NVIDIA's Newton Physics Engine and Robotics Innovations Unveiled
At the recent GTC 2025 conference, NVIDIA CEO Jensen Huang unveiled "Newton," an open-source physics engine developed in collaboration with Google DeepMind and Disney Research. Newton is designed to enhance robotic simulation by providing accurate virtual representations of robot behavior and interactions within realistic environments, thereby facilitating the development of advanced robotic control algorithms and designs.
During the keynote, Huang introduced "Blue," a robot developed in partnership with Disney Research and Google DeepMind. Blue demonstrated expressive movements and interactions, showcasing the capabilities of the Newton physics engine in creating lifelike robotic behaviors. The audience was captivated as Blue responded to Huang's questions with beeps, head nods, and body wiggles, demonstrating the potential of next-generation robotics.
Newton is built on NVIDIA's Warp framework, enabling developers to write GPU-accelerated programs for simulation, AI, robotics, and machine learning. It is compatible with learning frameworks such as MuJoCo Playground and NVIDIA Isaac Lab, allowing for significant performance gains in humanoid simulations and in-hand manipulation tasks. The engine supports differentiable physics, enabling the computation of gradients through simulations, which is essential for optimizing robotic system parameters.
In addition to Newton, NVIDIA announced Isaac GR00T N1, the world's first open, fully customizable foundation model for generalized humanoid reasoning and skills. This model features a dual-system architecture inspired by human cognition, with one system for fast-thinking actions and another for deliberate decision-making. Isaac GR00T N1 is trained on human demonstration data and synthetic data generated by the NVIDIA Omniverse platform, aiming to accelerate the development of humanoid robots across various industries.
During the conference, Huang also revealed NVIDIA's roadmap for next-generation AI chips, including the transition from Blackwell to Blackwell Ultra, expected later this year, and the upcoming Vera Rubin platform. The Vera Rubin platform, named after astronomer Vera Rubin, will feature a new CPU and networking architecture with twice the performance of Hopper and more memory.
These advancements reflect NVIDIA's commitment to driving innovation in robotics and AI, providing tools and frameworks that enable the creation of more capable and lifelike robots. As Huang noted, the future of humanoid robots replacing factory workers is not decades away but could become a reality in just a few years.
Comments
Post a Comment