Seeed-Projects/reBot-DevArm
Open Source Robotic Arm for All Developers
What it solves
reBot-DevArm provides an open-source, full-stack robotic arm platform designed to lower the barrier to entry for developers learning Embodied AI. It eliminates the need for proprietary hardware and software, providing everything from 3D design files and BOMs to high-level AI frameworks.
How it works
The project offers two hardware models—the B601-DM and B601-RS—which are 6-DOF robotic arms with grippers. The software stack is designed for compatibility with mainstream robotics ecosystems, including:
- Control & Kinematics: Python SDK, ROS1/2, and Pinocchio for forward/inverse kinematics and gravity compensation.
- Simulation: NVIDIA Isaac Sim and MuJoCo digital twins for sim-to-real transfer.
- AI Training: Integration with Hugging Face's LeRobot framework for training embodied agents.
- Perception: Support for depth cameras (e.g., Orbbec Gemini) and YOLO for visual grasping.
- Interaction: Voice-driven control via reSpeaker Flex 4-mic arrays.
Who it’s for
Developers, researchers, and students interested in Embodied AI, robotics, and sim-to-real training who want a fully transparent, open-source hardware and software stack.
Highlights
- True Open Source: Includes all sheet-metal and 3D-printed part files, detailed BOMs, and source code.
- Multi-Framework Support: Compatible with ROS2, LeRobot, Pinocchio, and Isaac Sim.
- Embodied Agent Architecture: Supports natural-language commands for autonomous planning and grasping.
- Sim-to-Real Pipeline: Provides digital twins and simulation environments to facilitate training in simulation and deploying to physical hardware.
- Flexible Hardware: Offers two models with different payload capacities (1.5kg for DM, 2.5kg for RS) and various kit options from pre-assembled to DIY.
Related
- Project
- Project
- Dispatch
- Project
- Project