi. The mechanical assembly is manageable if you have built articulated prints before, but getting stable walking is the real challenge. The pretrained policy works on the reference hardware, but any variation in servo calibration, print shrinkage, or weight distribution will require retuning or retraining the RL model, and that means working in Python with MuJoCo and understanding PPO hyperparameters. If you are comfortable with that workflow and want a walking robot that you can extend with on-device AI, this is one of the most complete open-source legged platforms available. If reinforcement learning is new territory, expect a steep learning curve before it walks reliably. The one thing most likely to go wrong: the sim-to-real gap — the policy trained in simulation does not transfer cleanly to hardware without careful tuning, and debugging gait instability without RL experience is frustrating.
IN THE REPO
NOT IN THE REPO
- —All STL files, full BOM with vendor links, complete PPO training pipeline, and assembly documentation are in the repository.
- —The RL policy is trained in MuJoCo simulation and exported to run on the Raspberry Pi 5 on-device.
- —The BOM targets under $400 but real builds cost more — the 10 Dynamixel XL430-W250-T servos alone run $240-280 depending on sourcing.
- —No commercial restrictions: Apache-2.0 licence throughout.
- —Active development with frequent commits; the I/O 2026 Gemma demo is documented in a separate branch.
- —You will need to train or tune the walking policy yourself unless you use the pretrained weights — the README links to the training guide but warns that sim-to-real transfer is sensitive to print tolerances and servo tuning.
Can I build this?
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
Partner · KickstarterHeyGears G1: 10M+ colors and transparent parts in one print, plus UV printing on flat objects. Figures, parts and labels — no painting.
Videos
Walking Star Wars Droid Prototype
No official build video in the repository; community walkthroughs exist but are not linked in the README.
Gallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Read the main README and the assembly guide (The README links to detailed assembly instructions and the BOM spreadsheet.)
- 2.Source the Dynamixel servos and Raspberry Pi(The 10 XL430-W250-T servos are the single biggest cost and lead time — check Robotis distributors and expect 1-2 weeks shipping.)
- 3.Print the parts and assemble the body(Follow the assembly PDF; print tolerances matter for gait stability, so calibrate your printer first.)
- 4.Flash the Pi, install dependencies, and test servo communication(The software setup guide walks through installing the Python environment and connecting to the Dynamixel bus.)
- 5.Load the pretrained policy or train your own(Pretrained weights are provided; if you modify the hardware or want custom gaits, follow the MuJoCo training guide.)
Resources
Documentation, files and community threads for this build — we link straight to the original sources and never rehost the creator’s files.
Partner · KickstarterCORE POWER BANK S: 0.2" thin, 2.8 oz, 5000 mAh with 22.5W fast charging. Semi-solid battery, carbon fiber, IPX7 waterproof.
KNOWN ISSUES
- The $400 BOM is a best-case number — real builds run $500-600, and international shipping on servos can add another $50-100.
- Sim-to-real transfer is sensitive: the pretrained policy may not walk stably if your prints shrink differently or servos are not calibrated to the same zero positions.
- Dynamixel servos require a specific USB adapter (U2D2 or similar) — a generic FTDI cable will not work.
- The repository assumes you are comfortable with ROS-style Python packages and command-line workflows; there is no GUI and no one-click installer.
- Reinforcement learning tuning is a research skill — if the gait does not transfer, debugging requires understanding PPO reward shaping and domain randomization, not just tweaking config files.
- No official assembly video, so expect to cross-reference the PDF with photos in the issues and Discord.
Can I use cheaper servos?
Not directly — the kinematics, torque curves, and Dynamixel protocol are baked into the design and the trained policy. Switching to analogue servos would require a full mechanical and software redesign.
Do I have to train the walking policy myself?
No, pretrained weights are included and work on the reference hardware. You only retrain if you modify the design or want different gaits.
Will it work without the Raspberry Pi 5?
A Pi 4 will run the base walking controller, but the Gemma language model demo requires the Pi 5 for acceptable inference speed.
Is there a kit or can I buy a pre-assembled one?
No — this is a DIY-only project. You source, print, and assemble everything yourself.
How long does the battery last?
The repository does not specify a reference battery, but community builds report 20-40 minutes of walking on a 2S LiPo depending on gait speed and load.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
3,392 stars and a walking robot trained entirely in simulation — how tight were your print tolerances when the policy finally transferred to hardware?
Antoine Pirrone
Antoine built Open Duck Mini as a personal robotics platform and showed it running Google's Gemma model locally at I/O 2026. The project grew out of his work in reinforcement learning for legged robots and is actively maintained.
DISCLAIMER
- Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
- Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
- Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.

CompareeTEAM25d agoedited
Practical notes from our verification: the repository is under active development with commits every few days, the I/O 2026 Gemma demo lives in a separate feature branch, and the pretrained walking policy is committed directly as a .zip checkpoint. The BOM spreadsheet links to Robotis and AliExpress but does not include tax or shipping, which is where the $400 estimate breaks — real-world builds we found in Issues report $520-650 all-in depending on region. The mechanical CAD is parametric Fusion 360 exports, so tweaking dimensions means editing the source files and re-exporting. No official assembly video, but the PDF is detailed and the community has filled the gaps on Discord.