NVIDIA's Newton Engine Enhances Robotics with Advanced Simulation Capabilities
Tony Kim Sep 29, 2025 14:32
NVIDIA Isaac Lab and Newton are transforming robotics by offering advanced simulation for quadruped locomotion and cloth manipulation, bridging the sim-to-real gap.
NVIDIA's latest developments in robotic simulation, particularly through the integration of its Newton engine with the Isaac Lab, are set to revolutionize the field by offering a robust and scalable solution for simulating complex physical interactions. According to the NVIDIA Developer Blog, this initiative aims to close the sim-to-real gap that often challenges robotics developers.
Newton: A Collaborative Endeavor
Newton, an open-source physics engine collaboratively developed by NVIDIA, Google DeepMind, and Disney Research, is managed by the Linux Foundation. Built on NVIDIA Warp and OpenUSD, Newton is designed to enhance robot learning by providing an extensible platform compatible with various robot learning frameworks, including MuJoCo Playground and Isaac Lab.
Innovative Training and Simulation Techniques
The integration of Newton within Isaac Lab facilitates the training of quadruped robots to navigate complex environments. The process involves setting up a multiphysics simulation to fold clothes using an industrial manipulator. By leveraging the Newton Solver API, the platform allows seamless integration with training environments, enabling robots to learn tasks with greater precision and efficiency.
Key Features and Performance Improvements
The Newton Beta release introduces significant performance enhancements. MuJoCo Warp, the primary Newton solver, is reported to be up to 152 times faster for locomotion tasks and 313 times faster for manipulation tasks on the GeForce RTX 4090. Additionally, the NVIDIA RTX PRO 6000 Blackwell Series offers substantial speed improvements, making Newton a powerful tool for real-time robotic simulations.
Adoption and Ecosystem Expansion
The adoption of Newton is growing rapidly among leading universities and companies. ETH Zurich's Robotic Systems Lab employs Newton for simulating earthmoving applications, while Peking University integrates it into tactile sensing domains. These collaborations highlight Newton's potential in advancing robotic learning and reducing the sim-to-real gap.
Overall, NVIDIA's Newton engine, with its advanced capabilities and collaborative development, marks a significant step forward in robotic simulation and training, paving the way for more realistic and effective robotic applications.
Image source: Shutterstock.jpg)