newton-physics/newton

An open-source, GPU-accelerated physics simulation engine built upon NVIDIA Warp, specifically targeting roboticists and simulation researchers.

What it solves

Newton provides a GPU-accelerated physics simulation engine designed for roboticists and simulation researchers. It addresses the need for scalable, differentiable, and extensible simulations that can handle complex robotic systems, cables, and cloth.

How it works

Built upon NVIDIA Warp, Newton generalizes the warp.sim module and uses MuJoCo Warp as its primary backend. It leverages GPU-based computation to achieve high performance and integrates with OpenUSD for scene description and support. The engine is designed to be differentiable, allowing for gradients to be used in optimization and learning tasks.

Who it’s for

It is primarily intended for roboticists and researchers specializing in simulation who require high-performance physics engines for rapid iteration and scalable testing.

Highlights

  • GPU-accelerated computation via NVIDIA Warp.
  • Integration with MuJoCo Warp as the primary backend.
  • Support for OpenUSD.
  • Differentiable simulation capabilities.
  • User-defined extensibility.
  • Broad support for various simulation types including rigid bodies (URDF), cables, and cloth.

Related

  • Project
    google-deepmind/mujoco_warp

    A GPU-accelerated version of the MuJoCo physics simulator designed for NVIDIA hardware to provide high-throughput, accurate simulation for robotics research.

  • Project
    google-deepmind/mujoco_playground

    A GPU-accelerated suite of robot learning environments built with MuJoCo MJX, designed for efficient reinforcement learning and sim-to-real research.

  • Project
    NVIDIA/warp

    A Python framework for GPU-accelerated, differentiable simulations used in robotics, physics, and machine learning.

  • Project
    isaac-sim/IsaacLab

    A GPU-accelerated open-source framework built on NVIDIA Isaac Sim for simplifying robotics research in reinforcement learning, imitation learning, and motion planning.

  • Project
    google-deepmind/mujoco

    A high-performance physics engine for the fast and accurate simulation of articulated structures, widely used in robotics and machine learning research.