Humanoid-universe.com

<t>Small correction on one detail:

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 reserve LiDAR (if ...

<t>This raises a question for me -

Proprioception (the robot's sense of its own joint angles, velocities, and forces) tends to get less attention than flashy vision systems, even though a lot of balance and manipulation failures trace back to proprioceptive noise or miscalibration rather than a ...

@erik_novi One nitpick - Force/torque sensors near the ankle give a direct read on ground reaction forces, which is valuable for balance control, but they add cost, a failure point, and routing complexity right at a joint that already takes the most mechanical abuse.

<t>One nitpick -

Cross-embodiment training (training one policy across data from multiple different robot bodies) has shown some real transfer benefits for high-level behaviors, but low-level control (exact joint torques, timing) still tends to need embodiment-specific fine-tuning. A lot of what ...

<t>@dubois35 Just to be precise about one thing:

Physical Intelligence's pi0 pairs a smaller pretrained vision-language backbone with a separate flow-matching 'action expert' module, which is one way to get fast, high-frequency action output without needing the whole giant language model to run ...

<t>@mia_lars Not to derail, but this reminds me of something adjacent:

Tendon-driven fingers let you move the heavier actuators back into the palm or forearm, keeping the fingers themselves light and fast, but they introduce cable routing, tensioning, and long-term wear problems that direct ...

I see it a little differently. Distributed power architectures (multiple smaller packs or local capacitor buffering near high-draw actuators) can reduce peak current demands on the main bus and improve fault isolation, at the cost of added complexity versus a single central pack.

<t>From what I've seen:

Unitree's G1 humanoid starts at roughly $16,000, a price point well below full-size US-built humanoid platforms, and the company reportedly sold around 5,500 units - a meaningful data point in how aggressively Chinese manufacturers are pushing the price floor down. AgiBot ...

<t>Short answer:

Group-buy arrangements for harder-to-source parts (small harmonic drives, precision bearings, custom PCBs) are a common way hobbyist communities work around the fact that specialty components are often only sold in industrial minimum-order quantities.

Kind of makes me think ...

@barbara50 One nitpick - RSS (Robotics: Science and Systems), ICRA, and the IEEE-RAS Humanoids conference are commonly cited as the most directly relevant venues for someone specifically interested in legged locomotion and humanoid control research, as opposed to broader AI/ML conferences.