With the booming development of the humanoid robot market, it is expected to reach a scale of US$38 billion in the future. In order to accelerate the development of the next generation of humanoid robots, NVIDIA released a series of important tools and platforms at the CES exhibition, aiming to solve the data acquisition and training problems faced in the development of humanoid robots. These tools will greatly promote technological progress in the field of robotics and pave the way for future intelligent robot applications.
In the future, the humanoid robot market is expected to reach US$38 billion. To meet this huge market demand, especially in the industrial and manufacturing fields, NVIDIA recently announced the launch of a series of basic robot models, data pipelines and simulation frameworks to accelerate the development of next-generation humanoid robots. Nvidia founder and CEO Jensen Huang announced the Isaac GR00T blueprint at the CES show, which is dedicated to generating synthetic motion to help developers use imitation learning to generate large amounts of synthetic motion data to train humanoid robots. Imitation learning is a subset of robotic learning that enables humanoid robots to acquire new skills by observing and imitating expert human demonstrations. Collecting these high-quality datasets is time-consuming and expensive in the real world, but with the Isaac GR00T blueprint, developers can easily generate large synthetic datasets with a small amount of human demonstration. Developers can use Apple Vision Pro to capture human motion in digital twins, starting with the GR00T-Teleop workflow. Next, these human demonstrations are extended into larger synthetic motion datasets via the GR00T-Mimic workflow, and finally, the GR00T-Gen workflow is further extended with domain randomization and 3D upscaling, supported by the NVIDIA Omniverse and Cosmos platforms. data set. These data sets can be used as input into robot strategies, helping robots interact with their environment more efficiently and safely. In addition, NVIDIA also released the Cosmos platform at CES, which contains a series of open pre-trained world basic models, designed to generate physical consciousness videos and world states, and promote the development of physical artificial intelligence. The models were trained on 18 trillion units of data, covering 2 million hours of autonomous driving, robotics and drone video. Cosmos not only generates large-scale data sets, but also bridges the gap between simulation and reality through 3D-to-realistic upscaling of images. Taken together, Nvidia's Isaac GR00T, Omniverse and Cosmos are driving innovation in physical artificial intelligence and humanoid robots. Many robotics companies, including Boston Dynamics and Figure, have begun using the Isaac GR00T and achieved good results. Humanoid software, hardware and robot manufacturers can apply for early access to NVIDIA’s Humanoid Robots Developer Program.
This series of NVIDIA initiatives marks a new stage in the development of humanoid robots, which will greatly promote the integration and development of artificial intelligence and robotics and create a more intelligent future for us. In the future, more innovative technologies and applications are expected to emerge, which is worth looking forward to.