Welcome Here  Shenzhen Mingjiada Electronics Co., Ltd.

On 15 August, NVIDIA unveiled an innovative workflow that deeply integrates artificial intelligence technology with its Omniverse platform to accelerate the development of humanoid robots. By combining the advanced capabilities of Apple's Vision Pro with NVIDIA's AI and simulation technologies, this process significantly reduces the amount of data required to train humanoid robots, solving the costly and time-consuming problem of data acquisition in traditional methods.

In the traditional path, the training of humanoid robots is highly dependent on a large amount of manual demonstration data, which is not only economically costly, but also inefficient. NVIDIA's new strategy here cleverly uses teleoperation technology to collect limited human demonstration data as a starting point, and then relies on its powerful Isaac Sim simulator and innovative MimicGen NIM microservices to generate massive amounts of high-quality synthetic data. This synthetic data complements the real data, and together they drive the training process of NVIDIA's Project GR00T humanoid base model, achieving a quantum leap in training efficiency.

Taking it a step further, NVIDIA also provides Robocasa NIM microservices and the Isaac Lab platform, tools that generate rich robotics learning experiences and offer unlimited possibilities for model optimisation. NVIDIA's OSMO cloud platform acts as the ‘commander’ of the entire development process, efficiently managing and optimising the allocation of computing resources to ensure a smooth and efficient development process.

In summary, the launch of NVIDIA's innovative workflow not only opens up a new path for the development of humanoid robots, but also signals that future robotics technology will take more solid steps on the road to intelligence and efficiency.