Building Hospital Automation Using NVIDIA Isaac for Healthcare

In this talk

1. 0:02

   Prepare the hospital for automation

   Hospitals face ongoing staffing shortages, and robots could help doctors and nurses expand capacity and improve outcomes. Project Rheo aims to make hospital environments suitable for developing and training automation.

2. 0:26

   Build a simulation-ready hospital

   NuRec and Isaac Sim create a physics-accurate digital twin of a hospital. Robot skills are gathered through teleoperation and real-world data.

3. 0:38

   Expand training data and vary conditions

   MimicGen scales the collected examples into a much larger training set. Cosmos Transfer varies layouts, lighting, and object placement, while photorealistic domain transfer is used to align simulation with real environments.

4. 0:56

   Train and validate robot behavior

   Isaac Lab uses imitation and reinforcement learning with GR00T to train robots to carry skills into real settings. Thousands of parallel experiments optimize precision, timing, and safety, with simulated edge cases and environmental changes tested before policies move to hardware.

5. 1:13

   Turn hospital experience into deployed skills

   Project Rheo is being used in smart hospital environments with PeritasAI, Light Wheel, and AdventHealth. The approach uses hospital experience to develop intelligence that can be deployed on robots.