|
Take heed to this text |
Construction professionals leverage Apple’s innovative AR technology to remotely control robots, streamlining the process of gathering demonstration footage. Supply: NVIDIA
As part of the prestigious SIGGRAPH conference being held in Denver, NVIDIA Corporation is taking center stage. Explored in-depth analysis and decision-making frameworks for simulation, generative synthetic intelligence, and robotics applications. The corporation is providing a comprehensive range of companies, fashion options, and computing platforms to empower robotics and AI developers to design, build, and deploy the next generation of humanoid robotics.
Jensen Huang, founder and CEO of NVIDIA, notes that the next major wave of AI advancements will be in robotics, with the development of humanoid robots being a particularly exciting breakthrough. We’re significantly accelerating the development of the entire NVIDIA robotics stack, making it accessible to a global community of humanoid robotics developers and companies, empowering them to leverage the power of the platforms, acceleration libraries, and AI models tailored to their unique needs.
Is there something exciting at SIGGRAPH? Rev Labaredian, vice chairman of Omniverse, underscored NVIDIA’s long-standing commitment to virtualization techniques during a press conference, highlighting the company’s extensive experience in graphics analysis dating back to 2001. This established foundation has allowed for seamless connections between graphics, simulation, and robotics.
NVIDIA’s NIMs empower developers to create photorealistic digital twins.
Accordingly, NVIDIA asserts that its technology is capable of aiding in the design of both robots and their environments, as well as being utilised for training manufacturing programmes. Its Nim microservices are prebuilt containers utilizing NVIDIA’s inference software. The corporation plans to significantly reduce the frequency of deploying new mannequins, shortening the time frame from weeks to mere minutes.
Kari Briski, vice chair of generative AI software product management at NVIDIA, concedes that the time to leverage generative AI is indeed now, despite the potential for initial intimidation. Companies seek expedited routes to production to realize returns on their investments. This prompted NVIDIA’s NIMs to standardize the deployment of AI frameworks, largely reliant on CUDA to dominate the field.
“By introducing two novel AI microservices, researchers can significantly enhance their simulation workflows within RoboSim, a cutting-edge platform for robotics simulation, and unlock new possibilities for innovation.”
The MimicGen NIM produces synthetic motion data primarily through recordings from spatial computing devices, such as Apple Vision Pro.
The Robocasa NIM produces realistic robotic tasks and simulation-ready settings within NVIDIA’s Common Scene Description framework, enabling seamless collaboration and growth within immersive 3D environments.
The corporation also introduced innovations to facilitate the growth of data-intensive applications, alongside USD connectors allowing customers to stream massive ray-traced datasets to Apple Vision Pro.
The USD Search API allows for text-based searching of 3D assets when working with OpenUSD. Supply: NVIDIA
OSMO enables seamless cloud-based orchestration of robots
Here: Out there, NVIDIA offers OSMO, a managed cloud service enabling robotics developers to manage and scale complex workflows across distributed computing resources, both on-premises and in the cloud, seamlessly.
“NVIDIA reports that OSMO significantly streamlines robotic coaching and simulation processes, reducing the time it takes to deploy and scale new initiatives from months to just a few weeks.” Customers can visualise and interact with a diverse range of tasks, including generating synthetic data, training models, performing reinforcement learning, and conducting software-in-the-loop testing at scale for autonomous cellular robots and manipulators.
NVIDIA Workflows Seamlessly Fuse Human Insight with AI-Powered Intelligence.
Based on coaching bases for humanoid and different robots, typically, large amounts of data are required, as highlighted by renowned technology company NVIDIA. One potential approach to harness human demonstration data is, however, it may be expensive and labor-intensive, as noted.
NVIDIA has unveiled a groundbreaking workflow that leverages the power of artificial intelligence (AI) and its Omniverse platform, empowering builders to train robots using significantly less data than previously necessary. Initially, construction professionals employ Apple’s Xcode Pro to capture a limited number of tele-operated test scenarios.
In a subsequent step, they employ Isaac Sim’s simulation capabilities to replicate the recordings, subsequently utilizing MimicGen NIM to create synthetic datasets derived from these simulated records. NVIDIA reports that builders can hone their skills on a humanoid robotic simulator using both real-world data and synthetic information, thus reducing costs and expediting development.
Within the framework of robotic studies, roboticists can leverage Robocasa NIM to create experiences that enable retraining of their robotic model. Throughout the entire workflow, NVIDIA claimed that OSMO is capable of assigning computational tasks to entirely distinct sources, thereby obviating weeks’ worth of administrative responsibilities.
NVIDIA’s renowned expertise in AI is expected to aid the development of synthetic data used to train and refine its general-purpose computing platform.
“For Fourier’s CEO Alex Gu, developing humanoid robots poses a daunting challenge – one that demands an enormous volume of accurate data meticulously gathered from the real world.” “NVIDIA’s innovative simulation and generative AI development tools are poised to significantly accelerate our model-building processes, streamlining the workflow from inception to delivery.”
.
Humanoid corporations take off with a head start.
NVIDIA’s latest offering provides early access to NVIDIA Isaac, including NIMs and OSMO, as well as its most recent software releases: Omniverse, Isaac Lab, Jetson Thor, and the Challenge platform.
The corporate has announced that several leading robotics builders, including ByteDance Analytics, Fourier, Galbot, LimX Dynamics, Mentee Robotics, RobotEra, and others, have already joined its early-access program.
“NVIDIA and Boston Dynamics share a longstanding history of collaborative innovation, driving advancements in robotics with every partnership.” “We’re thrilled to witness the tangible results of our efforts drive significant growth in trade, and the early-access program is a pioneering strategy for seamless access to top-tier expertise.”
At SIGGRAPH, NVIDIA’s CEO Jensen Huang engaged in a compelling conversation with Lauren Goode, senior author at Wired, exploring the transformative impact of robotics and AI on industrial digitalization. He has even met with Meta’s Mark Zuckerberg.
