NVIDIA Robotics Research

Primary research and open systems connecting scalable simulation, training and physical deployment.

No.Paper · what it does and showsIndustry contribution

NVIDIA Isaac Sim: Enabling Scalable, GPU-Accelerated Simulation for Robotics2026 ↗

An independent survey by external (UNSW) authors, not an NVIDIA publication: it reviews Isaac Sim’s architecture, compares it with other simulators, and analyses representative usage across five domains plus practical limits.

Gives engineering teams an evidence-based map of where GPU simulation pays off, shortening simulator and digital-twin decisions before hardware investment.

Isaac Lab: A GPU-Accelerated Simulation Framework for Multi-Modal Robot Learning2025 ↗

NVIDIA’s successor to Isaac Gym: GPU-parallel physics, photorealistic rendering, actuator and multi-frequency sensor models, data collection and domain randomization unified in one composable framework for reinforcement and imitation learning at scale.

Became the de-facto training substrate for robot learning, letting teams train whole-body, locomotion and dexterous policies in hours of wall-clock instead of weeks.