NVIDIA Isaac for Healthcare v0.4 enables end-to-end sim-to-real SO-ARM surgical assistant workflow

AI Impact Summary

NVIDIA Isaac for Healthcare v0.4 introduces an end-to-end sim-to-real workflow (SO-ARM Starter Workflow) that unifies data collection, policy training, and hardware deployment for autonomous surgical assistance. The pipeline relies on a mixed data strategy—roughly 70 simulation episodes with 10-20 real-world demonstrations—where over 93% of training data are synthetic, enabling rapid iteration while validating on LeRobot and WOWROBO vision sensors. Key components include GR00T N1.5 fine-tuning, RTI DDS-based policy deployment, and deployment on DGX Spark hardware, with SO-ARM101 hardware supporting both follower and leader roles.