Asimov open-sources humanoid locomotion policy; non-technical user fine-tunes fall recovery on a single 4090

freelerobot · x · 2026-09-27

Robotics startup Asimov has open-sourced the locomotion policy checkpoint for its humanoid Asimov 1, along with Isaac Lab-based training code covering reward definitions, hardware-tuned actuator configurations, and domain randomization settings, giving developers a foundation to adapt and extend.

Quoting the release, @eckartal — a self-described non-technical person — shows how low the barrier has become: starting from the base policy and training code, he is training a fall recovery policy on a single RTX 4090.

Original post →

More from Embodied

Embodied channel →