NVIDIA Open Sources GPU-Native Medical Physics Simulator for Healthcare Robots

12:00 · July 29, 2026

Healthcare robotics has long been held back by a shortage of real-world training data, since annotated surgical demonstrations are expensive to collect, many clinically important scenarios are rare by definition, and prototyping on physical hardware is slow. NVIDIA’s new Medical Physics Simulation framework, built inside its existing Isaac for Healthcare platform, tries to close that gap with GPU-native, modular simulation environments, including dedicated Endoluminal and Surgical Simulation Modules that model the physical interaction between a device and human anatomy in real time. The framework is built on NVIDIA’s Warp and Newton Physics libraries alongside CUDA, and it is designed to unify physics simulation with imaging pipelines so a single environment can generate both the mechanical behavior a robot needs to learn from and the visual data used to train its perception systems. NVIDIA says the approach lets 8,192 robot-training environments run in parallel, cutting training time for certain policies from more than five hours down to under two minutes. Several surgical robotics companies, including CMR Surgical, Johnson & Johnson MedTech, Medtronic, XCath, and Inner Logic, are already using the framework to build surgical digital twins, train endovascular navigation policies, and generate synthetic data for regulatory submissions.