Open Source Robotics

A curated collection of high-quality open source robotics projects, tools, and software to propel the robotics community forward.

Asimov v1

Open-source 1.2 m, 25-DoF humanoid you can fabricate, build and train yourself

License

CERN-OHL-S-2.0

Overview

Asimov v1 is a fully open-source humanoid robot by Asimov Inc (Menlo Research): a 1.2 m, 35 kg biped with 25 actuated degrees of freedom that you can fabricate, assemble, train and customize. The repository releases everything a self-source builder needs — mechanical CAD for all seven subassemblies, KiCad electrical design, wiring harness, a MuJoCo simulation model and onboard software — under CERN-OHL-S-2.0 (hardware) and GPL-2.0 (software), with a detailed public assembly manual and a DIY kit that the repository README says is shipping now for $20,000.

Asimov v1's mechanical design spans seven subassemblies with 167 STEP parts: 43 CNC-machined 7075 aluminium, 39 MJF PA12 nylon, 18 SLM 316L stainless and 67 off the shelf. The 25 actuated joints (six powered joints per leg with fixed flat feet, 5-DoF arms, waist yaw and a 2-DoF neck) use EC-A4310-class actuators on five CAN interfaces at 1 Mbps and one at 500 kbps, with squat, curl and raise load figures published (rated: squat 5 kg, bicep curl 5 kg per arm, front raise 5 kg per arm, lateral raise 6 kg per arm; peak: 15 kg, 15 kg and 18 kg per arm for the last three). Compute splits between a Raspberry Pi 5 for media and networking and a Radxa CM5 on a custom KiCad motion-control board; the head carries a 2 MP camera and stereo microphone array, the torso a speaker and 6-DoF IMU. The public manual at docs.menlo.ai covers preparation (part organisation, battery, compute module, firmware and software flashing, actuator IDs, electronics disassembly), assembly in the order legs, pelvis, arms, torso with head, then Finish Assembly, and an Operate chapter for calibration and first start, and the repo ships a MuJoCo MJCF model with per-link meshes for simulation. Self-source BOM cost is approximately $16k, dominated by actuators (~$7k) and chassis fabrication (~$7k).

Why it matters

Asimov v1 is one of the most complete full-size open humanoid releases to date: not a teaser repo but the actual CNC drawings, metal-print files, PCB sources, sim model and step-by-step manual — at a $15-16k price point that puts a trainable biped within reach of labs and serious builders, in the middle of the humanoid boom.

Problem it solves

Capable humanoids are either closed commercial products or partial open releases missing the mechanical or electrical sources. Asimov v1 publishes the entire stack — fabricatable CAD, electronics, simulation and manual — so builders can self-source or kit-build a real 25-DoF biped and own every layer of it.

Key takeaways

  • 1.2 m / 35 kg biped, 25 actuated DoF: six powered joints per leg, 5-DoF arms, waist, 2-DoF neck
  • Complete fabrication release: 167 STEP parts across CNC 7075, MJF PA12 and SLM 316L classes
  • Custom KiCad motion-control board (Radxa CM5) + Raspberry Pi 5, five 1 Mbps and one 500 kbps CAN interfaces
  • MuJoCo MJCF simulation model with per-link meshes included in-repo
  • Public step-by-step manual: preparation, module assembly, verification
  • ~$16k self-source BOM; DIY kit shipping at $20,000; CERN-OHL-S hardware license

Who it is for

  • Humanoid robotics labs and researchers
  • Advanced builders self-sourcing a full biped
  • Robot-learning teams wanting a trainable open platform
  • Hardware engineers studying humanoid design

Limitations

  • Serious fabrication requirements: CNC aluminium and SLM 316L metal printing must be outsourced by most builders
  • ~$16k self-source cost, dominated by actuators and chassis fabrication
  • Locomotion policy, Asimov API and mobile app are still on the roadmap (not yet released)
  • Hands are not included in the platform or DIY kit
  • The manual has no bill of materials page, and its own banner says its mechanical tour from Structural Architecture onward is being updated. Its Calibrate Actuators page says joint-by-joint home-pose angles are not yet specified.

This website requires JavaScript to function properly. Please enable JavaScript to explore our collection of open source robotics projects.