Simulate before you build: our sim-to-real workflow
Hardware time is the most expensive time in robotics. Here is how simulation-first engineering keeps projects fast and predictable.
Read articleOpen-source robot arms now cover everything from classroom kits to research hardware. Here is how we think about the three we use most.
MeArm: learning the basics
MeArm is a classic 4-DOF desktop arm. It comes with complete laser-cut and 3D-print files and runs on Arduino or Raspberry Pi.
- Choose it for: STEM classes, first robotics projects, kinematics demos
- Skip it if: you need repeatable precision or want to train AI policies
SO-ARM100: low-cost data collection
SO-ARM100 is currently the first choice for low-cost AI arms. It is 3D-printed and driven by serial bus servos, and you can build a leader–follower pair for a few hundred dollars.
You move the leader arm by hand, the follower copies it, and every movement is recorded. Those demonstrations train imitation-learning policies with LeRobot.
- Choose it for: teleoperation datasets, imitation learning, affordable lab setups
- Skip it if: your task needs high payload or force-sensitive contact
OpenArm: physical-AI research
OpenArm is a fully open humanoid arm built for contact-rich interaction. It ships with URDF models, a CAN-bus communication library, bilateral teleoperation, and reinforcement-learning interfaces for Genesis and MuJoCo.
- Choose it for: force-aware manipulation, bilateral teleoperation, sim-to-real research
- Skip it if: budget or build time is tight
Our rule of thumb
Start with the cheapest arm that can collect the data your task needs, then move up only when the data says so.
Not sure which fits your project? Ask us: we are happy to help you compare.