NVIDIA's robot simulator and the robot-learning framework that runs on it, used to train and test robot control policies in simulation before they run on real robots.
AI Infrastructure
Simulators, models, runtimes and operations platforms that agents and autonomous machines are built on and run on.
v0.1 inclusion rule
Must be built for agents or autonomous machines and usable by third parties today, not general-purpose cloud, chips or MLOps with an AI label.
The open-source framework for building robot software: messaging between a robot's parts, shared libraries and tools, with long-term-support releases.
Open-source software for managing fleets of robots from different makers in one facility, connecting them to each other and to the facility's other systems.
NVIDIA's open-source SDK for real-time pipelines that take in sensor data, run AI on it and respond within the machine, used in medical devices, industrial inspection and edge computing.
An open-source library, dataset format and hub for teaching robots from demonstrations: record data, share datasets, then train and run policies on real robots.
Open robot foundation models (π0, π0-FAST and π0.5) that take camera images and an instruction and output robot actions, with code to fine-tune them on a robot maker's own data.