FANUC cobot takes plain-language commands, trained in Isaac Sim before hardware existed

lukas_m_ziegler · x · 2026-09-25

At AMB, a FANUC CRX cobot takes natural-language pick commands: vision-based object recognition plus NVIDIA's GR00T vision-language-action model plans the grasp (Cosmos used for some applications). The team developed against an Isaac Sim digital twin since hardware wasn't available, using the same simulation to prep fine-tuning data — so the robot learned its task before the physical arm existed, decoupling software progress from hardware availability.

Original post →

More from Embodied

Embodied channel →