Bimo is a 45 cm tall bipedal robot created to democratize bipedal robotics research in a compact platform. Get started with pre-built kits or customize the design. Includes complete CAD files (coming soon), RP2040 firmware, Isaac Lab training environment, and a Python API for real-time control and model deployment
Product Page • Discord Community • Roadmap
- Fully Open Source: CAD files (coming soon), firmware, and code. Build, modify, and improve.
- Sim-to-Real Ready: Train policies in Isaac Lab, deploy directly on hardware.
- Baseline Walking Model: Deployed directly from simulation, without adaptation.
- Community-Driven: Share your implementations and improvements with the community.
- Two Configurations: Available as a fully assembled SLS kit or a DIY (FDM) printable edition.
- Fast Training: Vectorized Isaac Lab environments to train policies in <20 minutes (depends on hardware).
| Feature | Specification |
|---|---|
| Height | 45 cm |
| Weight | ~1.6 kg |
| Actuators | 8 servo motors (STS-3215) |
| Sensors | BNO08x 9-DOF IMU, 4× TOF distance sensors, 2x 180ºFOV Cameras |
| Microcontroller | Custom PCB, based on the RP2040 |
| Communication | Direct through USB |
| Power | 12V Includes adapters for different power supplies |
| Compute | On-board MCU, offboard inference via Python API (compatible with SBCs) |
| Design | Hip-head biped for simpler mechanics, includes empty head volume for custom hardware |
| Control Loop | 20 Hz |
Use Isaac Lab + RSL-RL to train bipedal behaviors in simulation.
cd IsaacLab/
./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/train.py --task Bimo --num_envs 2048 --headlessExport trained policies as ONNX and run inference on your robot in real-time.
Control your robot with Python, no ML required.
from bimo import Bimo, BimoRoutines
bimo = Bimo()
bimo.initialize()
routines = BimoRoutines()
routines.perform(bimo, "stand")
routines.perform(bimo, "sit")
# Or command servo positions directly
bimo.send_positions([-30, -30, 0, 0, 60, 60, 30, 30]) # DegreesWARNING: Always double-check manual positions! The above example sends a standing pose as an example AFTER sitting down, which would make the robot launch itself backwards!
- Add new routines or behaviors and submit to the community
- Improve sim-to-real transfer
- Optimize hardware design
- Create new Bimo based projects
the-bimo-project/
├── README.md # Main project README
├── LICENSE # Apache 2.0
├── .gitignore
│
├── BimoAPI/ # Python control library
│ ├── setup.py # Package setup (mekion-bimo)
│ ├── README.md # BimoAPI documentation
│ ├── CHANGELOG.md # BimoAPI version history
│ ├── bimo/
│ │ ├── __init__.py
│ │ ├── bimo.py # Core Bimo control class
│ │ └── routines.py # Pre-programmed behaviors
│ └── examples/
│ └── api_example.py # Complete inference example
│
├── IsaacLab/ # RL training environment
│ ├── README.md # Training guide
│ └── bimo/
│ ├── __init__.py
│ ├── bimo_config.py # Robot & actuator configuration
│ ├── bimo_task_env.py # Full Isaac Lab task environment
│ ├── agents/
│ │ ├── __init__.py
│ │ └── rsl_rl.py # RSL-RL PPO hyperparameters
│ └── assets/
│ └── Bimo.usd # USD robot model
│
└── MCU/ # Microcontroller firmware
├── README.md # MCU protocol & setup
├── CHANGELOG.md # MCU version history
└── micro_bimo.ino # RP2040 firmware (Arduino IDE)
Q1-Q2 2026:
- Community development
- Progressive Bimo updates
- CE/FCC Certification
Late Q2 2026:
- First kits ship: SLS, DIY options available
- Manuals, DIY assembly tutorials, CAD models
Q3 2026:
- R&D Team Constitution
- Project Maintenance & Development
- Contribution setup
| Organization | Details |
|---|---|
 |
EOI: Escuela de Organización Industrial |
 |
JLCCNC: High quality CNC machining services |
All code and CAD designs are, and will, be released under the Apache 2.0 License. See LICENSE for details.
- Website: Mekion
- GitHub: github.com/mekion/the-bimo-project
- Discord: Join Community
- Issues & Discussions: GitHub Issues
For questions, partnerships, or press inquiries:
- Email: info@mekion.com
- X(Twitter): @mekionlabs
Built with ❤️ by Mykhaylo Ilyin. Making bipedal robots accessible.
