Newton Beta: A New Era in Robot Physics Simulation
The laws of physics apply everywhere — and now, robots are learning them faster than ever.
A powerful new open-source robotics platform, Newton Beta, has been unveiled — co-developed by NVIDIA, Google DeepMind, and Disney Research | Walt Disney Imagineering, and managed under the Linux Foundation. This collaboration represents a major leap forward in the simulation and training infrastructure for embodied AI and robotics.
The Core of Newton Beta
Newton Beta is designed to bring together state-of-the-art physics simulation, reinforcement learning, and robotic control under one unified framework. With support for both NVIDIA Isaac Lab and MuJoCo Warp, the platform enables researchers and developers to simulate multiphysics environments — everything from robotic quadruped locomotion to complex material interactions — at warp speed.
By leveraging GPU-accelerated computation and deep reinforcement learning, Newton Beta allows for high-fidelity, real-time simulation that dramatically reduces the time it takes to train robots to understand and respond to the real world.
Collaboration at the Cutting Edge
This collaboration marks a rare convergence of expertise across academia, industry, and entertainment:
NVIDIA provides the GPU-accelerated simulation foundation through its Isaac Lab ecosystem.
Google DeepMind contributes its breakthroughs in AI learning algorithms and embodied intelligence.
Disney Research | Walt Disney Imagineering brings decades of experience in robotics, animatronics, and physics-based motion design.
The Linux Foundation ensures open access and community-driven governance — keeping Newton Beta transparent, extensible, and widely accessible.
From Quadrupeds to Multiphysics Simulation
Early demonstrations of Newton Beta highlight the training of quadruped robots — machines that can walk, run, and adapt dynamically to uneven terrain with physical accuracy that mirrors reality. The integration of MuJoCo Warp adds multiphysics support, meaning the same system can handle soft-body dynamics, fluid simulations, and contact-rich environments — a key step toward true general-purpose robot training.
Developers can already explore these capabilities through NVIDIA’s technical blog, which walks through the process of training quadrupeds and running advanced physics simulations in real time.
Why It Matters
Newton Beta is more than a robotics toolkit — it’s the infrastructure for the next generation of robot intelligence.
By merging AI learning, cinematic-grade simulation, and open-source collaboration, this project lays the groundwork for an era where robots not only move like humans — they understand the physics that govern our world.
Learn more about Newton Beta and explore how to train a quadruped robot using Isaac Lab and MuJoCo Warp
