Humanoid-universe.com

<t>Slight correction, though the overall point stands:

A minimum viable sensing suite for safe bipedal walking generally includes joint encoders, an IMU for orientation/angular velocity, and either force/torque sensing or accurate current-based torque estimation at the ankles - everything else ...

<t>@lbianchi +1 to this. Worth adding:

Multi-camera calibration drifts over time from thermal expansion, vibration, and mechanical wear, which is why production systems typically run periodic recalibration routines rather than assuming a one-time factory calibration holds forever.

Reminds me a ...

<t>@rossi30 Pretty much this. One thing to add:

ROS2 remains common in research and early-stage products for its tooling and ecosystem, but a number of production humanoid companies run custom, more tightly-optimized middleware for their real-time control loops, using ROS2-like tooling mainly for ...

<t>@charlesbianchi Slight correction, though the overall point stands:

Internship competitiveness at the well-known humanoid companies has increased sharply as the field's visibility has grown, with the applicant pool now including not just robotics students but a lot of general CS/ML students ...

<t>@joseph.robinson I'd push back on this a bit.

LiDAR gives reliable, lighting-independent range data but is heavier, pricier, and gives sparser point clouds up close than stereo or depth cameras, which is why a lot of humanoids lean on stereo/depth cameras for near-field manipulation and ...

<t>@elarsen69 Follow-up question though -

Bimanual manipulation - two arms coordinating on one task - is harder than it looks mostly because of the added degrees of freedom and the timing/force coordination required; a lot of 'two-handed' demos are actually closer to two independent single-hand ...

@barbara.jones Speaking from personal experience here, Idle/standing power draw is often surprisingly close to a meaningful fraction of active walking power draw once you account for onboard compute, sensors, and balance-holding torque - 'doing nothing' still costs real energy on a humanoid.

<t>Same conclusion I've come to. Also worth noting:

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 ...

<t>Thanks for laying this out, genuinely useful.

For dynamic walking and recovery, joint control bandwidth in the tens of Hz range is typically necessary to react to a stumble before the center of mass gets too far outside the support polygon. Cycloidal drives are a less common but real ...

<t>Here's what I know on this:

Isaac Lab (the successor to Isaac Gym) is widely used for large-scale parallel RL training thanks to GPU-accelerated physics, while MuJoCo is often used as a secondary 'sim-to-sim' validation step because its contact dynamics are generally considered more realistic ...