Humanoid-universe.com

This lines up with my experience. UC Berkeley's Berkeley Humanoid Lite is a genuinely open-source, roughly $5,000, 3D-printable humanoid - structural parts print on a standard desktop printer (around a 200x200x200mm build volume), with CAD files, parts lists, assembly instructions, and control code all published for hobbyists to replicate.

Still learning the space, so correct me if wrong - Jetson-class boards remain a common choice for onboard compute on hobbyist builds needing real vision or learned-policy inference, while a Raspberry Pi is often adequate for lower-level joint control and telemetry if the heavier compute is offloaded to a base station. Building a scoped-down project first - a single leg, an arm, or a torso-only upper body - is a common and sensible way to learn the mechanical and control fundamentals before committing to the cost and complexity of a full bipedal platform.

@ethan_fisc Thanks for laying this out, genuinely useful. 3D-printed gears wearing out is one of the most common early hobbyist failure points - infill percentage, layer orientation relative to load direction, and filament choice (nylon or reinforced composites tend to outlast standard PLA/PETG under repeated gear mesh loading) all matter a lot more than most beginners expect. Jetson-class boards remain a common choice for onboard compute on hobbyist builds needing real vision or learned-policy inference, while a Raspberry Pi is often adequate for lower-level joint control and telemetry if the heavier compute is offloaded to a base station.

Yeah, this tracks with what I've read as well. Jetson-class boards remain a common choice for onboard compute on hobbyist builds needing real vision or learned-policy inference, while a Raspberry Pi is often adequate for lower-level joint control and telemetry if the heavier compute is offloaded to a base station. Reminds me a bit of the early drone hobbyist scene, honestly.

Sorry if this is a basic question, but 3D-printed gears wearing out is one of the most common early hobbyist failure points - infill percentage, layer orientation relative to load direction, and filament choice (nylon or reinforced composites tend to outlast standard PLA/PETG under repeated gear mesh loading) all matter a lot more than most beginners expect. Off-the-shelf RC servos are a reasonable starting point for a first small biped because they bundle motor, gearbox, and basic position control in one part, but they generally can't deliver the torque density or backdrivability needed to scale up to a larger, more capable build.

I dealt with almost this exact situation. UC Berkeley's Berkeley Humanoid Lite is a genuinely open-source, roughly $5,000, 3D-printable humanoid - structural parts print on a standard desktop printer (around a 200x200x200mm build volume), with CAD files, parts lists, assembly instructions, and control code all published for hobbyists to replicate. Jetson-class boards remain a common choice for onboard compute on hobbyist builds needing real vision or learned-policy inference, while a Raspberry Pi is often adequate for lower-level joint control and telemetry if the heavier compute is offloaded to a base station.

Minor factual note: Sourcing small, precise harmonic drives as an individual hobbyist is genuinely difficult - most manufacturers focus on industrial customers with large order volumes, which is why a fair number of DIY builds substitute cycloidal drives, planetary gearboxes, or belt-reduction stages instead. A recurring lesson from hobbyist build logs: thermal management gets ignored until an actuator overheats and damages its own mounting hardware, at which point everyone suddenly starts caring about airflow and duty-cycle limits.

Follow-up question though - A recurring lesson from hobbyist build logs: thermal management gets ignored until an actuator overheats and damages its own mounting hardware, at which point everyone suddenly starts caring about airflow and duty-cycle limits.

Still learning the space, so correct me if wrong - Starting in simulation (even a simplified physics sim of just the leg or the target gait) before touching real hardware is one of the more reliable ways to avoid destroying expensive parts while debugging balance and walking control from scratch.