Open Low Cost Humanoid (WIP)

OpenLCH is an open-source ultra-low-cost humanoid robot designed for experimenting with machine learning methods for robot control.

[!WARNING] The project is currently in progress (V0.3.1)

We designed our robot in OnShape, trained the PPO model in Isaac Gym, and are now transfering it onto the physical robot. Check out our public roadmap for updates here.

Our goal is to build and deploy a large amount (20-30) of small humanoid robots to the physical world and create an affordable open-source platform for humanoid research and competitions. The robot design is inspired by Robotis OP3, while the initative is inspired by Alex Koch's robot arms.

We made the first version of this humanoid robot at a hackathon in 24 hours on 2024/08/31.

[!NOTE]
Interested in updates, contributing, or building your own mini humanoid? Let us know through our interset form here: https://forms.gle/AvDzMEFUYeVNtFvj6!

Mechatronics

CAD

OnShape: https://cad.onshape.com/documents/cacc96f8a7850b951e7aa69a/w/3a0a4ee9d8251956ba5e5e92/e/b92662619a7718ffa83530f2

URDF/MJCF: https://kscale.store/file/5b9b5eecb7ffcab1

BoM

Parts Assembly Guide

WIP, coming soon...

We're currently 3D printing and testing the hardware.

Software

Firmware

WIP, coming soon...

We're currently writing our firmware for sensors and controllers.

Runtime

WIP, coming soon...

We're currently working on robot control that will be written in Rust for performance and safety (and enjoyment). :)

Simulation

We're using the K-Scale simulation library (based on Isaac Gym) to simulate and training our robot.

Left: URDF Model, Right: Isaac Gym Training

NVIDIA Isaac (Humanoid) Gym

We use NVIDIA Isaac Gym to simulate, train, and test the robot for locomotion based on the K-Scale simulation library.

Link: https://github.com/jingxiangmo/sim/tree/master Docs: https://docs.kscale.dev/software/simulation/isaac

To test policy in MuJoCo, you can use the following command:

export MODEL_DIR=sim/resources

python sim/sim2sim.py --load_model examples/standing.pt --embodiment stompymicro

PyBullet

Currently the URDF model also support PyBullet using K-Scale OnShape library: https://docs.kscale.dev/software/onshape

ML

Locomotion

RL (PPO)

We use PPO to train the robot to stand and walk. The training is done in Isaac Gym with the K-Scale simulation and training library: https://github.com/jingxiangmo/sim/tree/master.

Manipulation

E-VLA (WIP)

Integration of E-VLA will be in V2. For more details, please refer to the E-VLA documentation.

K-Lang (WIP)

Integration of K-Lang will be in V2. For more details, please refer to the K-Lang documentation.

License

This project is licensed under the MIT License.

Contributors

Core contributors:

• Kelsey Pool - mechanical & design
• Denys Bezmenov - electrical & embedded
• Jingxiang Mo - electrical, software, & ML

Open Source Acknowledgment:

• Henri Lemoine - ML
• Advait Patel - ML

Hackathon Team

- **Kelsey Pool** - Mechanical design - **Denys Bezmenov** - Electrical and software control - **Jingxiang Mo** - Mechanical assembly, electrical, and software control - **Baaqer Farhat** - Mechanical assembly, software Acknowledgement: - **Jacob Zietek** - AI/ML and simulation help - **Saad Sharief** - Teleoperation help

Last updated: 2024/09/27