 |
 |
 |
We are making a miniature version of the BDX Droid by Disney. It is about 42 centimeters tall with its legs extended. The full BOM cost should be under $400 !
This repo is kind of a hub where we centralize all resources related to this project. This is a working repo, so there are a lot of undocumented scripts :) We'll try to clean things up at some point.
The gait is getting better !
open_duck_mini_real_demo.mp4
open_duck_mini_v2_real_back_push_smaller.mp4
Update 02/04/2024: You can try two policies we trained : this one and this one I run with the following arguments : python v2_rl_walk_mujoco.py --onnx_model_path ~/BEST_WALK_ONNX_2.onnx -p 32 --commands --cutoff_frequency 40
Update 15/03/2025: join our discord server to get help or show us your duck :) https://discord.gg/UtJZsgfQGe
Update 07/02/2025: Big progress on sim2real, see videos above :)
Update 24/02/2025: Working hard on sim2real !
Update 07/02/2025 : We are writing documentation on the go, but the design and BOM should not change drastically. Still missing the "expression" features, but they can be added after building the robot!
Update 22/01/2025 : The mechanical design is pretty much finalized (fixing some mistakes here and there). The current version does not include all the "expression" features we want to include in the final robot (LEDs for the eyes, a camera, a speaker and a microphone). We are now working on making it walk with reinforcement learning !
See this document for getting from a onshape design to a simulated robot in MuJoCo (Warning, outdated. Has not been updated in a while)
We are switching to Mujoco Playground, see this repo
open_duck_mini_v2-2025-02-18_18.35.34.mp4
We are still working with AWD too
open_duck_mini_v2-2025-01-19_20.42.31.mp4
ref_motion_demo.mp4
See this repo
We used Rhoban's BAM
https://docs.google.com/spreadsheets/d/1gq4iWWHEJVgAA_eemkTEsshXqrYlFxXAPwO515KpCJc/edit?usp=sharing
Chinese: https://zihao-ai.feishu.cn/wiki/AfAtw69vRigXaRk5UkbcrAiLnJw?from=from_copylink
Chinese: https://zihao-ai.feishu.cn/wiki/space/7488517034406625281
See print_guide.
See assembly guide (incomplete).
This repo contains the code to run the policies on the onboard computer (Raspberry pi zero 2w) https://github.com/apirrone/Open_Duck_Mini_Runtime
When we get a very nice stable and robust walk that transfers to the real robot, we'll provide it here (you can ask for a checkpoint on the discord too if you want to try).
If you want to train your own policies, and contribute to making the ducks walk nicely, see this document
- Centralize the URDF/MJCF. Separate repo ? Menagerie ?
Thanks a lot to HuggingFace and Pollen Robotics for sponsoring this project !