news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

IT Home reported on August 6 that NVIDIA today announced a new workflow that uses artificial intelligence and Omniverse to accelerate the development of humanoid robots. This workflow significantly reduces the amount of data required to train humanoid robots by combining Apple Vision Pro with NVIDIA's AI and simulation technology.

According to IT Home, traditionally, training humanoid robots requires a large amount of manual demonstration data, which is usually an expensive and time-consuming process.Nvidia’s new approach collects a small amount of human demonstration data via teleoperation, then generates a large amount of synthetic data using its Isaac Sim simulator and MimicGen NIM microservices.These synthetic data, along with real data, are used to train NVIDIA’s Project GR00T humanoid base model, saving time and cost.

In addition, NVIDIA's Robocasa NIM microservices and Isaac Lab platform can also be used to generate robot learning experiences and further optimize robot models. NVIDIA's OSMO cloud platform is responsible for efficiently managing the allocation of computing resources throughout the development process, further improving development efficiency.