Nvidia Announces New Workflow to Accelerate Humanoid Robot Development

TapTechNews August 6th news, 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's VisionPro with Nvidia's AI and simulation technologies.

According to TapTechNews, traditionally, training humanoid robots requires a large amount of manual demonstration data, which is usually an expensive and time-consuming process. Nvidia's new method collects a small amount of human demonstration data through remote operation, and then uses its IsaacSim simulator and MimicGenNIM microservice to generate a large amount of synthetic data. These synthetic data, together with real data, are used to train Nvidia's ProjectGR00T humanoid basic model, thereby saving time and cost.

In addition, Nvidia's RobocasaNIM microservice and IsaacLab platform can also be used to generate robot learning experiences to further optimize the robot model. Nvidia's OSMO cloud platform is responsible for efficiently managing the allocation of computing resources in the entire development process to further improve development efficiency.