Embodied AI Glossary中文NVIDIA OSMO
AdvancedNVIDIA's cloud orchestration platform that schedules robotics and physical-AI training, simulation, and data generation across different compute.
OSMO is a cloud-native orchestration platform from NVIDIA for robotics and physical-AI development. A robotics project usually chains together many steps — synthetic data generation, model training, simulation-based evaluation, and software-in-the-loop/hardware-in-the-loop testing — often running on different machines: cloud GPU clusters, local workstations, or edge devices like Jetson. OSMO lets developers describe this pipeline as a configuration file, and the platform handles allocating compute, queuing jobs, and tracking data, removing the need to manually move data and coordinate jobs across machines by hand. It's one piece of NVIDIA's “three computers” strategy, the layer that connects training and simulation, and is commonly used alongside Isaac Sim, Isaac Lab, and Cosmos.
ExampleTurn “Isaac Sim generates synthetic data → train a grasping policy → Isaac Lab runs batch evaluation” into a single OSMO workflow, submit it once, and the platform automatically distributes the jobs across different GPU nodes.
- Also called
- OSMO
- Related
- NVIDIA Three-Computer Solution · NVIDIA Isaac Sim · NVIDIA Isaac Lab · NVIDIA Cosmos · Synthetic Data · Hardware-in-the-Loop / Software-in-the-Loop Simulation
- Sources
- NVIDIA OSMO 官方页面 (Chinese)
- As of
- 2025