newton-physics/newton
Newton
Newton is a GPU-accelerated physics engine and robotics simulation platform designed for high-performance modeling of rigid bodies and complex articulations. It functions as a differentiable physics engine, calculating gradients to enable mathematical optimization and machine learning.
The platform is distinguished by its ability to execute multiple parallel physics worlds on a single GPU, which accelerates data collection for reinforcement learning. It also supports the simulation of deformable bodies, such as cloth and cables, using particle-based methods and multi-physics coupling.
Newton covers a broad range of robotic capabilities, including inverse kinematics for joint parameter calculation and the generation of synthetic sensor data for IMUs and contact sensors. The system includes tools for importing robot models and asset meshes, as well as exporting simulation states to universal 3D formats for visualization and analysis.
Features
- Differentiable Physics Engines - Functions as a differentiable physics engine that calculates gradients for mathematical optimization and machine learning.
- GPU-Accelerated Robot Simulators - Provides a GPU-accelerated simulation environment for modeling rigid bodies and complex robotic articulations.
- Parallel Simulations - Executes multiple independent physics worlds simultaneously on the GPU to accelerate reinforcement learning data collection.
- Robotics Simulators - Provides a complete virtual environment for modeling robotic joints, kinematics, and synthetic sensor data.
- Robotic Physics and Sensor Simulators - Simulates rigid bodies, articulated systems, and sensors specifically for realistic robotic interaction.
- Reinforcement Learning Accelerations - Accelerates reinforcement learning by executing multiple parallel physics worlds on a single GPU for rapid data collection.
- Robotic Sensor Simulation - Simulates the functional output of robotic perception hardware like IMUs and contact sensors.
- Inverse Kinematics - Computes the required joint configurations for robotic arms to reach specific end-effector poses.
- GPU-Accelerated Physics Simulations - Implements high-throughput physical simulations using parallel compute kernels on the GPU.
- Inverse Kinematics - Computes required joint parameters for articulated robots to reach specific spatial targets.
- Parallel World Simulations - Runs multiple independent physics worlds simultaneously on a single GPU to accelerate RL data collection.
- Synthetic Sensor Simulation - Generates artificial sensor readings by calculating intersections and contact points within the physics engine.
- Simulation Visualizers - Renders the physical world and dynamic properties for real-time observation or offline analysis.
- Robot Model Importers - Ingests robot models from standard descriptive formats into the simulation environment.
- Simulation Environment Builders - Builds simulation environments by adding geometric shapes, ground planes, and predefined templates.
- Soft Body Simulations - Models deformable objects such as cloth and cables using particle-based dynamics and multi-physics coupling.
- Particle Physics Simulations - Simulates soft bodies and flexible materials like cloth and cables using particle-based dynamics.
- Physics Simulation - GPU-accelerated differentiable physics engine for robotics.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
