Menlo open-sources Asimov 1 humanoid locomotion policy and Isaac Lab training code

freelerobot · x · 2026-09-25

Menlo Research open-sourced the locomotion policy and full training stack of its Asimov 1 humanoid robot, including a trained checkpoint, an Isaac Lab-based RL framework using PPO and adversarial motion priors, reward definitions, hardware-tuned actuator configs, and domain randomization settings—giving developers a foundation to adapt to their own hardware and experiments.

Original post →

More from Embodied

Embodied channel →